A pipeline to connect GWAS Variant-to-Gene-to-Program (V2G2P) Approach
Help improve this workflow!
This workflow has been published but could be further improved with some additional meta data:- Keyword(s) in categories input, output
You can help improve this workflow by suggesting the addition or removal of keywords, suggest changes and report issues, or request to become a maintainer of the Workflow .
Variant-to-Gene-to-Program (V2G2P) Approach
Description
A pipeline to connect GWAS variants to genes to disease-associate gene programs. This pipeline uses snakemake and cNMF from Kotliar et al. The V2G2P approach could be applied to any GWAS studies with the correct cell type(s).
Overview of the V2G2P pipeline
The V2G2P approach has three components: V2G, G2P, and V2G2P enrichment test. Each one works as a stand-alone pipeline. Together these three components are essential for V2G2P. Below is an overview of the relations among the three steps:
The V2G pipeline is linked here . Author: Rosa Ma.
Details about G2P and V2G2P enrichment test
Below is a figure showing the different modules and features within the G2P pipeline:
Usage
Step 1: Clone this github repository
Step 2: Install conda environment
Install Snakemake and conda environment using conda:
bash
conda env create -f conda_env/cnmf_env.yml
conda env create -f conda_env/cnmf_analysis_R.yml
cnmf_env contains snakemake. If you do not have snakemake installed already, you can activate the environment via
conda activate cnmf_env
, then run the pipeline.
Step 3: Gather all input data
Necessary inputs:
Config file slots: | field | meaning | total workers | Number of processes to run in parallel | | seed | A number to set seed for reproducibility | | num runs | Number of NMF run (recommend 100 for the actual data analysis, and 10 for testing the pipeline)
Step 4: Run the pipeline
sh
conda activate cnmf_env
snakemake -n --configfile /path/to/config.json --quiet ## always recommend doing a dry run
Execute the workflow locally via
sh
snakemake --configfile /path/to/config.json
Please see the log.sh file in this github page for more examples.
For more snakemake usage and configuration, please visit snakemake documentation page.
Outputs
The output files are in the folders specified in analysisDir and figDir fields in the config file.
Analysis files (in analysisDir)
Summary output for choosing the number of components
Output can be found in config['analysisDir']/{cNMF_gene_selection}/{sampleName}/acrossK/
Outputs for each model (each choice of # components)
Output can be found in config['analysisDir']/{cNMF_gene_selection}/{sampleName}/K*/threshold_*/
Code Snippets
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/perturbationAnalysis.R \ --sampleName {wildcards.sample} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir}/ \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --cell.count.thr {wildcards.min_cell_per_guide} \ --guide.count.thr {wildcards.min_guide_per_ptb} \ --recompute F \ --motif.enhancer.background /oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv \ --motif.promoter.background /oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv \ ' " |
930 931 932 933 934 935 936 937 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/MAST_DE_Topics_scatter.R \ --barcode.names {params.barcode_names} \ --sampleName {wildcards.sample} \ --num.genes.per.MAST.runGroup {params.num_genes_per_MAST_runGroup} \ --scatteroutput {params.scatteroutput} ' " |
955 956 957 958 959 960 961 962 963 964 965 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/MAST_DE_Topics_preparation.R \ --barcode.names {params.barcode_names} \ --outdirsample {params.outdirsample} \ --scatteroutput {params.scatteroutput} \ --numCtrl {params.numCtrl} \ --sampleName {wildcards.sample} \ --K.val {wildcards.k} \ --density.thr {params.threshold} ' " |
984 985 986 987 988 989 990 991 992 993 994 995 996 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/MAST_DE_Topics_runGroups.R \ --barcode.names {params.barcode_names} \ --outdirsample {params.outdirsample} \ --scatteroutput {params.scatteroutput} \ --gene.group.list {input.MAST_gene_groups} \ --scatter.gene.group {wildcards.MAST_run_index} \ --sampleName {wildcards.sample} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --scriptdir {params.perturbAnalysis_scriptdir} ' " |
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/MAST_DE_Topics_gatherGroups.R \ --barcode.names {params.barcode_names} \ --outdirsample {params.outdirsample} \ --scatteroutput {params.scatteroutput} \ --total.scatter.gene.group {params.MAST_num_runs} \ --sampleName {wildcards.sample} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --scriptdir {params.perturbAnalysis_scriptdir} ' " |
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/aggregate_across_K_perturb-seq.R \ --figdir {params.figdir} \ --outdir {params.analysisdir} \ --datadir {params.datadir} \ --sampleName {wildcards.sample} \ --K.list {params.klist_comma} \ --K.table {params.K_spectra_threshold_table} ' " ## how to create this automatically? rule findK_plot_perturb_seq: input: toplot = os.path.join(config["analysisDir"], "{folder}/{sample}/acrossK/aggregated.outputs.findK.perturb-seq.RData") output: percent_batch_topics_plot = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/percent.batch.topics.pdf") params: time = "3:00:00", mem_gb = "64", figdir = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/"), analysisdir = os.path.join(config["analysisDir"], "{folder}/{sample}/acrossK/"), GO_threshold = 0.1, partition = "owners,normal" |
1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_findK_plots_perturb-seq.R \ --figdir {params.figdir} \ --outdir {params.analysisdir} \ --sampleName {wildcards.sample} \ --p.adj.threshold 0.1 \ --aggregated.data {input.toplot} \ ' " |
150 151 152 153 154 155 156 157 158 159 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/seurat_to_h5ad.R \ --sampleName {params.sampleName} \ --inputSeuratObject {input.seurat_object} \ --output_h5ad {output.h5ad_mtx} \ --output_gene_name_txt {output.gene_name_txt} \ --minUMIsPerCell {params.min_UMIs_per_cell} \ --minUniqueGenesPerCell {params.min_unique_genes_per_cell} ' " |
283 284 285 286 287 288 289 290 291 292 293 294 295 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}/{wildcards.sample}; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {wildcards.k} \ --n-iter {params.run_per_worker} \ --total-workers {params.run_per_worker} \ --seed {params.seed} \ --numgenes {wildcards.num_genes} ' " |
323 324 325 326 327 328 | shell: " bash -c 'source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.todir}/{wildcards.sample}; \ cp -r {params.fromdir}/{wildcards.sample}/cnmf_tmp {params.todir}/{wildcards.sample}/; \ cp {params.fromdir}/{wildcards.sample}/{wildcards.sample}.overdispersed_genes.txt {params.todir}/{wildcards.sample}/ ' " |
422 423 424 425 426 427 428 429 430 431 432 433 434 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}/{wildcards.sample}; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {wildcards.k} \ --n-iter {params.run_per_worker} \ --total-workers {params.total_workers} \ --seed {params.seed} \ --genes-file {input.genes} ' " |
461 462 463 464 465 466 | shell: " bash -c 'source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.todir}/{wildcards.sample}; \ cp -r {params.fromdir}/{wildcards.sample}/cnmf_tmp {params.todir}/{wildcards.sample}/; \ cp {params.fromdir}/{wildcards.sample}/{wildcards.sample}.overdispersed_genes.txt {params.todir}/{wildcards.sample}/ ' " |
530 531 532 533 534 535 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/cNMF/cnmf.py factorize \ --output-dir {params.outdir} \ --name {wildcards.sample} ' " |
554 555 556 557 558 559 560 561 562 563 | run: cmd = "mkdir -p " + os.path.join(params.inputdir, wildcards.sample, "cnmf_tmp") shell(cmd) for worker in range(params.num_workers): for run in range(params.run_per_worker): from_file = os.path.join(params.inputdir, "worker" + str(worker), wildcards.sample, "cnmf_tmp/" + wildcards.sample + ".spectra.k_" + wildcards.k + ".iter_" + str(run) + ".df.npz") index_here = worker * params.run_per_worker + run ### give the runs a new index to_file = os.path.join(params.outdir, wildcards.sample, "cnmf_tmp/" + wildcards.sample + ".spectra.k_" + wildcards.k + ".iter_" + str(index_here) + ".df.npz") cmd = "cp " + from_file + " " + to_file shell(cmd) |
588 589 590 591 592 593 594 595 596 597 598 599 600 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}/K{wildcards.k}/; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {wildcards.k} \ --n-iter {params.num_runs} \ --total-workers 1 \ --seed {params.seed} \ --numgenes {wildcards.num_genes} ' " |
627 628 629 630 631 632 633 634 635 636 637 638 639 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}/K{wildcards.k}/; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {wildcards.k} \ --n-iter {params.num_runs} \ --total-workers 1 \ --seed {params.seed} \ --genes-file {input.genes} ' " |
663 664 665 666 667 668 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/cNMF/cnmf.py combine \ --output-dir {params.outdir} \ --name {wildcards.sample} ' " |
685 686 687 688 689 | run: cmd = "mkdir -p " + os.path.join(params.outdir, wildcards.folder + "_acrossK", wildcards.sample, "cnmf_tmp") shell(cmd) cmd = "cp " + input.merged_result + " " + output.merged_copied_result shell(cmd) |
717 718 719 720 721 722 723 724 725 726 727 728 729 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {params.klist} \ --n-iter {params.num_runs} \ --total-workers 1 \ --seed {params.seed} \ --numgenes {wildcards.num_genes} ' " |
758 759 760 761 762 763 764 765 766 767 768 769 770 | shell: " bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {params.outdir}; \ python workflow/scripts/cNMF/cnmf.py prepare \ --output-dir {params.outdir} \ --name {wildcards.sample} \ -c {input.h5ad_mtx} \ -k {params.klist} \ --n-iter {params.num_runs} \ --total-workers 1 \ --seed {params.seed} \ --genes-file {input.genes} ' " |
815 816 817 818 819 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/cNMF/cnmf.modified.py k_selection_plot --output-dir {params.outdir} --name {wildcards.sample}; \ cp {output.plot} {output.plot_new_location} ' " |
892 893 894 895 896 897 898 899 900 | run: threshold_here = wildcards.threshold.replace("_",".") shell("bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/cNMF/cnmf.modified.py consensus \ --output-dir {params.outdir} \ --name {wildcards.sample} \ --components {wildcards.k} \ --local-density-threshold {threshold_here} ' ") # --show-clustering |
924 925 926 927 928 929 930 931 932 933 934 | run: threshold_here = wildcards.threshold.replace("_",".") shell("bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/cNMF/cnmf.modified.py consensus \ --output-dir {params.outdir} \ --name {wildcards.sample} \ --components {wildcards.k} \ --local-density-threshold {threshold_here} \ --show-clustering; \ cp {params.outdir}/{wildcards.sample}/{wildcards.sample}.clustering.k_{wildcards.k}.dt_{wildcards.threshold}.png {params.figdir}/{wildcards.folder}/{wildcards.sample}/K{wildcards.k}/ ' ") |
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/calcUMAP.only.R \ --outdir {params.outdir}/ \ --inputSeuratObject {input.input_seurat_object} \ --sampleName {wildcards.sample} \ --maxMt 50 \ --maxCount 25000 \ --minUniqueGenes 0 \ --UMAP.resolution 0.6' " ## can make this a wildcard rule plot_UMAP: input: seurat_object_withUMAP = os.path.join(config["analysisDir"], "data/{sample}.withUMAP_SeuratObject.RDS"), cNMF_Results = os.path.join(config["analysisDir"], "{folder}/{sample}/K{k}/threshold_{threshold}/cNMF_results.k_{k}.dt_{threshold}.RData") output: factor_expression_UMAP = os.path.join(config["figDir"], "{folder}/{sample}/K{k}/{sample}_K{k}_dt_{threshold}_Factor.Expression.UMAP.pdf") params: time = "6:00:00", mem_gb = "200", datadir = config["dataDir"], outdir = os.path.join(config["analysisDir"], "{folder}_acrossK/{sample}"), figdir = os.path.join(config["figDir"], "{folder}"), analysisdir = os.path.join(config["analysisDir"], "{folder}"), # K{k}/threshold_{threshold} threshold = get_cNMF_filter_threshold_double, partition = "owners,normal" |
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_UMAP_plot.R \ --sampleName {wildcards.sample} \ --inputSeuratObject {input.seurat_object_withUMAP} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --recompute F ' " |
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/variance_explained_v2.py \ --path_to_topics {params.path_to_topics} \ --topic_sampleName {wildcards.sample} \ --X_normalized {params.X_normalized_path} \ --outdir {params.outdir} \ --k {wildcards.k} \ --density_threshold {params.threshold} ' " |
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ conda info --env; \ Rscript workflow/scripts/cNMF_analysis.R \ --topic.model.result.dir {params.outdir}/ \ --sampleName {wildcards.sample} \ --barcode.names {params.barcode_names} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir}/ \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --recompute F \ --organism {params.organism} \ --motif.enhancer.background /oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv \ --motif.promoter.background /oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv \ ' " |
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_analysis_topic_plot.R \ --sampleName {wildcards.sample} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir}/ \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --recompute F ' " |
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/batch.topic.correlation.R \ --figdir {params.figdir} \ --outdir {params.analysisdir} \ --sampleName {wildcards.sample} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --barcode.names {input.barcode_names} \ --recompute F ' " |
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/motif_enrichment.R \ --sampleName {wildcards.sample} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --recompute F \ --ep.type {wildcards.ep_type} \ --organism {params.organism} \ --motif.match.thr.str {wildcards.motif_match_thr} \ --motif.enhancer.background {input.fimo_formatted} \ --motif.promoter.background {input.fimo_formatted} '" ## to do # --motif.enhancer.background {input.enhancer_fimo_formatted} \ # --motif.promoter.background {input.promoter_fimo_formatted} \ # ' " |
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_analysis_motif.enrichment_plot.R \ --sampleName {wildcards.sample} \ --ep.type {wildcards.ep_type} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --motif.match.thr.str {wildcards.motif_match_thr} \ --recompute F ' " |
1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_analysis_gsea_clusterProfiler.R \ --topic.model.result.dir {params.outdir} \ --sampleName {wildcards.sample} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --ranking.type {wildcards.ranking_type} \ --GSEA.type {wildcards.GSEA_type} \ --organism {params.organism} \ --recompute F ' " |
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/plot_gsea_clusterProfiler.R \ --sampleName {wildcards.sample} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --ranking.type {wildcards.ranking_type} \ --GSEA.type {wildcards.GSEA_type} '" |
1396 1397 1398 1399 1400 1401 1402 1403 1404 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/create_program_summary_table.R \ --sampleName {wildcards.sample} \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --perturbSeq {params.perturbseq} '" |
1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/create_comprehensive_program_summary_table.R \ --sampleName {wildcards.sample} \ --outdir {params.analysisdir} \ --scratch.outdir {params.scratch_outdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --perturbSeq {params.perturbseq} '" |
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/aggregate_across_K.R \ --figdir {params.figdir} \ --outdir {params.analysisdir} \ --sampleName {wildcards.sample} \ --K.list {params.klist_comma} \ --K.table {params.K_spectra_threshold_table} ' " ## how to create this automatically? rule findK_plot: input: toplot = os.path.join(config["analysisDir"], "{folder}/{sample}/acrossK/aggregated.outputs.findK.RData") output: GSEA_plots = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/All_GSEA.pdf"), TFMotifEnrichment_plots = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/All_TFMotifEnrichment.pdf"), topic_clustering_plot = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/cluster.topic.zscore.by.Pearson.corr.pdf"), variance_explained_plot = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/variance.explained.by.model.pdf") params: time = "3:00:00", mem_gb = "64", figdir = os.path.join(config["figDir"], "{folder}/{sample}/acrossK/"), analysisdir = os.path.join(config["analysisDir"], "{folder}/{sample}/acrossK/"), GO_threshold = 0.1, partition = "owners,normal" |
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/cNMF_findK_plots.R \ --figdir {params.figdir} \ --outdir {params.analysisdir} \ --sampleName {wildcards.sample} \ --p.adj.threshold 0.1 \ --aggregated.data {input.toplot} \ ' " |
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ mkdir -p {output.munged_features} {params.raw_features_dir}; \ cp -r {params.external_features}/* {params.raw_features_dir}/; \ cd {params.munged_features_dir}; \ python {params.pipelineDir}/workflow/scripts/pops/munge_feature_directory.py \ --gene_annot_path {params.gene_annot_path} \ --feature_dir {params.raw_features_dir} \ --save_prefix {wildcards.magma_prefix} \ --nan_policy zero ' " |
1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ rm -r {params.raw_features_dir}; \ mkdir -p {output.munged_features} {params.raw_features_dir}; \ cp -r {params.external_features}/* {params.raw_features_dir}/; \ cp {input.cNMF_ENSG_topic_zscore_scaled} {params.raw_features_dir}/; \ cd {params.munged_features_dir}; \ python {params.pipelineDir}/workflow/scripts/pops/munge_feature_directory.py \ --gene_annot_path {params.gene_annot_path} \ --feature_dir {params.raw_features_dir} \ --save_prefix {wildcards.magma_prefix}_cNMF{wildcards.k} \ --nan_policy zero ' " |
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python {params.pipelineDir}/workflow/scripts/pops/pops.py \ --gene_annot_path {params.gene_annot_path} \ --feature_mat_prefix {params.munged_features_dir}/{wildcards.magma_prefix}_cNMF{wildcards.k} \ --num_feature_chunks {params.num_munged_feature_chunks} \ --control_features_path {params.PoPS_control_features} \ --magma_prefix {params.magma_dir}/{wildcards.magma_prefix} \ --out_prefix {params.outdir}/{wildcards.magma_prefix}_cNMF{wildcards.k} \ --verbose ' " |
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python {params.pipelineDir}/workflow/scripts/pops/pops.py \ --gene_annot_path {params.gene_annot_path} \ --feature_mat_prefix {params.munged_features_dir}/{wildcards.magma_prefix} \ --num_feature_chunks {params.num_munged_feature_chunks} \ --control_features_path {params.PoPS_control_features} \ --magma_prefix {params.magma_dir}/{wildcards.magma_prefix} \ --out_prefix {params.outdir}/{wildcards.magma_prefix} \ --verbose ' " |
1670 1671 1672 1673 1674 1675 1676 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ mkdir -p {params.outdir}; \ Rscript workflow/scripts/PoPS_aggregate_features.R \ --feature.dir {params.raw_features_dir} \ --output {params.outdir}/ ' " |
1690 1691 1692 1693 1694 1695 1696 1697 1698 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ mkdir -p {params.outdir}; \ Rscript workflow/scripts/PoPS_aggregate_features_with_cNMF.R \ --feature.RDS {input.all_features_RDS} \ --cNMF.features {input.cNMF_ENSG_topic_zscore_scaled} \ --output {params.outdir}/ \ --prefix {wildcards.magma_prefix}_cNMF{wildcards.k} ' " |
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/PoPS.data.processing.R \ --output {params.outdir}/ \ --scratch.output {params.scratch_outdir}/ \ --prefix {wildcards.magma_prefix}_cNMF{wildcards.k} \ --external.features.metadata {params.external_features_metadata} \ --coefs_with_cNMF {input.coefs_with_cNMF} \ --preds_with_cNMF {input.preds_with_cNMF} \ --marginals_with_cNMF {input.marginals_with_cNMF} \ --coefs_without_cNMF {input.coefs_without_cNMF} \ --preds_without_cNMF {input.preds_without_cNMF} \ --marginals_without_cNMF {input.marginals_with_cNMF} \ --cNMF.features {input.cNMF_ENSG_topic_zscore_scaled} \ --all.features {input.all_features_with_cNMF_RDS} \ --recompute F ' " |
1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/PoPS.plots.R \ --sampleName {wildcards.sample} \ --output {params.outdir}/ \ --figure {params.figdir}/ \ --scratch.output {params.scratch_outdir}/ \ --prefix {wildcards.magma_prefix}_cNMF{wildcards.k} \ --k.val {wildcards.k} \ --coefs_with_cNMF {input.coefs_with_cNMF} \ --preds_with_cNMF {input.preds_with_cNMF} \ --marginals_with_cNMF {input.marginals_with_cNMF} \ --coefs_without_cNMF {input.coefs_without_cNMF} \ --preds_without_cNMF {input.preds_without_cNMF} \ --marginals_without_cNMF {input.marginals_with_cNMF} \ --cNMF.features {input.cNMF_ENSG_topic_zscore_scaled} \ --all.features {input.all_features_with_cNMF_RDS} \ --external.features.metadata {params.external_features_metadata} \ --combined.preds {input.combined_preds} \ --coefs.defining.top.topic.RDS {input.coe.afs_defining_top_topic_RDS} \ --preds.importance.score.key.columns {input.preds_importance_score_key_columns} ' " |
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/output_IGVF_format.R \ --sampleName {wildcards.sample} \ --outdir {params.analysisdir}/ \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --level {params.level} \ --cell.type {params.cell_type} \ ' " |
1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/create_cellxgene_h5ad_IGVF_format.py \ --path_to_topics {params.cNMF_outdir} \ --topic_sampleName {wildcards.sample} \ --outdir {params.analysisdir} \ --k {wildcards.k} \ --density_threshold {params.threshold} \ --barcode_dir {params.barcode_dir} \ ' " |
21 22 23 24 25 26 27 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_env; \ python workflow/scripts/filter_to_h5ad.py \ --inputPath {input.raw_h5ad_mtx} \ --output_h5ad {output.h5ad_mtx} \ --output_gene_name_txt {output.gene_name_txt} ' " |
12 13 14 15 | shell: "bash -c ' source ~/.bashrc; \ conda activate cnmf_env; \ bash workflow/scripts/fimo_motif_match.sh {input.coord} {input.fasta} {output.fasta} ' " |
27 28 29 30 31 32 33 34 35 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/program_prioritization/create_input_table.R \ --sampleName {wildcards.sample} \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --density.thr {params.threshold} \ --perturbSeq {params.perturbseq}' " # popsdir = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/" |
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/program_prioritization/compute_enrichment.R \ --input.GWAS.table {input.input_GWAS_table} \ --coding.variant.df {input.coding_variant_table} \ --sampleName {wildcards.sample} \ --outdirsample {params.outdirsample}/ \ --celltype {params.celltype} \ --figdir {params.figdir}/ \ --outdir {params.analysisdir} \ --K.val {wildcards.k} \ --trait.name {wildcards.GWAS_trait} \ --density.thr {params.threshold} \ --cNMF.table {input.input_table_for_compute_enrichment} \ --regulator.analysis.type {wildcards.regulator_analysis_type} \ --perturbSeq {params.perturbseq}' " |
12 13 14 15 16 17 18 | shell: "bash -c ' source $HOME/.bashrc; \ conda activate cnmf_analysis_R; \ Rscript workflow/scripts/seurat_to_h5ad.R \ --inputSeuratObject {input.seurat_object} \ --output_h5ad {output.h5ad_mtx} \ --output_gene_name_txt {output.gene_name_txt} ' " |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 | packages <- c("optparse","dplyr", "data.table", "reshape2", "conflicted","ggplot2", "tidyr", "textshape","readxl") # , "IsoplotR" xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/", help="Figure directory"), # "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes" make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), # "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_2min" make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), # "/oak/stanford/groups/engreitz/Users/kangh/process_sequencing_data/210912_FT010_fresh_Telo_sortedEC/multiome_FT010_fresh_2min/outs/filtered_feature_bc_matrix" make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), make_option("--K.list", type="character", default="14,15,60", help="K values available for analysis"), # make_option("--K.val", type="numeric", default=14, help="K value to analyze"), make_option("--K.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210625_snakemake_output/analysis/2kG.library/K.spectra.threshold.table.txt", help="table for defining spectra threshold"), # opt$K.table <-"/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/all_genes/2kG.library/K.spectra.threshold.table.txt" make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), # make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), # make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), # make_option("--raw.mtx.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), # make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), # make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), # make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), make_option("--adj.p.value.thr", type="numeric", default=0.05, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## 2n1.99x ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/figures/all_genes" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/" ## opt$sampleName <- "Perturb_2kG_dup4" ## scratch sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/" ## ## opt$datadir <- "/oak/stanford/groups/engreitz/Users/kangh/process_sequencing_data/210912_FT010_fresh_Telo_sortedEC/multiome_FT010_fresh_2min/outs/filtered_feature_bc_matrix" ## opt$sampleName <- "FT010_fresh_3min" mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) ## ## all genes (for interactive sessions) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K.spectra.threshold.table.txt" ## control only perturb-seq (for sdev) ## ## for testing findK_plots for control only cells ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/" ## opt$sampleName <- "2kG.library.ctrl.only" SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) # STATIC.SAMPLE=c("Telo_no_IL1B_T200_1", "Telo_no_IL1B_T200_2", "Telo_plus_IL1B_T200_1", "Telo_plus_IL1B_T200_2", "no_IL1B", "plus_IL1B", "pooled") DATADIR=opt$datadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir OUTDIR.ACROSS.K=paste0(OUTDIR,"/",SAMPLE,"/acrossK/") ## OUTDIR.ACROSS.K=paste0(OUTDIR,"/",SAMPLE,"/acrossK/threshold_", DENSITY.THRESHOLD, "/") # SEP=opt$sep K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() ## k <- opt$K.val FIGDIR=opt$figdir ## adjusted p-value threshold p.value.thr <- opt$adj.p.value.thr ## ## directories for factor motif enrichment ## FILENAME=opt$filename ## # create dir if not already ## if(SEP) check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE, ".sep/"), paste0(FIGDIR,SAMPLE, ".sep/K",k,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE, paste0(OUTDIR,SAMPLE, ".sep/"),FGSEADIR, FGSEAFIG, OUTDIR.ACROSS.K) else ## check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE, FGSEADIR, FGSEAFIG, OUTDIR.ACROSS.K) check.dir <- c(OUTDIR, FIGDIR, OUTDIR.ACROSS.K) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x) })) ## palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## selected.gene <- c("EDN1", "NOS3", "TP53", "GOSR2", "CDKN1A") ## # ABC genes ## gene.set <- c("INPP5B", "SF3A3", "SERPINH1", "NR2C1", "FGD6", "VEZT", "SMAD3", "AAGAB", "GOSR2", "ATP5G1", "ANGPTL4", "SRBD1", "PRKCE", "DAGLB") # ABC_0.015_CAD_pp.1_genes ## # cell cycle genes ## gene.list.three.groups <- read.delim(paste0(DATADIR,"/ptbd.genes_three.groups.txt"), header=T, stringsAsFactors=F) ## enhancer.set <- gene.list.three.groups$Gene[grep("E_at_", gene.list.three.groups$Gene)] ## CAD.focus.gene.set <- gene.list.three.groups %>% subset(Group=="CAD_focus") %>% pull(Gene) %>% append(enhancer.set) ## EC.pos.ctrl.gene.set <- gene.list.three.groups %>% subset(Group=="EC_pos._ctrls") %>% pull(Gene) ## cell.count.thr <- opt$cell.count.thr # greater than this number, filter to keep the guides with greater than this number of cells ## guide.count.thr <- opt$guide.count.thr # greater than this number, filter to keep the perturbations with greater than this number of guides ## guide.design = read.delim(file=paste0(opt$datadir, "/200607_ECPerturbSeqMiniPool.design.txt"), header=T, stringsAsFactors = F) ## ## add GO pathway log2FC ## GO <- read.delim(file=paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.table.brief.txt"), header=T, check.names=FALSE) ## GO.list <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.list.brief.txt", header=T, check.names=F) ## colnames(GO)[1] <- "Gene" ## colnames(GO.list)[1] <- "Gene" ## ## load all sample, K, topic's top 100 genes (by TopFeatures() KL-score measure) ## ## allGeneKtopic100 <- read.delim(paste0(TMDIR, "no.plus.pooled.top100.topicStats.txt"), header=T) ## # load non-expressed control gene list ## non.expressed.genes <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/non.expressed.ctrl.genes.txt", header=F, stringsAsFactors=F) %>% unlist %>% as.character() %>% sort() ## # perturbation type list ## gene.set.type.df <- data.frame(Gene=guide.design %>% pull(guideSet) %>% unique(), ## type=rep("other", guide.design %>% pull(guideSet) %>% unique() %>% length())) ## gene.set.type.df$Gene <- gene.set.type.df$Gene %>% as.character() ## gene.set.type.df$type <- gene.set.type.df$type %>% as.character() ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% non.expressed.genes)] <- "non-expressed" ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% CAD.focus.gene.set)] <- "CAD focus" ## gene.set.type.df$type[grepl("^safe|^negative", gene.set.type.df$Gene)] <- "negative-control" ## gene.set.type.df$Gene[which(gene.set.type.df$Gene == "negative_control")] <- "negative-control" ## gene.set.type.df$Gene[which(gene.set.type.df$Gene == "safe_targeting")] <- "safe-targeting" ## # gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" ## # convert enhancer SNP rs number to enhancer target gene name ## enh.snp.to.gene <- read.delim(paste0(DATADIR, "/enhancer.SNP.to.gene.name.txt"), header=T, stringsAsFactors = F) %>% mutate(Enhancer_name=gsub("_","-", Enhancer_name)) ## # gene corresponding pathway ## gene.def.pathways <- read_excel(paste0(DATADIR,"topic.gene.definition.pathways.xlsx"), sheet="Gene_Pathway") # K spectra threshold table if(file.exists(opt$K.table)) { K.spectra.threshold <- read.table(file=paste0(opt$K.table), header=T, stringsAsFactors=F) } else { K.spectra.threshold <- data.frame(K = K.list, density.threshold=rep(0.2, length(K.list))) ## assume 0.2 is the best threshold for filtering out outlier topics } ## load cNMF pipeline aggregated results (?) ## initialize storage variables # promoter.fisher.df.list <- enhancer.fisher.df.list <- fgsea.results <- all.test.df.list <- all.fdr.df.list <- count.by.GWAS.list <- count.by.GWAS.withTopic.list <- theta.zscore.list <- theta.raw.list <- all.enhancer.fisher.df.list <- all.promoter.fisher.df.list <- all.enhancer.fisher.df.10en6.list <- promoter.wide.10en6.list <- promoter.wide.binary.10en6.list <- all.promoter.fisher.df.10en6.list <- all.promoter.ttest.df.list <- all.promoter.ttest.df.10en6.list <- all.enhancer.ttest.df.list <- all.enhancer.ttest.df.10en6.list <- vector("list", nrow(K.spectra.threshold)) batch.percent.df.list <- all.test.df.list <- MAST.df.list <- vector("list", nrow(K.spectra.threshold)) ## loop over all values of K and aggregate results for (n in 1:nrow(K.spectra.threshold)) { k <- K.spectra.threshold[n,"K"] DENSITY.THRESHOLD <- K.spectra.threshold[n,"density.threshold"] %>% gsub("\\.","_",.) ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## directories # FIGDIRSAMPLE=ifelse(SEP, paste0(FIGDIR,SAMPLE,".sep/K",k,"/"), paste0(FIGDIR, SAMPLE, "/K",k,"/")) # FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k, "/threshold_", DENSITY.THRESHOLD) # FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") # FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## batch topic correlation file batch.correlation.file.name <- paste0(OUTDIRSAMPLE, "/batch.correlation.RDS") if(file.exists(batch.correlation.file.name)) { print(paste0("loading batch correlation file from: ", batch.correlation.file.name)) load(batch.correlation.file.name) batch.percent.df.list[[n]] <- batch.percent.df } else { print(paste0("file ", batch.correlation.file.name, " not found")) } # ## load motif enrichment results # file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.factorMotifEnrichment.",SUBSCRIPT.SHORT,".RData") # print(file.name) # if(file.exists((file.name))) { # load(file.name) # print(paste0("loading ", file.name)) # } # motif.enrichment.variables <- c("all.enhancer.fisher.df", "all.promoter.fisher.df", # "promoter.wide", "enhancer.wide", "promoter.wide.binary", "enhancer.wide.binary", # "enhancer.wide.10en6", "enhancer.wide.binary.10en6", "all.enhancer.fisher.df.10en6", # "promoter.wide.10en6", "promoter.wide.binary.10en6", "all.promoter.fisher.df.10en6", # "all.promoter.ttest.df", "all.promoter.ttest.df.10en6", "all.enhancer.ttest.df", "all.enhancer.ttest.df.10en6") # motif.enrichment.variables.missing <- (!(motif.enrichment.variables %in% ls())) %>% as.numeric %>% sum # if ( motif.enrichment.variables.missing > 0 ) { # warning(paste0(motif.enrichment.variables[!(motif.enrichment.variables %in% ls())], " not available")) # } else { # promoter.fisher.df.list[[n]] <- all.promoter.fisher.df %>% mutate(K = k) # enhancer.fisher.df.list[[n]] <- all.enhancer.fisher.df %>% mutate(K = k) # all.promoter.ttest.df.list[[n]] <- all.promoter.ttest.df %>% mutate(K = k) # all.promoter.ttest.df.10en6.list[[n]] <- all.promoter.ttest.df.10en6 %>% mutate(K = k) # all.enhancer.ttest.df.list[[n]] <- all.enhancer.ttest.df %>% mutate(K = k) # all.enhancer.ttest.df.10en6.list[[n]] <- all.enhancer.ttest.df.10en6 %>% mutate(K = k) # all.promoter.fisher.df.list[[n]] <- all.promoter.fisher.df # all.enhancer.fisher.df.list[[n]] <- all.enhancer.fisher.df # } ## all Wilcoxon statistical tests file.name <- paste0(OUTDIRSAMPLE, "/all.test.", SUBSCRIPT, ".txt") print(file.name) if(file.exists(file.name)) { print(paste0("loading ", file.name)) all.test.df.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) } else { print("all.test file not found") } ## MAST statistical test file.name <- paste0(OUTDIRSAMPLE, "/", SAMPLE, "_MAST_DEtopics.txt") print(file.name) if(file.exists(file.name)) { print(paste0("loading ", file.name)) MAST.df.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F, fill=T, check.names=F) %>% mutate(K = k) } else { print("MAST result file not found") } # file.name <- paste0(OUTDIRSAMPLE, "/all.expressed.genes.pval.fdr.", SUBSCRIPT, ".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # all.fdr.df.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } # # load count.by.GWAS # file.name <- paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # count.by.GWAS.list[[n]] <- read.delim(file=file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } # # load count.by.GWAS.with.topic # file.name <- paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes.withTopic_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # count.by.GWAS.withTopic.list[[n]] <- read.delim(file=file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } } batch.percent.df <- do.call(rbind, batch.percent.df.list) MAST.df <- do.call(rbind, MAST.df.list) all.test.df <- do.call(rbind, all.test.df.list) file.name <- paste0(OUTDIR.ACROSS.K, "/aggregated.outputs.findK.perturb-seq.RData") save(batch.percent.df, all.test.df, MAST.df, file = file.name) |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 | packages <- c("optparse","dplyr", "data.table", "reshape2", "conflicted","ggplot2", "tidyr", "textshape","readxl") # , "IsoplotR" xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/", help="Figure directory"), # "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes" make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), # "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_2min" make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), # "/oak/stanford/groups/engreitz/Users/kangh/process_sequencing_data/210912_FT010_fresh_Telo_sortedEC/multiome_FT010_fresh_2min/outs/filtered_feature_bc_matrix" make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), make_option("--K.list", type="character", default="14,30,60", help="K values available for analysis"), # make_option("--K.val", type="numeric", default=14, help="K value to analyze"), make_option("--K.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K.spectra.threshold.table.txt", help="table for defining spectra threshold"), # opt$K.table <-"/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/all_genes/2kG.library/K.spectra.threshold.table.txt" # make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), # make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), # make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), # make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), # make_option("--raw.mtx.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), # make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), # make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), # make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## sdev for 2n1.99x singlets ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/figures/all_genes" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/" ## ## opt$datadir <- "/oak/stanford/groups/engreitz/Users/kangh/process_sequencing_data/210912_FT010_fresh_Telo_sortedEC/multiome_FT010_fresh_2min/outs/filtered_feature_bc_matrix" ## opt$sampleName <- "Perturb_2kG_dup4" mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) ## ## all genes (for interactive sessions) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K.spectra.threshold.table.txt" ## ## overdispersed Genes ## opt$figdir <- "//oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/figures/top2000VariableGenes/" ## opt$outdir <- "//oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes/" ## opt$K.table <- "" ## opt$K.list <- "3,5,7,12,14,19,21,23,25,27,29,31,35,40,45,50,60,70,80,90,100,120" ## opt$sampleName <- "2kG.library_overdispersedGenes" ## ## sdev K562 gwps 2k most dispersed genes cNMF ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes" ## opt$sampleName <- "WeissmanK562gwps" ## opt$K.list <- "3,5,10,15,20,25,30,35,40,45,50,55,60,70,80,90,100,110,120" ## opt$K.table <- "/to/use/for/specifying/spectra/cut/off/threshold//default/0.2" SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) # STATIC.SAMPLE=c("Telo_no_IL1B_T200_1", "Telo_no_IL1B_T200_2", "Telo_plus_IL1B_T200_1", "Telo_plus_IL1B_T200_2", "no_IL1B", "plus_IL1B", "pooled") DATADIR=opt$datadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir OUTDIR.ACROSS.K=paste0(OUTDIR,"/",SAMPLE,"/acrossK/") K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() FIGDIR=opt$figdir ## adjusted p-value threshold p.value.thr <- opt$adj.p.value.thr ## # create dir if not already check.dir <- c(OUTDIR, FIGDIR, OUTDIR.ACROSS.K) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x) })) # K spectra threshold table if(file.exists(opt$K.table)) { K.spectra.threshold <- read.table(file=paste0(opt$K.table), header=T, stringsAsFactors=F) } else { K.spectra.threshold <- data.frame(K = K.list, density.threshold=rep(0.2, length(K.list))) ## assume 0.2 is the best threshold for filtering out outlier topics } ## initialize storage variables GSEA.types <- c("GOEnrichment", "PosGenesGOEnrichment", "ByWeightGSEA", "GSEA") for (j in 1:length(GSEA.types)) { GSEA.type <- GSEA.types[j] to.eval <- paste0("clusterProfiler.", GSEA.type, ".list <- vector(\"list\",nrow(K.spectra.threshold))") eval(parse(text = to.eval)) } all.MAST.df.list <- all.test.df.list <- all.fdr.df.list <- count.by.GWAS.list <- count.by.GWAS.withTopic.list <- theta.zscore.list <- theta.raw.list <- all.promoter.ttest.df.list <- all.enhancer.ttest.df.list <- varianceExplainedByModel.list <- varianceExplainedPerProgram.list <- vector("list", nrow(K.spectra.threshold)) ## loop over all values of K and aggregate results for (n in 1:nrow(K.spectra.threshold)) { k <- K.spectra.threshold[n,"K"] DENSITY.THRESHOLD <- K.spectra.threshold[n,"density.threshold"] %>% gsub("\\.","_",.) ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) # SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## directories # FIGDIRSAMPLE=ifelse(SEP, paste0(FIGDIR,SAMPLE,".sep/K",k,"/"), paste0(FIGDIR, SAMPLE, "/K",k,"/")) # FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k, "/threshold_", DENSITY.THRESHOLD) ## if (SEP) { ## guideCounts <- loadGuides(n, sep=T) %>% mutate(Gene=Gene.marked) ## tmp.labels <- guideCounts$Gene %>% unique() %>% strsplit("-") %>% sapply("[[",2) %>% unique() ## tmp.labels <- tmp.labels[!(tmp.labels %in% c("control","targeting"))] ## rep1.label <- paste0("-",tmp.labels[1]) ## rep2.label <- paste0("-",tmp.labels[2]) ## } else guideCounts <- loadGuides(n) %>% mutate(Gene=Gene.marked) ## cNMF direct output file (GEP) cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") ## cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.k_", k, ".dt_", density.threshold, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } else { print(paste0("file ", cNMF.result.file, " not found")) } # ## cNMF analysis results file # file.name <- ifelse(SEP, # paste0(OUTDIRSAMPLE,"/cNMFAnalysis.",SUBSCRIPT,".sep.RData"), # paste0(OUTDIRSAMPLE,"/cNMFAnalysis.",SUBSCRIPT,".RData")) # print(file.name) # if(file.exists(file.name)) { # print("loading the file") # load(file.name) # } else { # print("file not found") # } ## theta.zscore ## theta.zscore.list[[n]] <- theta.zscore %>% as.data.frame %>% `colnames<-`(paste0("factor_", colnames(.))) %>% mutate(Gene=rownames(.)) %>% melt(value.name="weight", id.vars="Gene", variable.name="Factor") %>% mutate(K=k) theta.zscore.list[[n]] <- theta.zscore %>% `colnames<-`(paste0("K",k,"_factor_", colnames(.))) ## theta.raw ## theta.raw.list[[n]] <- theta.raw %>% as.data.frame %>% `colnames<-`(paste0("factor_", colnames(.))) %>% mutate(Gene=rownames(.)) %>% melt(value.name="weight", id.vars="Gene", variable.name="Factor") %>% mutate(K=k) theta.raw.list[[n]] <- theta.raw %>% `colnames<-`(paste0("K",k,"_factor_", colnames(.))) ## ## theta.KL ## ## load KL score ## file.name <- paste0(OUTDIRSAMPLE, "/topic.KL.score_", SUBSCRIPT.SHORT, ".txt") %>% gsub("_k_", "_K", .) ## if(file.exists(file.name)) { ## print(paste0("Loading ", file.name)) ## theta.KL.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F) ## } else { ## print(paste0(file.name, " does not exist.")) ## } # ## motif enrichment file (old.211025) # file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.factorMotifEnrichment.",SUBSCRIPT.SHORT, ".RData") # print(file.name) # if(file.exists(file.name)) { # print("loading the file") # load(file.name) # } else { # print("file not found") # } # if ("all.promoter.fisher.df" %in% ls()) { # promoter.fisher.df.list[[n]] <- all.promoter.fisher.df %>% mutate(K = k) # enhancer.fisher.df.list[[n]] <- all.enhancer.fisher.df %>% mutate(K = k) # rm(list=c("all.promoter.fisher.df", "all.enhancer.fisher.df")) # } else { # print("missing all.promoter.fisher.df and/or all.enhancer.fisher.df") # } ## load motif enrichment results for(ep.type in c("promoter", "enhancer")){ num.top.genes <- 300 file.name <- paste0(OUTDIRSAMPLE, "/", ep.type, ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.pval1e-6_", SUBSCRIPT.SHORT,".txt") if(file.exists(file.name)) { eval(parse(text = paste0("all.", ep.type, ".ttest.df.list[[n]] <- read.delim(file.name, stringsAsFactors=F) %>% mutate(K = k)"))) ## store in all.{promoter, enhancer}.ttest.df.list } else { message(paste0(file.name, " does not exist")) } } # file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.factorMotifEnrichment.",SUBSCRIPT.SHORT,".RData") # print(file.name) # if(file.exists((file.name))) { # load(file.name) # print(paste0("loading ", file.name)) # } # motif.enrichment.variables <- c("all.enhancer.fisher.df", "all.promoter.fisher.df", # "promoter.wide", "enhancer.wide", "promoter.wide.binary", "enhancer.wide.binary", # "enhancer.wide.10en6", "enhancer.wide.binary.10en6", "all.enhancer.fisher.df.10en6", # "promoter.wide.10en6", "promoter.wide.binary.10en6", "all.promoter.fisher.df.10en6", # "all.promoter.ttest.df", "all.promoter.ttest.df.10en6", "all.enhancer.ttest.df", "all.enhancer.ttest.df.10en6") # motif.enrichment.variables.missing <- (!(motif.enrichment.variables %in% ls())) %>% as.numeric %>% sum # if ( motif.enrichment.variables.missing > 0 ) { # warning(paste0(motif.enrichment.variables[!(motif.enrichment.variables %in% ls())], " not available")) # } else { # promoter.fisher.df.list[[n]] <- all.promoter.fisher.df %>% mutate(K = k) # enhancer.fisher.df.list[[n]] <- all.enhancer.fisher.df %>% mutate(K = k) # all.promoter.ttest.df.list[[n]] <- all.promoter.ttest.df %>% mutate(K = k) # all.promoter.ttest.df.10en6.list[[n]] <- all.promoter.ttest.df.10en6 %>% mutate(K = k) # all.enhancer.ttest.df.list[[n]] <- all.enhancer.ttest.df %>% mutate(K = k) # all.enhancer.ttest.df.10en6.list[[n]] <- all.enhancer.ttest.df.10en6 %>% mutate(K = k) # all.promoter.fisher.df.list[[n]] <- all.promoter.fisher.df # all.enhancer.fisher.df.list[[n]] <- all.enhancer.fisher.df # } ## GSEA results ranking.types <- c("zscore", "raw", "median_spectra", "median_spectra_zscore") for (j in 1:length(GSEA.types)) { GSEA.type <- GSEA.types[j] to.eval <- paste0("clusterProfiler.", GSEA.type, ".list.here <- vector(\"list\",length(ranking.types))") eval(parse(text = to.eval)) for (i in 1:length(ranking.types)) { ranking.type <- ranking.types[i] file.name <- paste0(OUTDIRSAMPLE,"/clusterProfiler_GeneRankingType",ranking.type,"_EnrichmentType", GSEA.type, ".txt") if(file.exists(file.name)) { message("Loading ", file.name) to.eval <- paste0("clusterProfiler.", GSEA.type, ".list.here[[i]] <- read.delim(file.name, header=T, stringsAsFactors = F) %>% mutate(type = ranking.type, K = k)") eval(parse(text = to.eval)) } else { warning(paste0(file.name, " file does not exist")) } } to.eval <- paste0("clusterProfiler.", GSEA.type, ".list[[n]] <- do.call(rbind, clusterProfiler.", GSEA.type, ".list.here)") eval(parse(text = to.eval)) } ## variance explained by the model file.name <- paste0(OUTDIRSAMPLE, "/summary.varianceExplained.df.txt") if(file.exists(file.name)) { varianceExplainedByModel.list[[n]] <- read.delim(file.name, stringsAsFactors=F) %>% mutate(K = k) } else { message(paste0(file.name, " does not exist")) } ## variance explained per program file.name <- paste0(OUTDIRSAMPLE, "/metrics.varianceExplained.df.txt") if(file.exists(file.name)) { varianceExplainedPerProgram.list[[n]] <- read.delim(file.name, stringsAsFactors=F) } else { message(paste0(file.name, " does not exist")) } # ## all statistical tests # file.name <- paste0(OUTDIRSAMPLE, "/all.test.", SUBSCRIPT, ".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # all.test.df.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("all.test file not found") # } # file.name <- paste0(OUTDIRSAMPLE, "/all.expressed.genes.pval.fdr.", SUBSCRIPT, ".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # all.fdr.df.list[[n]] <- read.table(file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } # # load count.by.GWAS # file.name <- paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # count.by.GWAS.list[[n]] <- read.delim(file=file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } # # load count.by.GWAS.with.topic # file.name <- paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes.withTopic_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt") # print(file.name) # if(file.exists(file.name)) { # print(paste0("loading ", file.name)) # count.by.GWAS.withTopic.list[[n]] <- read.delim(file=file.name, header=T, stringsAsFactors=F) %>% mutate(K = k) # } else { # print("file not found") # } } # promoter.fisher.df <- do.call(rbind, promoter.fisher.df.list) # enhancer.fisher.df <- do.call(rbind, enhancer.fisher.df.list) GSEA.types <- c("GOEnrichment", "ByWeightGSEA", "GSEA") ## use external input clusterProfiler.GO.list.here <- clusterProfiler.GSEA.list.here <- clusterProfiler.enricher.GSEA.list.here <- vector("list",length(GSEA.types)) for (j in 1:length(GSEA.types)) { GSEA.type <- GSEA.types[j] to.eval <- paste0("clusterProfiler.", GSEA.type, ".df <- do.call(rbind, clusterProfiler.", GSEA.type, ".list)") eval(parse(text = to.eval)) } ## clusterProfiler.GO.df <- do.call(rbind, clusterProfiler.GO.list) ## clusterProfiler.GSEA.df <- do.call(rbind, clusterProfiler.GSEA.list) ## clusterProfiler.enricher.GSEA.df <- do.call(rbind, clusterProfiler.enricher.GSEA.list) # all.test.df <- do.call(rbind, all.test.df.list) # all.fdr.df <- do.call(rbind, all.fdr.df.list) # count.by.GWAS <- do.call(rbind, count.by.GWAS.list) # count.by.GWAS.withTopic <- do.call(rbind, count.by.GWAS.withTopic.list) theta.zscore.df <- do.call(cbind, theta.zscore.list) theta.raw.df <- do.call(cbind, theta.raw.list) # theta.KL.df <- do.call(rbind, theta.KL.list) all.promoter.ttest.df <- do.call(rbind, all.promoter.ttest.df.list) # all.promoter.ttest.df.10en6 <- do.call(rbind, all.promoter.ttest.df.10en6.list) all.enhancer.ttest.df <- do.call(rbind, all.enhancer.ttest.df.list) # all.enhancer.ttest.df.10en6 <- do.call(rbind, all.enhancer.ttest.df.10en6.list) varianceExplainedByModel.df <- do.call(rbind, varianceExplainedByModel.list) varianceExplainedPerProgram.df <- do.call(rbind, varianceExplainedPerProgram.list) file.name <- paste0(OUTDIR.ACROSS.K, "/aggregated.outputs.findK.RData") save(clusterProfiler.GOEnrichment.df, clusterProfiler.ByWeightGSEA.df, clusterProfiler.GSEA.df, theta.zscore.df, theta.raw.df, all.promoter.ttest.df, all.enhancer.ttest.df, varianceExplainedByModel.df, varianceExplainedPerProgram.df, file=file.name) |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel","ramify", "ggpubr","gridExtra", "org.Hs.eg.db","limma","fgsea", "conflicted", "cluster","textshape","readxl", "ggdist", "gghalves", "Seurat", "writexl", "purrr") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/", help="Output directory"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), make_option("--barcode.names", type="character", default="", help="metadata CBC and sample information data table"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 # ## control genes directories (for sdev) # opt$sampleName <- "2kG.library.ctrl.only" # opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/" # opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/all_genes_acrossK/2kG.library.ctrl.only/" # opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/" # opt$K.val <- 60 ## ## overdispersed gene directories (for sdev) ## opt$sampleName <- "2kG.library_overdispersedGenes" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/figures/top2000VariableGenes" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes" ## opt$K.val <- 120 ## ## K562 gwps sdev ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$K.val <- 90 ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## ## ENCODE Mouse Heart data ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/figures/top2000VariableGenes" ## opt$K.val <- 55 ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/collab_data/IGVF/mouse_ENCODE_heart/auxiliary_data/snrna/heart_Parse_10x_integrated_metadata.csv" ## ## teloHAEC no_IL1B 200 gene library ## opt$sampleName <- "no_IL1B" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/analysis/all_genes/" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/figures/all_genes/" ## opt$K.val <- 20 ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/no_IL1B.barcodes.txt" SAMPLE=strsplit(opt$sampleName,",") %>% unlist() ## DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr # create dir if not already check.dir <- c(OUTDIR, FIGDIR, OUTDIRSAMPLE, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## graphing constants and helpers palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ################################################## ## load data cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") if(file.exists(cNMF.result.file)) { message(paste0("loading cNMF result file: \n", cNMF.result.file)) load(cNMF.result.file) } else { print(paste0(cNMF.result.file, " does not exist")) } ## annotate omega to get Gene, Guide, Sample, and CBC if(grepl("2kG.library", SAMPLE)) { ann.omega <- cbind(omega, barcode.names) ## %>% } else { if(grepl("[.]csv", opt$barcode.names)) barcode.names <- read.delim(opt$barcode.names, stringsAsFactors=F, sep=",") else barcode.names <- read.delim(opt$barcode.names, stringsAsFactors=F) ann.omega <- merge(omega, barcode.names %>% select(CBC, sample), by.x=0, by.y="CBC", all.x=T) } ## Batch Effect QC: correlate batch binary labels with topic expression ## Batch binary label if(grepl("2kG.library", SAMPLE)) { ann.omega.batch <- ann.omega %>% mutate(sample.short = gsub("scRNAseq_2kG_", "", sample) %>% gsub("_.*$","", .)) ann.omega.batch.binary <- ann.omega.batch %>% mutate(tmp.value = 1) %>% spread(key="sample.short", fill=0, value="tmp.value") ann.omega.batch.binary.mtx <- ann.omega.batch.binary %>% select(-long.CBC,-Gene.full.name,-Guide,-CBC,-sample,-Gene) %>% as.matrix() m <- cor(ann.omega.batch.binary.mtx, method="pearson") %>% as.matrix() batch.correlation.mtx <- m[1:k,(k+1):(dim(m)[2])] } ann.omega.sample.batch.binary <- ann.omega %>% mutate(tmp.value = 1) %>% spread(key="sample", fill=0, value="tmp.value") ann.omega.sample.batch.binary.mtx <- ann.omega.sample.batch.binary %>% select(-Row.names) %>% as.matrix() m <- cor(ann.omega.sample.batch.binary.mtx, method="pearson") %>% as.matrix() sample.batch.correlation.mtx <- m[1:k, (k+1):(dim(m)[2])] ## calculate percent of topics with correlation past a threshold (0.1, 0.2, 0.4, 0.6) correlation.threshold.list <- c(0.1, 0.2, 0.4, 0.6) batch.passed.threshold.df <- do.call(rbind, lapply(correlation.threshold.list, function(threshold) { df <- sample.batch.correlation.mtx %>% apply(1, function(x) (x > threshold) %>% as.numeric %>% sum) %>% as.data.frame %>% `colnames<-`("num.batch.correlated") %>% mutate(batch.thr = threshold) %>% mutate(ProgramID = rownames(.), K = k) })) batch.percent.df <- batch.passed.threshold.df %>% group_by(batch.thr) %>% summarize(percent.correlated = ((num.batch.correlated > 0) %>% as.numeric %>% sum) / k) %>% mutate(K = k) ## max batch correlation per topic max.batch.correlation.df <- sample.batch.correlation.mtx %>% apply(1, function(x) { out <- max(abs(x)) }) %>% as.data.frame %>% `colnames<-`("maxPearsonCorrelation") %>% mutate(ProgramID = row.names(.)) %>% as.data.frame ## store batch and sample correlation matrix if(grepl("2kG.library", SAMPLE)) write.table(batch.correlation.mtx, file=paste0(OUTDIRSAMPLE, "/batch.correction.mtx.txt"), sep="\t", quote=F) write.table(sample.batch.correlation.mtx, file=paste0(OUTDIRSAMPLE, "/sample.batch.correction.mtx.txt"), sep="\t", quote=F) write.table(batch.passed.threshold.df, file=paste0(OUTDIRSAMPLE, "/batch.passed.thr.df.txt"), sep="\t", quote=F, row.names=F) write.table(batch.percent.df, file=paste0(OUTDIRSAMPLE, "/batch.percent.df.txt"), sep="\t", quote=F, row.names=F) write.table(max.batch.correlation.df, file=paste0(OUTDIRSAMPLE, "/max.batch.correlation.df.txt"), sep="\t", quote=F, row.names=F) if(grepl("2kG.library", SAMPLE)) { save(batch.correlation.mtx, sample.batch.correlation.mtx, batch.passed.threshold.df, batch.percent.df, max.batch.correlation.df, file=paste0(OUTDIRSAMPLE, "/batch.correlation.RDS")) } else { save(sample.batch.correlation.mtx, batch.passed.threshold.df, batch.percent.df, max.batch.correlation.df, file=paste0(OUTDIRSAMPLE, "/batch.correlation.RDS")) } ## Batch correlation heatmap plotHeatmap <- function(mtx, title){ heatmap.2( mtx, Rowv=T, Colv=T, trace='none', key=T, col=palette, labCol=colnames(mtx), ## margins=c(15,5), cex.main=0.1, cexCol=1/(nrow(mtx)^(1/7)), cexRow=1/(ncol(mtx)^(1/7)), main=title ) } pdf(paste0(FIGDIRTOP, "batch.correlation.heatmap.pdf"),width=0.15*ncol(sample.batch.correlation.mtx)+5, height=0.1*nrow(sample.batch.correlation.mtx)+5) if(grepl("2kG.library", SAMPLE)) plotHeatmap(batch.correlation.mtx, title=paste0(SAMPLE, ", K=", k, ", topic batch correlation")) plotHeatmap(sample.batch.correlation.mtx, title=paste0(SAMPLE, ", K=", k, ", topic sample correlation")) dev.off() ## automate batch topic selection Pearson.correlation.threshold <- 0.1 ## is this a good threshold for all values of K? ## check CDF of average correlation? batch.topic <- apply(sample.batch.correlation.mtx, 1, function(x) sum(as.numeric(abs(x) > Pearson.correlation.threshold))) %>% keep(function(x) x > 0) %>% names write.table(batch.topic, file=paste0(OUTDIRSAMPLE, "batch.topics.txt"), quote=F, sep="\t", row.names=F, col.names=F) |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 | suppressPackageStartupMessages(library(optparse)) option.list <- list( make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/outputs/", help="Output directory"), make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/outputs/", help="Output directory"), make_option("--sampleName",type="character",default="FT005_gex_new_pipeline", help="Sample name"), make_option("--project",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/",help="Project Directory"), make_option("--inputSeuratObject", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/outputs/FT005_gex/withUMAP.SeuratObject.RDS", help="Path to the Seurat Object"), make_option("--compareSeuratObject", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/outputs/FT005_gex_new_pipeline/withUMAP.SeruatObject.RDS", help="Path to the Seurat Object"), make_option("--maxMt", type="numeric", default=50, help="filter out cells with percent mitochondrial gene higher than this threhsold"), make_option("--maxCount", type="numeric", default=25000, help="filter out cells with UMI count more than this threshold"), make_option("--minUniqueGenes", type="numeric", default=0, help="filter out cells with unique gene detected less than this threshold"), make_option("--UMAP.resolution", type="numeric", default=0.06, help="UMAP resolution. The default is 0.06") ) opt <- parse_args(OptionParser(option_list=option.list)) library(SeuratObject) library(Seurat) suppressPackageStartupMessages(library(dplyr)) suppressPackageStartupMessages(library(ggplot2)) suppressPackageStartupMessages(library(data.table)) suppressPackageStartupMessages(library(tidyr)) suppressPackageStartupMessages(library(readxl)) suppressPackageStartupMessages(library(ggrepel)) ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") mytheme <- theme_classic() + theme(axis.text = element_text(size = 13), axis.title = element_text(size = 15), plot.title = element_text(hjust = 0.5)) ####################################################################### ## Constants PROJECT=opt$project OUTDIR=opt$outdir FIGDIR=opt$figdir SAMPLE=opt$sampleName # OUTDIRSAMPLE=paste0(OUTDIR,"/",SAMPLE,"/") FIGDIRSAMPLE=paste0(FIGDIR,"/",SAMPLE,"/") palette = colorRampPalette(c("#38b4f7", "white", "red"))(n=100) # create dir if not already # check.dir <- c(OUTDIR, OUTDIRSAMPLE, FIGDIR, FIGDIRSAMPLE) check.dir <- c(OUTDIR, FIGDIR, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x) })) ## function for calculating UMAP calcUMAP <- function(s) { s <- SCTransform(s) s <- RunPCA(s, verbose = FALSE) s <- FindNeighbors(s, dims = 1:10) s <- FindClusters(s, resolution = opt$UMAP.resolution) s <- RunUMAP(s, dims = 1:10) } ## load data s <- readRDS(opt$inputSeuratObject) s <- calcUMAP(s) ## Choose filters, use cells with Good_singlet, and filter MT/RP s[["percent.mt"]] <- PercentageFeatureSet(s, pattern = "^MT-") s[["percent.ribo"]] <- PercentageFeatureSet(s, pattern = "^RPS|^RPL") ## plot QC ## s.meta <- SeuratObject::FetchData(s, colnames(s[[]])) ## This weird Seurat syntax gets the list of all metadata vars and then fetches a matrix of the data # plotSingleCellStats(s.meta, mtMax=NULL, nCountMax=NULL, paste0(FIGDIRSAMPLE,"/QC.single.cell.stats.UMAPres.", opt$UMAP.resolution, ".pdf")) # saveRDS(s, paste0(OUTDIRSAMPLE,"/", SAMPLE, ".withUMAP.", opt$UMAP.resolution, ".SeuratObject.RDS")) # saveRDS(s, paste0(OUTDIRSAMPLE,"/", SAMPLE, ".withUMAP_SeuratObject.RDS")) saveRDS(s, paste0(OUTDIR,"/", SAMPLE, ".withUMAP_SeuratObject.RDS")) |
7
of
scripts/calcUMAP.only.R
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 | packages <- c("optparse", "data.table", "reshape2", "fgsea", "conflicted", "readxl", "writexl", "org.Hs.eg.db", "tidyr", "dplyr", "clusterProfiler", "msigdbr") xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/", help="Output directory"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210707_snakemake_maxParallel/all_genes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), make_option("--K.list", type="character", default="2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,19,21,23,25", help="K values available for analysis"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--raw.mtx.dir",type="character",default="stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), # the first lane: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.expandedMultiTargetGuide.RDS" make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), # make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), make_option("--enhancer.fimo.threshold", type="character", default="1.0E-4", help="Enhancer fimo motif match threshold"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute"), ## GSEA parameters make_option("--ranking.type", type="character", default="zscore", help="{zscore, raw} ranking for the top program genes"), make_option("--GSEA.type", type="character", default="GOEnrichment", help="{GOEnrichment, ByWeightGSEA, GSEA}"), ## make_option("--", type="", default= , help="") ## Organism flag make_option("--organism", type="character", default="human", help="Organism type, accept org.Hs.eg.db. Only support human and mouse.") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 ## ## ## K562 gwps sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$K.val <- 35 ## opt$sampleName <- "WeissmanK562gwps" ## opt$GSEA.type <- "ByWeightGSEA" ## opt$ranking.type <- "median_spectra_zscore" ## ## ENCODE mouse heart ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes" ## opt$K.val <- 10 ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$GSEA.type <- "ByWeightGSEA" ## opt$ranking.type <- "zscore" ## ## teloHAEC no_IL1B 200 gene library ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/analysis/top2000VariableGenes/" ## opt$K.val <- 20 ## opt$sampleName <- "no_IL1B" ## opt$GSEA.type <- "ByWeightGSEA" ## opt$ranking.type <- "median_spectra" ## ## IGVF b01_LeftCortex ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/" ## opt$K.val <- 20 ## opt$sampleName <- "IGVF_b01_LeftCortex" ## opt$GSEA.type <- "GSEA" ## opt$ranking.type <- "median_spectra" ## opt$organism <- "mouse" ## ## RCA Pt4 ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220124_snakemake_RCA/analysis/all_genes_acrossK/RCA" ## opt$sampleName <- "RCA" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220124_snakemake_RCA/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220124_snakemake_RCA/analysis/all_genes" ## opt$K.val <- 60 ## opt$ranking.type <- "median_spectra_zscore" ## opt$GSEA.type <- "GOEnrichment" ## opt$organism <- "human" SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir TMDIR=opt$topic.model.result.dir # SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") ## FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") ## FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) # SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr db <- ifelse(grepl("mouse|org.Mm.eg.db", opt$organism), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) ## load the appropriate database # create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## helper function to map between ENSGID and SYMBOL map.ENSGID.SYMBOL <- function(df) { ## need column `Gene` to be present in df ## detect gene data type (e.g. ENSGID, Entrez Symbol) gene.type <- ifelse(nrow(df) == sum(as.numeric(grepl("^ENS", df$Gene))), "ENSGID", "Gene") if(gene.type == "ENSGID") { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "ENSEMBL", column = "SYMBOL") df <- df %>% mutate(ENSGID = Gene, Gene = mapped.genes) } else { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "SYMBOL", column = "ENSEMBL") df <- df %>% mutate(ENSGID = mapped.genes) } return(df) } ###################################################################### ## Load topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } ## get list of topic defining genes theta.rank.list <- vector("list", ncol(theta.zscore))## initialize storage list for(i in 1:ncol(theta.zscore)) { topic <- paste0("topic_", colnames(theta.zscore)[i]) theta.rank.list[[i]] <- theta.zscore %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("topic.zscore") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.zscore), .before="topic.zscore") %>% mutate(zscore.specificity.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.rank.df <- do.call(rbind, theta.rank.list) %>% ## combine list to df `colnames<-`(c("topic.zscore", "Gene", "zscore.specificity.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## get list of topic genes by raw weight theta.raw.rank.list <- vector("list", ncol(theta.raw))## initialize storage list for(i in 1:ncol(theta.raw)) { topic <- paste0("topic_", colnames(theta.raw)[i]) theta.raw.rank.list[[i]] <- theta.raw %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("topic.raw") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.raw), .before="topic.raw") %>% mutate(raw.score.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.raw.rank.df <- do.call(rbind, theta.raw.rank.list) %>% ## combine list to df `colnames<-`(c("topic.raw", "Gene", "raw.score.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## get list of topic genes by median spectra weight median.spectra.rank.list <- vector("list", ncol(median.spectra))## initialize storage list for(i in 1:ncol(median.spectra)) { topic <- paste0("topic_", colnames(median.spectra)[i]) median.spectra.rank.list[[i]] <- median.spectra %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("median.spectra") %>% mutate(Gene = rownames(.)) %>% arrange(desc(median.spectra), .before="median.spectra") %>% mutate(median.spectra.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } median.spectra.rank.df <- do.call(rbind, median.spectra.rank.list) %>% ## combine list to df `colnames<-`(c("median.spectra", "Gene", "median.spectra.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## median.spectra.zscore.df <- median.spectra.zscore.df %>% mutate(Gene = ENSGID) ## quick fix, need to add "Gene" column to this dataframe in analysis script ###################################################################### ## run cluster profiler GSEA on top 300 genes ## ## map between EntrezID and Gene Symbol ## z <- org.Hs.egSYMBOL ## z_mapped_genes <- mappedkeys(z) ## entrez.to.symbol <- as.list(z[z_mapped_genes]) ## ## entrez.to.symbol <- as.list(org.Hs.egSYMBOL) ## symbol.to.entrez <- as.list(org.Hs.egSYMBOL2EG) ## ## map to entrez id (function) ## symbolToEntrez <- function(df) df %>% mutate(EntrezID = symbol.to.entrez[.$gene %>% as.character] %>% sapply("[[",1) %>% as.character) ## subset to top 300 genes ranking.type.ary <- c("zscore", "raw", "median_spectra_zscore", "median_spectra") score.colname.ary <- c("zscore", "raw", "median.spectra.zscore", "median.spectra") ranking.rank.colname.ary <- c("zscore.specificity.rank", "raw.score.rank", "median.spectra.zscore.rank", "median.spectra.rank") ranking.type.varname.ary <- c("theta.rank.df", "theta.raw.rank.df", "median.spectra.zscore.df", "median.spectra.rank.df") getData <- function(t) { i <- which(ranking.type.ary == opt$ranking.type) programID.here <- paste0("K", k, "_", t) ranking.type.varname.here <- ranking.type.varname.ary[i] if(grepl("median.spectra", ranking.type.varname.here)) { ranking.score.colname.here <- score.colname.ary[i] } else { ranking.score.colname.here <- paste0("topic.", ranking.type.ary[i]) } gene.df <- get(ranking.type.varname.here) %>% subset(ProgramID == programID.here) gene.type <- ifelse(nrow(gene.df) == sum(as.numeric(grepl("^ENS", gene.df$Gene))), "ENSGID", "Gene") mapped.genes <- mapIds(get(db), keys=gene.df$Gene, keytype = ifelse(gene.type == "Gene", "SYMBOL", "ENSEMBL"), column = ifelse(gene.type == "Gene", "ENSEMBL", "SYMBOL")) mapped.entrez.genes <- mapIds(get(db), keys=gene.df$Gene, keytype = ifelse(gene.type == "Gene", "SYMBOL", "ENSEMBL"), column = "ENTREZID") gene.df <- gene.df %>% mutate(!!gene.type := Gene, !!ifelse(gene.type=="ENSGID", "Gene", "ENSGID") := mapped.genes, EntrezID = mapped.entrez.genes) %>% as.data.frame gene.weights <- gene.df %>% pull(get(ranking.score.colname.here)) %>% `names<-`(gene.df$EntrezID) gene.weights[gene.weights < 0] <- 0 top.gene.df <- gene.df %>% subset(get(ranking.rank.colname.ary[i]) <= 300) %>% as.data.frame ## top.genes <- unlist(mget(top.gene.df$Gene, envir=org.Hs.egSYMBOL2EG, ifnotfound=NA)) ## old top.genes <- top.gene.df %>% pull(EntrezID) ## same as above pos.gene.df <- gene.df %>% subset(get(ranking.score.colname.here) > 0) %>% as.data.frame pos.genes <- pos.gene.df %>% pull(EntrezID) ## pos.genes <- unlist(mget(pos.gene.df$Gene, envir=org.Hs.egSYMBOL2EG, ifnotfound=NA)) ## geneUniverse <- unlist(mget(get(ranking.type.varname.ary[i])$Gene %>% unique, envir=org.Hs.egSYMBOL2EG, ifnotfound=NA)) geneUniverse <- gene.df$EntrezID return(list(top.genes = top.genes, pos.genes = pos.genes, geneUniverse = geneUniverse, gene.weights = gene.weights)) } m_df <- msigdbr(species = ifelse(grepl("mouse", opt$organism), "Mus musculus", "Homo sapiens")) ## save this as a txt file and read in ## for future if needed functionsToRun <- list(GOEnrichment = "out <- enrichGO(gene = top.genes, ont = 'ALL', OrgDb = db, universe = geneUniverse, readable=T, pvalueCutoff=1, pAdjustMethod = 'fdr') %>% as.data.frame %>% mutate(fdr.across.ont = p.adjust, ProgramID = paste0('K', k, '_', t))", PosGenesGOEnrichment = "out <- enrichGO(gene = pos.genes, ont = 'ALL', OrgDb = db, universe = geneUniverse, readable=T, pvalueCutoff=1, pAdjustMethod='fdr') %>% as.data.frame %>% mutate(fdr.across.ont = p.adjust, ProgramID = paste0('K', k, '_', t))", ByWeightGSEA = "out <- GSEA(gene.weights, TERM2GENE = m_df %>% select(gs_name, entrez_gene), pAdjustMethod = 'fdr', pvalueCutoff = 1) %>% as.data.frame %>% mutate(ProgramID = paste0('K', k, '_', t)) ", GSEA = "out <- enricher(top.genes, TERM2GENE = m_df %>% select(gs_name, entrez_gene), universe = geneUniverse, pAdjustMethod = 'fdr', qvalueCutoff=1) %>% as.data.frame %>% mutate(ProgramID = paste0('K', k, '_', t))" ) ## for(i in 1:length(ranking.type.ary)) { ranking.type.here <- opt$ranking.type GSEA.type <- opt$GSEA.type ## ranking.type.here <- ranking.type.ary[i] ## GO enrichment analysis. Include all of: ## MF: Molecular Function ## CC: Cellular Component ## BP: Biological Process ## ans.go <- do.call(rbind, lapply(1:60, function(t) { message("starting enrichment") ## out.list <- lapply(1:length(functionsToRun), function(j) { out <- do.call(rbind, lapply(c(1:k) %>% rev, function(t) { data.here <- getData(t) top.genes <- data.here$top.genes if(sum(as.numeric(is.na(names(top.genes)))) > 0) top.genes <- top.genes[-which(is.na(names(top.genes)))] ## remove genes that doesn't have matched Entrez ID ## print(head(top.genes)) pos.genes <- data.here$pos.genes if(sum(as.numeric(is.na(names(pos.genes)))) > 0) pos.genes <- pos.genes[-which(is.na(names(pos.genes)))] ## print(head(pos.genes)) geneUniverse <- data.here$geneUniverse ## print(head(geneUniverse)) gene.weights <- data.here$gene.weights if (sum(as.numeric(is.na(names(gene.weights)))) > 0) gene.weights <- gene.weights[-which(is.na(names(gene.weights)))] if (which(gene.weights==0) %>% length > 0) gene.weights <- gene.weights[-which(gene.weights==0)] ## can't have zero weights? if (length(which(is.na(gene.weights))) > 0) gene.weights <- gene.weights[-which(is.na(gene.weights))] ## can't have NA ## print(head(gene.weights)) message(paste0("Ranking type: ", ranking.type.here, ", Program ", t, ", out of ", k, ", function ", GSEA.type, ", top gene class: ", class(top.genes), "\n geneUniverse class: ", class(geneUniverse), ", gene.weights class: ", class(gene.weights))) ## message(paste0("Function to run: \n", functionsToRun[[GSEA.type]])) eval(parse(text = functionsToRun[[GSEA.type]])) return(out) })) ## if(j == 1) { file.name <- paste0(OUTDIRSAMPLE, "/clusterProfiler_GeneRankingType", ranking.type.here, "_EnrichmentType", GSEA.type,".txt") ## } else if (j == 2) { ## file.name <- paste0(OUTDIRSAMPLE, "/clusterProfiler_allGene", ranking.type.here, "_ByWeight_GSEA.txt") ## } else { ## file.name <- paste0(OUTDIRSAMPLE, "/clusterProfiler_top300Genes", ranking.type.here, "_GSEA.txt") ## } message(paste0("output table to ", file.name)) write.table(out, file.name, sep="\t", row.names=F, quote=F) |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 | packages <- c("optparse","dplyr", "ggplot2", "reshape2", "ggrepel", "conflicted") xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--ep.type", type="character", default="enhancer", help="motif enrichment for enhancer or promoter, specify 'enhancer' or 'promoter'"), make_option("--adj.p.value.thr", type="numeric", default=0.05, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute"), make_option("--motif.match.thr.str", type="character", default="pval1e-6", help="threshold for subsetting motif matches") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 mytheme <- theme_classic() + theme(axis.text = element_text(size = 7), axis.title = element_text(size = 8), plot.title = element_text(hjust = 0.5, face = "bold", size=8)) SAMPLE=strsplit(opt$sampleName,",") %>% unlist() # STATIC.SAMPLE=c("Telo_no_IL1B_T200_1", "Telo_no_IL1B_T200_2", "Telo_plus_IL1B_T200_1", "Telo_plus_IL1B_T200_2", "no_IL1B", "plus_IL1B", "pooled") # DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir ## TMDIR=opt$topic.model.result.dir ## SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val num.top.genes <- 300 ## number of top topic defining genes ep.type <- opt$ep.type DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) # SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr motif.match.thr.str <- opt$motif.match.thr.str ## ## directories for factor motif enrichment ## FILENAME=opt$filename ## ## modify motif.enhancer.background input directory ##HERE: perhaps do a for loop for all the desired thresholds (use strsplit on enhancer.fimo.threshold) ## opt$motif.enhancer.background <- paste0(opt$motif.enhancer.background, opt$enhancer.fimo.threshold, "/fimo.formatted.tsv") # create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE, FGSEADIR, FGSEAFIG) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ###################################################################### ## load data cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } else { warning(paste0(cNMF.result.file, " does not exist")) } ## load motif enrichment results all.ttest.df.path <- paste0(OUTDIRSAMPLE,"/", ep.type, ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.", motif.match.thr.str, "_", SUBSCRIPT.SHORT,".txt") ttest.df <- read.delim(all.ttest.df.path, stringsAsFactors=F) # file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.factorMotifEnrichment.",SUBSCRIPT.SHORT,".RData") # print(file.name) # if(file.exists((file.name))) { # load(file.name) # print(paste0("loading ", file.name)) # } # motif.enrichment.variables <- c("all.enhancer.fisher.df", "all.promoter.fisher.df", # "promoter.wide", "enhancer.wide", "promoter.wide.binary", "enhancer.wide.binary", # "enhancer.wide.10en6", "enhancer.wide.binary.10en6", "all.enhancer.fisher.df.10en6", # "promoter.wide.10en6", "promoter.wide.binary.10en6", "all.promoter.fisher.df.10en6", # "all.promoter.ttest.df", "all.promoter.ttest.df.10en6", "all.enhancer.ttest.df", "all.enhancer.ttest.df.10en6") # motif.enrichment.variables.missing <- (!(motif.enrichment.variables %in% ls())) %>% as.numeric %>% sum # if ( motif.enrichment.variables.missing > 0 ) { # warning(paste0(motif.enrichment.variables[!(motif.enrichment.variables %in% ls())], " not available")) # } ## End of data loading ########################################################################## ## Plots ## volcano plots volcano.plot <- function(toplot, ep.type, ranking.type, label.type="") { if( label.type == "pos") { label <- toplot %>% subset(two.sided.p.adjust < fdr.thr & enrichment.log2fc > 0) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) } else { label <- toplot %>% subset(two.sided.p.adjust < fdr.thr) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) } t <- gsub("topic_", "", toplot$topic[1]) p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(two.sided.p.adjust))) + geom_point(size=0.5) + mytheme + ggtitle(paste0(SAMPLE[1], " Topic ", t, " Top ", num.top.genes, " ", ranking.type,"\n", ifelse(ep.type=="promoter", "Promoter", "Enhancer"), " Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(adjusted p-value)") + xlim(0,max(toplot$enrichment.log2fc)) + geom_hline(yintercept=-log10(fdr.thr), linetype="dashed", color="gray") + geom_text_repel(data=label, box.padding = 0.25, aes(label=motif.toshow), size=2.5, max.overlaps = 15, color="black")# + theme(text=element_text(size=16), axis.title=element_text(size=16), axis.text=element_text(size=16), plot.title=element_text(size=14)) print(p) p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(two.sided.p.value))) + geom_point(size=0.25) + mytheme + ggtitle(paste0(SAMPLE[1], " Topic ", t, " Top ", num.top.genes," ", ranking.type,"\n", ifelse(ep.type=="promoter", "Promoter", "Enhancer"), " Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(p-value)") + xlim(0,max(toplot$enrichment.log2fc)) + geom_hline(yintercept=-log10(fdr.thr), linetype="dashed", color="gray") + geom_text_repel(data=label, box.padding = 0.25, aes(label=motif.toshow), size=2.5, max.overlaps = 15, color="black") #+ theme(text=element_text(size=16), axis.title=element_text(size=16), axis.text=element_text(size=16), plot.title=element_text(size=14)) return(p) } ## function for all volcano plots all.volcano.plots <- function(all.fisher.df, ep.type, ranking.type, label.type="") { for ( t in 1:k ){ toplot <- all.fisher.df %>% subset(topic==paste0("topic_",t)) volcano.plot(toplot, ep.type, ranking.type, label.type) %>% print() } } ########################################################################## ## motif enrichment plot pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.count.ttest.enrichment_motif.thr.", motif.match.thr.str, ".pdf"), width=3, height=3) all.volcano.plots(get(paste0("ttest.df")) %>% subset(top.gene.mean != 0 & !grepl("X.NA.",motif)), ep.type, ranking.type="z-score") dev.off() |
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel","ramify", "ggpubr","gridExtra","RNOmni", "org.Hs.eg.db","limma","fgsea", "conflicted", "cluster","textshape","readxl", "ggdist", "gghalves", "Seurat", "writexl") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") source("./workflow/scripts/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/", help="Output directory"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210707_snakemake_maxParallel/all_genes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), make_option("--K.list", type="character", default="2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,19,21,23,25", help="K values available for analysis"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--raw.mtx.dir",type="character",default="stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), # the first lane: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.expandedMultiTargetGuide.RDS" make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/2kG.library.barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), make_option("--enhancer.fimo.threshold", type="character", default="1.0E-4", help="Enhancer fimo motif match threshold"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute"), make_option("--perturb.seq", type="character", default="False", help="True for perturb-seq. The pipeline will perform statistical test if True."), ## Organism flag make_option("--organism", type="character", default="human", help="Organism type, accept org.Hs.eg.db. Only support human and mouse.") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## 2n dataset (for sdev) ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/figures/all_genes/" ## opt$K.val <- 60 ## opt$sampleName <- "Perturb_2kG_dup4" ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/all_genes_acrossK/all_genes_acrossK/2kG.library/" ## opt$K.val <- 60 ## ## debug ctrl ## opt$topic.model.result.dir <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210810_snakemake_ctrls/all_genes_acrossK/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation" ## opt$sampleName <- "2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210810_snakemake_ctrls/figures/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210810_snakemake_ctrls/analysis/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation/all_genes/" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210806_curate_ctrl_mtx/outputs/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation.barcodes.tsv" ## opt$K.val <- 60 ## ## ctrl 2nd round ## opt$sampleName <- "2kG.library.ctrl.only" ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes_acrossK/2kG.library.ctrl.only/" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210806_curate_ctrl_mtx/211206_ctrl_mtx_for_cNMF_pipeline/outputs/ctrl_mtx/barcodes.tsv" ## opt$subsample.type <- "ctrl" ## opt$K.val <- 8 ## ## debug scRNAseq_2kG_11AMDox_1 ## opt$sampleName <- "scRNAseq_2kG_11AMDox_1" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/" ## opt$K.val <- 14 ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes_acrossK/scRNAseq_2kG_11AMDox_1" ## ## debug K562 gwps sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes_acrossK/WeissmanK562gwps/" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/WeissmanLab_data/K562_gwps_raw_singlecell_01_metadata.txt" ## opt$K.val <- 25 ## opt$sampleName <- "WeissmanK562gwps" ## ## debug mouse ENCODE heart sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes/" ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes_acrossK/mouse_ENCODE_heart/" ## opt$K.val <- 45 ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/collab_data/IGVF/mouse_ENCODE_adrenal/auxiliary_data/snrna/adrenal_Parse_10x_integrated_metadata.csv" ## sdev for mouse ENCODE ## ## debug IGVF b01_LeftCortex sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/" ## opt$topic.model.result.dir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes_acrossK/IGVF_b01_LeftCortex/" ## opt$K.val <- 15 ## opt$sampleName <- "IGVF_b01_LeftCortex" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_igvf_b01_LeftCortex_data/IGVF_b01_LeftCortex.barcodes.txt" ## opt$organism <- "mouse" mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir TMDIR=opt$topic.model.result.dir SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") ## FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") ## FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr ## ## directories for factor motif enrichment ## FILENAME=opt$filename ## ## modify motif.enhancer.background input directory ##HERE: perhaps do a for loop for all the desired thresholds (use strsplit on enhancer.fimo.threshold) ## opt$motif.enhancer.background <- paste0(opt$motif.enhancer.background, opt$enhancer.fimo.threshold, "/fimo.formatted.tsv") # create dir if not already check.dir <- c(OUTDIR, FIGDIR, OUTDIRSAMPLE, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) # selected.gene <- c("EDN1", "NOS3", "TP53", "GOSR2", "CDKN1A") # # ABC genes # gene.set <- c("INPP5B", "SF3A3", "SERPINH1", "NR2C1", "FGD6", "VEZT", "SMAD3", "AAGAB", "GOSR2", "ATP5G1", "ANGPTL4", "SRBD1", "PRKCE", "DAGLB") # ABC_0.015_CAD_pp.1_genes #200 gene library # # cell cycle genes # ## need to update these for 2kG library # gene.list.three.groups <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/ptbd.genes_three.groups.txt"), header=T, stringsAsFactors=F) # enhancer.set <- gene.list.three.groups$Gene[grep("E_at_", gene.list.three.groups$Gene)] # CAD.focus.gene.set <- gene.list.three.groups %>% subset(Group=="CAD_focus") %>% pull(Gene) %>% append(enhancer.set) # EC.pos.ctrl.gene.set <- gene.list.three.groups %>% subset(Group=="EC_pos._ctrls") %>% pull(Gene) cell.count.thr <- opt$cell.count.thr # greater than this number, filter to keep the guides with greater than this number of cells guide.count.thr <- opt$guide.count.thr # greater than this number, filter to keep the perturbations with greater than this number of guides # guide.design = read.delim(file=paste0(DATADIR, "/200607_ECPerturbSeqMiniPool.design.txt"), header=T, stringsAsFactors = F) # ## add GO pathway log2FC # GO <- read.delim(file=paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.table.brief.txt"), header=T, check.names=FALSE) # GO.list <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.list.brief.txt", header=T, check.names=F) # colnames(GO)[1] <- "Gene" # colnames(GO.list)[1] <- "Gene" # ## load all sample, K, topic's top 100 genes (by TopFeatures() KL-score measure) # ## allGeneKtopic100 <- read.delim(paste0(TMDIR, "no.plus.pooled.top100.topicStats.txt"), header=T) # # load non-expressed control gene list # non.expressed.genes <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/non.expressed.ctrl.genes.txt", header=F, stringsAsFactors=F) %>% unlist %>% as.character() %>% sort() # # perturbation type list # gene.set.type.df <- data.frame(Gene=guide.design %>% pull(guideSet) %>% unique(), # type=rep("other", guide.design %>% pull(guideSet) %>% unique() %>% length())) # gene.set.type.df$Gene <- gene.set.type.df$Gene %>% as.character() # gene.set.type.df$type <- gene.set.type.df$type %>% as.character() # gene.set.type.df$type[which(gene.set.type.df$Gene %in% non.expressed.genes)] <- "non-expressed" # gene.set.type.df$type[which(gene.set.type.df$Gene %in% CAD.focus.gene.set)] <- "CAD focus" # gene.set.type.df$type[grepl("^safe|^negative", gene.set.type.df$Gene)] <- "negative-control" # gene.set.type.df$Gene[which(gene.set.type.df$Gene == "negative_control")] <- "negative-control" # gene.set.type.df$Gene[which(gene.set.type.df$Gene == "safe_targeting")] <- "safe-targeting" # # gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # gene.set.type.df.200 <- gene.set.type.df # # reference table # ref.table <- read_xlsx(opt$reference.table, sheet="2000_gene_library_annotated") # gene.set.type.df <- ref.table %>% select(Symbol, `Class(es)`) %>% `colnames<-`(c("Gene", "type")) # gene.set.type.df$type[grepl("EC_ctrls", gene.set.type.df$type)] <- "EC_ctrls" # gene.set.type.df$type[grepl("NonExpressed", gene.set.type.df$type)] <- "non-expressed" # gene.set.type.df$type[grepl("abc.015", gene.set.type.df$type)] <- "ABC" # gene.set.type.df <- rbind(gene.set.type.df, c("negative-control", "negative-control"), c("safe-targeting", "safe-targeting")) # non.expressed.genes <- gene.set.type.df %>% subset(type == "non-expressed") %>% pull(Gene) # # ABC genes # gene.set <- gene.set.type.df %>% subset(grepl("ABC", type)) %>% pull(Gene) # ## add GWAS classification # modified.ref.table <- ref.table %>% mutate(GWAS.classification="") # CAD.index <- which(grepl("CAD_Loci",ref.table$`Class(es)`)) # EC_ctrls.index <- which(grepl("^EC_ctrls",ref.table$`Class(es)`)) # ABC_linked.index <- which(grepl("MIG_etc",ref.table$`Class(es)`)) # IBD.index <- which(grepl("Non-CAD_loci_IBD",ref.table$`Class(es)`)) # non.expressed.index <- which(grepl("NonExpressed",ref.table$`Class(es)`)) # poorly.annotated.9p21.index <- which(grepl("9p21",ref.table$`Class(es)`)) # # length(CAD.index) + length(EC_ctrls.index) + length(ABC_linked.index) + length(IBD.index) + length(non.expressed.index) + length(poorly.annotated.9p21.index) # modified.ref.table$GWAS.classification[ABC_linked.index] <- "ABC" # modified.ref.table$GWAS.classification[IBD.index] <- "IBD" # modified.ref.table$GWAS.classification[non.expressed.index] <- "NonExpressed" # modified.ref.table$GWAS.classification[poorly.annotated.9p21.index] <- "9p21.poorly.annotated" # modified.ref.table$GWAS.classification[EC_ctrls.index] <- "EC_ctrls" # modified.ref.table$GWAS.classification[CAD.index] <- "CAD" # modified.ref.table <- modified.ref.table %>% group_by(GWAS.classification) %>% mutate(gene.count.per.GWAS.category = n()) # ref.table <- modified.ref.table # ## add TSS distance to SNP # modified.ref.table <- ref.table %>% mutate(TSS.dist.to.SNP = abs(`TSS v. SNP loc`)) # not.in.SNP.index <- which(is.na(modified.ref.table$`TSS v. SNP loc`)) # modified.ref.table$TSS.dist.to.SNP[not.in.SNP.index] <- NA # ref.table <- modified.ref.table %>% ungroup() # ## add closest gene to top GWAS loci ranking # modified.ref.table <- ref.table %>% # group_by(`Top SNP ID`) %>% # per SNP metrics # arrange(abs(`TSS v. SNP loc`)) %>% # mutate(TSS.v.SNP.ranking = 1:n(), # total.gene.in.this.loci = n()) %>% ungroup() %>% # group_by(`Top SNP ID`, GWAS.classification) %>% # per SNP per GWAS class (CAD, IBD, NonExpressed, ABC, 9p21.poorly.annotated) # arrange(abs(`TSS v. SNP loc`)) %>% # mutate(TSS.v.SNP.ranking.in.GWAS.category = 1:n(), # total.gene.in.this.loci.in.GWAS.category = n()) %>% ungroup() # not.in.SNP.index <- which(is.na(modified.ref.table$`TSS v. SNP loc`)) # modified.ref.table$TSS.v.SNP.ranking.in.GWAS.category[not.in.SNP.index] <- NA # modified.ref.table$TSS.v.SNP.ranking[not.in.SNP.index] <- NA # ref.table <- modified.ref.table # ## add gene count per distance ranking per GWAS loci # modified.ref.table <- ref.table # modified.ref.table <- modified.ref.table %>% # group_by(TSS.v.SNP.ranking) %>% # per ranking, not considering which GWAS category the gene is from # mutate(total.TSS.v.SNP.ranking.count = n()) %>% ungroup() %>% # group_by(GWAS.classification, TSS.v.SNP.ranking.in.GWAS.category) %>% # per GWAS category and per ranking # mutate(total.TSS.v.SNP.ranking.count.per.GWAS.classification = n()) %>% ungroup() # not.in.SNP.index <- which(is.na(modified.ref.table$TSS.v.SNP.ranking)) # modified.ref.table$total.TSS.v.SNP.ranking.count[not.in.SNP.index] <- NA # modified.ref.table$total.TSS.v.SNP.ranking.count.per.GWAS.classification[not.in.SNP.index] <- NA # ref.table <- modified.ref.table # write.table(ref.table, file=paste0(opt$datadir, "/ref.table.txt"), row.names=F, quote=F, sep="\t") # ## ref.table ranking count summary table # ref.table.gene.to.SNP.dist.ranking.count.summary.allGWAS <- ref.table %>% select(TSS.v.SNP.ranking, total.TSS.v.SNP.ranking.count) %>% mutate(GWAS.classification="all") %>% unique() # ref.table.gene.to.SNP.dist.ranking.count.summary.indGWAS <- ref.table %>% select(TSS.v.SNP.ranking.in.GWAS.category, total.TSS.v.SNP.ranking.count.per.GWAS.classification, GWAS.classification) %>% `colnames<-`(c("TSS.v.SNP.ranking", "total.TSS.v.SNP.ranking.count", "GWAS.classification")) %>% unique() # ref.table.gene.to.SNP.dist.ranking.count.summary <- rbind(ref.table.gene.to.SNP.dist.ranking.count.summary.allGWAS, ref.table.gene.to.SNP.dist.ranking.count.summary.indGWAS) # ref.table.summary.na.index <- which(is.na(ref.table.gene.to.SNP.dist.ranking.count.summary$TSS.v.SNP.ranking)) # ref.table.gene.to.SNP.dist.ranking.count.summary <- ref.table.gene.to.SNP.dist.ranking.count.summary[-ref.table.summary.na.index,] # rm(ref.table.summary.na.index) # # convert enhancer SNP rs number to enhancer target gene name # need 2kG library version # enh.snp.to.gene <- read.delim(paste0(DATADIR, "/enhancer.SNP.to.gene.name.txt"), header=T, stringsAsFactors = F) %>% mutate(Enhancer_name=gsub("_","-", Enhancer_name)) # # gene corresponding pathway # gene.def.pathways <- read_excel(paste0(DATADIR,"topic.gene.definition.pathways.xlsx"), sheet="Gene_Pathway") # ## Gavin's new list # gene.classes.ranked <- read.table(paste0(opt$datadir, "Gene_Classes_Ranked_for_CAD_n_EC.txt"), header=T, stringsAsFactors = F) # summaries <- read.delim(paste0(opt$datadir, "Gene_Summaries_n_Classes.txt"), sep="\t", header=T, stringsAsFactors = F) # gene.summaries <- read_xlsx(paste0(opt$datadir, "Gene_Summaries.xlsx"), sheet="uniprot_summaries") # print("loaded all prerequisite data") ## for the guides that target multiple genes, we will split the gene annotation and duplicate the cell entry, so that each gene will get a cell read out. adjust.multiTargetGuide.rownames <- function(omega) { ## duplicate cells with guide that targets multiple genes cells.with.multiTargetGuide <- rownames(omega)[grepl("and",rownames(omega))] if(length(cells.with.multiTargetGuide) > 0) { ## split index by number of guide targets cells.with.multiTargetGuide.index <- which(grepl("and",rownames(omega))) cells.with.singleTargetGuide.index <- which(!grepl("and",rownames(omega))) ## get multi target gene names multiTarget.names <- cells.with.multiTargetGuide %>% strsplit(., split=":") %>% sapply("[[",1) ## full names: GeneA-and-GeneB multiTarget.Guide.CBC <- cells.with.multiTargetGuide %>% strsplit(., split=":") %>% sapply( function(x) paste0(x[[2]], ":", x[[3]]) ) multiTarget.names.1 <- multiTarget.names %>% strsplit(., split="-and-") %>% sapply ("[[",1) multiTarget.names.2 <- multiTarget.names %>% strsplit(., split="-and-") %>% sapply ("[[",2) multiTarget.names.all <- multiTarget.names.1 %>% append(multiTarget.names.2) %>% unique() ## get all the genes/enhancers that have guides targeting other gene/enhancer at the same time cells.with.multiTarget.gene.names.index <- which(grepl(paste0(multiTarget.names.all,collapse="|"), rownames(omega))) multiTarget.long.CBC.1 <- sapply(1:length(multiTarget.names), function(i) { paste0(multiTarget.names.1[i], "_multiTarget:", multiTarget.Guide.CBC[i]) }) multiTarget.long.CBC.2 <- sapply(1:length(multiTarget.names), function(i) { paste0(multiTarget.names.2[i], "_multiTarget:", multiTarget.Guide.CBC[i]) }) ## change original df's rownames multiTargetGuide.mtx <- omega[cells.with.multiTargetGuide.index,] multiTargetGuide.mtx.1 <- multiTargetGuide.mtx %>% `rownames<-`(multiTarget.long.CBC.1) multiTargetGuide.mtx.2 <- multiTargetGuide.mtx %>% `rownames<-`(multiTarget.long.CBC.2) ## pull cells with guides that has a single target, but the perturbed gene has multiTarget guide expanded.gene.name.df <- do.call(rbind, lapply(1:length(multiTarget.names.all), function(i) { gene.name.here <- multiTarget.names.all[i] toPaste.gene.name.here <- multiTarget.names[which(grepl(gene.name.here, multiTarget.names))] %>% gsub("-TSS2","",.) %>% unique() out <- do.call(rbind, lapply(1:length(toPaste.gene.name.here), function(j) { singleTarget.cell.index.here <- which(grepl(gene.name.here,rownames(omega)) & !grepl("and",rownames(omega))) singleTarget.cell.df <- omega[singleTarget.cell.index.here,] rownames(singleTarget.cell.df) <- gsub(gene.name.here,toPaste.gene.name.here[j],rownames(singleTarget.cell.df)) return(singleTarget.cell.df) })) return(out) })) # takes two minutes omega <- rbind(omega[cells.with.singleTargetGuide.index,], multiTargetGuide.mtx.1, multiTargetGuide.mtx.2, expanded.gene.name.df) } return(omega) } ###################################################################### ## Process topic model results ## for ( n in 1:length(SAMPLE) ) { ## if (SEP) { ## guideCounts <- loadGuides(n, sep=T) %>% mutate(Gene=Gene.marked) ## tmp.labels <- guideCounts$Gene %>% unique() %>% strsplit("-") %>% sapply("[[",2) %>% unique() ## tmp.labels <- tmp.labels[!(tmp.labels %in% c("control","targeting"))] ## rep1.label <- paste0("-",tmp.labels[1]) ## rep2.label <- paste0("-",tmp.labels[2]) ## } else guideCounts <- loadGuides(n) %>% mutate(Gene=Gene.marked) db <- ifelse(grepl("mouse|org.Mm.eg.db", opt$organism), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) ## load the appropriate database cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) print("finished loading cNMF result file") } else { theta.path <- paste0(TMDIR, "/", SAMPLE, ".gene_spectra_tpm.k_", k, ".dt_", DENSITY.THRESHOLD,".txt") theta.zscore.path <- paste0(TMDIR, "/", SAMPLE, ".gene_spectra_score.k_", k, ".dt_", DENSITY.THRESHOLD,".txt") median.spectra.path <- paste0(TMDIR, "/", SAMPLE, ".spectra.k_", k, ".dt_", DENSITY.THRESHOLD,".consensus.txt") print(theta.path) theta.raw <- read.delim(theta.path, header=T, stringsAsFactors=F, check.names=F, row.names=1) median.spectra <- read.delim(median.spectra.path, header=T, stringsAsFactors=F, check.names=F, row.names=1) ## theta.raw <- read.delim(theta.path, header=T, stringsAsFactors=F, check.names=F) %>% select(-``) print("finished reading raw weights for topics") tmp.theta <- theta.raw tmp.theta[tmp.theta==0] <- min(tmp.theta[tmp.theta > 0])/100 theta <- tmp.theta %>% apply(1, function(x) x/sum(x)) %>% `colnames<-`(c(1:k)) theta.raw <- theta.raw %>% t() %>% as.data.frame() %>% `colnames<-`(c(1:k)) median.spectra <- median.spectra %>% t() %>% as.data.frame %>% `colnames<-`(c(1:k)) print("loading topic z-score coefficient") theta.zscore <- read.delim(theta.zscore.path, header=T, stringsAsFactors=F, check.names=F, row.names=1) %>% t() %>% `colnames<-`(c(1:k)) tmp <- rownames(theta) %>% strsplit(., split=":") %>% sapply("[[",1) tmpp <- data.frame(table(tmp)) %>% subset(Freq > 1) # keep row names that have duplicated gene names but different ENSG names tmp.copy <- tmp tmp.copy[grepl(paste0(tmpp$tmp,collapse="|"),tmp)] <- rownames(theta)[grepl(paste0(tmpp$tmp,collapse="|"),rownames(theta))] rownames(theta) <- rownames(theta.raw) <- rownames(theta.zscore) <- tmp.copy ## median.spectra.names <- median.spectra %>% rownames %>% strsplit(split=":") %>% sapply(`[[`,1) ## rownames(median.spectra) <- median.spectra.names median.spectra.zscore <- apply(median.spectra, MARGIN=1, function(x) (x - mean(x)) / sd(x)) %>% t if(grepl("2kG.library", SAMPLE)) { median.spectra.zscore.df <- median.spectra.zscore %>% as.data.frame %>% mutate(Gene.full.name = rownames(.)) %>% separate(col="Gene.full.name", sep=":", remove=F, into = c("Gene", "ENSGID")) %>% melt(id.vars = c("Gene.full.name", "Gene", "ENSGID"), variable.name="ProgramID", value.name="median.spectra.zscore") %>% mutate(ProgramID = paste0("K", k, "_", ProgramID)) %>% as.data.frame %>% group_by(ProgramID) %>% arrange(desc(median.spectra.zscore)) %>% mutate(median.spectra.zscore.rank = 1:n()) %>% select(-Gene.full.name) %>% as.data.frame ## put median spectra zscore into ENSGID format for PoPS median.spectra.zscore.formatted <- median.spectra.zscore %>% as.data.frame %>% mutate(Gene.ENSGID = rownames(.)) %>% separate(col="Gene.ENSGID", sep=":", remove=F, into = c("Gene", "ENSGID_from_input")) %>% as.data.frame median.spectra.zscore.mappedENSGID <- mapIds(get(db), keys=median.spectra.zscore.formatted$Gene, keytype = "SYMBOL", column = "ENSEMBL") median.spectra.zscore.formatted <- median.spectra.zscore.formatted %>% mutate(ENSGID_mapped = median.spectra.zscore.mappedENSGID) %>% mutate(matchedENSGIDbool = ENSGID_from_input == ENSGID_mapped) median.spectra.zscore.formatted <- median.spectra.zscore.formatted %>% `rownames<-`(.$ENSGID_from_input) median.spectra.zscore.formatted <- median.spectra.zscore.formatted %>% select(-Gene, -ENSGID_from_input, -ENSGID_mapped, -matchedENSGIDbool, -Gene.ENSGID) %>% `colnames<-`(paste0("median_spectra_K", k, "_", colnames(.))) %>% as.data.frame print("save the data") ensembl.theta.zscore.names <- mapIds(get(db), keys = rownames(theta.zscore), keytype = "SYMBOL", column="ENSEMBL") ensembl.theta.zscore.names[ensembl.theta.zscore.names %>% is.na] <- rownames(theta.zscore)[ensembl.theta.zscore.names %>% is.na] theta.zscore.ensembl <- theta.zscore colnames(theta.zscore.ensembl) <- paste0("zscore_K", k, "_", colnames(theta.zscore.ensembl)) theta.zscore.ensembl <- theta.zscore.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_topic1")) ensembl.theta.raw.names <- mapIds(get(db), keys = rownames(theta.raw), keytype = "SYMBOL", column="ENSEMBL") ensembl.theta.raw.names[ensembl.theta.raw.names %>% is.na] <- rownames(theta.raw)[ensembl.theta.raw.names %>% is.na] theta.raw.ensembl <- theta.raw colnames(theta.raw.ensembl) <- paste0("tpm_K", k, "_topic", colnames(theta.raw.ensembl)) theta.raw.ensembl <- theta.raw.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.raw.names,.before=paste0("raw_K",k,"_topic1")) ## normalize to zero mean + unit variance theta.raw.ensembl.scaled <- theta.raw.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("tpm_K",k,"_1")) theta.zscore.ensembl.scaled <- theta.zscore.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_1")) median.spectra.zscore.formatted.scaled <- median.spectra.zscore.formatted %>% apply(2, scale) %>% as.data.frame %>% mutate(ENGSID = row.names(median.spectra.zscore.formatted), .before=paste0("median_spectra_K", k, "_1")) } else { ## detect gene data type (e.g. ENSGID, Entrez Symbol) gene.type <- ifelse(nrow(median.spectra.zscore) == sum(as.numeric(grepl("^ENS", median.spectra.zscore %>% rownames))), "ENSGID", "Gene") median.spectra.zscore.df <- median.spectra.zscore %>% as.data.frame %>% mutate(!!gene.type := rownames(.)) %>% melt(id.vars = c(gene.type), variable.name="ProgramID", value.name="median.spectra.zscore") %>% mutate(ProgramID = paste0("K", k, "_", ProgramID)) %>% as.data.frame %>% group_by(ProgramID) %>% arrange(desc(median.spectra.zscore)) %>% mutate(median.spectra.zscore.rank = 1:n()) %>% as.data.frame if(gene.type == "Gene") { ## put median spectra zscore into ENSGID format for PoPS mapped.genes <- mapIds(get(db), keys=median.spectra.zscore %>% rownames, keytype = "SYMBOL", column = "ENSEMBL") median.spectra.zscore.formatted <- median.spectra.zscore %>% as.data.frame %>% ## `colnames<-`(paste0("median_spectra_K", k, "_", colnames(.))) %>% mutate(Gene = rownames(.)) %>% mutate(ENSGID = mapped.genes) ## ensembl.theta.zscore.names <- mapIds(get(db), keys = rownames(theta.zscore), keytype = "SYMBOL", column="ENSEMBL") ## ensembl.theta.zscore.names[ensembl.theta.zscore.names %>% is.na] <- rownames(theta.zscore)[ensembl.theta.zscore.names %>% is.na] ## theta.zscore.ensembl <- theta.zscore ## colnames(theta.zscore.ensembl) <- paste0("zscore_K", k, "_", colnames(theta.zscore.ensembl)) ## theta.zscore.ensembl <- theta.zscore.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_1")) ## ensembl.theta.raw.names <- mapIds(get(db), keys = rownames(theta.raw), keytype = "SYMBOL", column="ENSEMBL") ## ensembl.theta.raw.names[ensembl.theta.raw.names %>% is.na] <- rownames(theta.raw)[ensembl.theta.raw.names %>% is.na] ## theta.raw.ensembl <- theta.raw ## colnames(theta.raw.ensembl) <- paste0("tpm_K", k, "_", colnames(theta.raw.ensembl)) ## theta.raw.ensembl <- theta.raw.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.raw.names,.before=paste0("tpm_K",k,"_1")) ## ## normalize to zero mean + unit variance ## theta.raw.ensembl.scaled <- theta.raw.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("tpm_K",k,"_1")) ## theta.zscore.ensembl.scaled <- theta.zscore.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_1")) ## median.spectra.zscore.formatted.scaled <- median.spectra.zscore.formatted %>% select(-ENSGID, -Gene) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENGSID = median.spectra.zscore.formatted$ENSGID, .before=paste0("median_spectra_K", k, "_1")) } else { mapped.genes <- mapIds(get(db), keys=median.spectra.zscore %>% rownames, keytype = "ENSEMBL", column = "SYMBOL") median.spectra.zscore.formatted <- median.spectra.zscore %>% as.data.frame %>% ## `colnames<-`(paste0("median_spectra_K", k, "_", colnames(.))) %>% mutate(ENSGID = rownames(.)) %>% mutate(Gene = mapped.genes) ## median.spectra.zscore.formatted.scaled <- median.spectra.zscore.formatted %>% select(-ENSGID, -Gene) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENGSID = median.spectra.zscore.formatted$ENSGID, .before=paste0("median_spectra_K", k, "_1")) } median.spectra.zscore.ensembl.names <- median.spectra.zscore.formatted$ENSGID median.spectra.zscore.formatted <- median.spectra.zscore.formatted %>% mutate(Gene_ENSGID = paste0(Gene, ":", ENSGID)) %>% `rownames<-`(.$Gene_ENSGID) %>% select(-Gene, -ENSGID, -Gene_ENSGID) %>% `colnames<-`(paste0("median_spectra_K", k, "_", colnames(.))) %>% as.data.frame if(gene.type == "Gene") { ensembl.theta.zscore.names <- mapIds(get(db), keys = rownames(theta.zscore), keytype = "SYMBOL", column="ENSEMBL") } else { ensembl.theta.zscore.names <- theta.zscore %>% rownames } ensembl.theta.zscore.names[ensembl.theta.zscore.names %>% is.na] <- rownames(theta.zscore)[ensembl.theta.zscore.names %>% is.na] theta.zscore.ensembl <- theta.zscore colnames(theta.zscore.ensembl) <- paste0("zscore_K", k, "_", colnames(theta.zscore.ensembl)) theta.zscore.ensembl <- theta.zscore.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_1")) if(gene.type == "Gene") { ensembl.theta.raw.names <- mapIds(get(db), keys = rownames(theta.raw), keytype = "SYMBOL", column="ENSEMBL") } else { ensembl.theta.raw.names <- theta.raw %>% rownames } ensembl.theta.raw.names[ensembl.theta.raw.names %>% is.na] <- rownames(theta.raw)[ensembl.theta.raw.names %>% is.na] theta.raw.ensembl <- theta.raw colnames(theta.raw.ensembl) <- paste0("tpm_K", k, "_", colnames(theta.raw.ensembl)) theta.raw.ensembl <- theta.raw.ensembl %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.raw.names,.before=paste0("tpm_K",k,"_1")) ## normalize to zero mean + unit variance theta.raw.ensembl.scaled <- theta.raw.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("tpm_K",k,"_1")) theta.zscore.ensembl.scaled <- theta.zscore.ensembl %>% select(-ENSGID) %>% apply(2, scale) %>% as.data.frame %>% mutate(ENSGID=ensembl.theta.zscore.names,.before=paste0("zscore_K",k,"_1")) median.spectra.zscore.formatted.scaled <- median.spectra.zscore.formatted %>% apply(2, scale) %>% as.data.frame %>% mutate(ENGSID = median.spectra.zscore.ensembl.names, .before=paste0("median_spectra_K", k, "_1")) } ## truncate.theta.names <- function(theta) { ## theta.gene.names <- rownames(theta) %>% strsplit(., split=":") %>% sapply("[[",1) # remove ENSG names ## rownames(theta) <- theta.gene.names ## return(theta) ## } ## theta.raw <- truncate.theta.names(theta.raw) ## theta.zscore <- truncate.theta.names(theta.zscore) omega.path <- paste0(TMDIR, "/", SAMPLE, ".usages.k_", k, ".dt_", DENSITY.THRESHOLD, ".consensus.txt") print(omega.path) omega.original <- omega <- read.delim(omega.path, header=T, stringsAsFactors=F, check.names=F, row.names = 1) %>% apply(1, function(x) x/sum(x)) %>% t() colnames(omega) <- paste0("topic_",colnames(omega)) print("finished loading omega") barcode.names <- read.table(opt$barcode.names, header=T, stringsAsFactors=F, sep=ifelse(grepl("csv$", opt$barcode.names), ",", "\t")) ## %>% `colnames<-`("long.CBC") if(grepl("2kG.library", SAMPLE)) { barcode.names <- read.table(opt$barcode.names, header=F, stringsAsFactors=F) ## %>% `colnames<-`("long.CBC") rownames(omega) <- rownames(omega.original) <- barcode.names %>% `colnames<-`("long.CBC") %>% pull(long.CBC) %>% gsub("CSNK2B-and-CSNK2B", "CSNK2B",.) %>% gsub("[(]'", "", .) %>% gsub("',[)]", "", .) omega <- adjust.multiTargetGuide.rownames(omega) barcode.names <- data.frame(long.CBC=rownames(omega)) %>% mutate(long.CBC = gsub("CSNK2B-and-CSNK2B", "CSNK2B", long.CBC)) %>% separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% separate(col="CBC", into=c("CBC", "sample"), sep="-scRNAseq_2kG_", remove=F) %>% mutate(Gene = gsub("-TSS2$", "", Gene.full.name), CBC = gsub("RHOA-and-", "", CBC), Guide = gsub("RHOA-and-", "", Guide)) %>% as.data.frame } ## helper function to map between ENSGID and SYMBOL map.ENSGID.SYMBOL <- function(df) { ## need column `Gene` to be present in df ## detect gene data type (e.g. ENSGID, Entrez Symbol) gene.type <- ifelse(nrow(df) == sum(as.numeric(grepl("^ENS", df$Gene))), "ENSGID", "Gene") if(gene.type == "ENSGID") { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "ENSEMBL", column = "SYMBOL") df <- df %>% mutate(ENSGID = Gene, Gene = mapped.genes) } else { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "SYMBOL", column = "ENSEMBL") df <- df %>% mutate(ENSGID = mapped.genes) } return(df) } ## get list of topic defining genes theta.rank.list <- vector("list", ncol(theta.zscore))## initialize storage list for(i in 1:ncol(theta.zscore)) { topic <- paste0("topic_", colnames(theta.zscore)[i]) theta.rank.list[[i]] <- theta.zscore %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("topic.zscore") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.zscore), .before="topic.zscore") %>% mutate(zscore.specificity.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.rank.df <- do.call(rbind, theta.rank.list) %>% ## combine list to df `colnames<-`(c("topic.zscore", "Gene", "zscore.specificity.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## get list of topic genes by raw weight theta.raw.rank.list <- vector("list", ncol(theta.raw))## initialize storage list for(i in 1:ncol(theta.raw)) { topic <- paste0("topic_", colnames(theta.raw)[i]) theta.raw.rank.list[[i]] <- theta.raw %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("topic.raw") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.raw), .before="topic.raw") %>% mutate(raw.score.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.raw.rank.df <- do.call(rbind, theta.raw.rank.list) %>% ## combine list to df `colnames<-`(c("topic.raw", "Gene", "raw.score.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## get list of topic genes by median spectra weight median.spectra.rank.list <- vector("list", ncol(median.spectra))## initialize storage list for(i in 1:ncol(median.spectra)) { topic <- paste0("topic_", colnames(median.spectra)[i]) median.spectra.rank.list[[i]] <- median.spectra %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("median.spectra") %>% mutate(Gene = rownames(.)) %>% arrange(desc(median.spectra), .before="median.spectra") %>% mutate(median.spectra.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } median.spectra.rank.df <- do.call(rbind, median.spectra.rank.list) %>% ## combine list to df `colnames<-`(c("median.spectra", "Gene", "median.spectra.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL ## median.spectra.zscore.df <- median.spectra.zscore.df %>% mutate(Gene = ENSGID) ## quick fix, need to add "Gene" column to this dataframe in analysis script write.table(theta.zscore, file=paste0(OUTDIRSAMPLE, "/topic.zscore_",SUBSCRIPT.SHORT, ".txt"), row.names=T, quote=F, sep="\t") write.table(theta.raw, file=paste0(OUTDIRSAMPLE, "/topic.tpm.score_",SUBSCRIPT.SHORT, ".txt"), row.names=T, quote=F, sep="\t") write.table(theta.zscore.ensembl, file=paste0(OUTDIRSAMPLE, "/topic.zscore.ensembl_",SUBSCRIPT.SHORT, ".txt"), row.names=F, quote=F, sep="\t") write.table(theta.raw.ensembl, file=paste0(OUTDIRSAMPLE, "/topic.tpm.ensembl_",SUBSCRIPT.SHORT, ".txt"), row.names=F, quote=F, sep="\t") write.table(theta.zscore.ensembl.scaled, file=paste0(OUTDIRSAMPLE, "/topic.zscore.ensembl.scaled_", SUBSCRIPT.SHORT, ".txt"), row.names=F, quote=F, sep = "\t") write.table(theta.raw.ensembl.scaled, file=paste0(OUTDIRSAMPLE, "/topic.tpm.ensembl.scaled_", SUBSCRIPT.SHORT, ".txt"), row.names=F, quote=F, sep = "\t") write.table(median.spectra.zscore.df, file=paste0(OUTDIRSAMPLE, "/median.spectra.zscore.df_", SUBSCRIPT.SHORT, ".txt"), sep="\t", quote=F, row.names=F) write.table(median.spectra.zscore.formatted.scaled, file=paste0(OUTDIRSAMPLE, "/median.spectra.zscore.ensembl.scaled_", SUBSCRIPT.SHORT, ".txt"), sep="\t", quote=F, row.names=F) save(theta, theta.raw, theta.raw.rank.df, theta.zscore, median.spectra.zscore.df, median.spectra, median.spectra.rank.df, omega, theta.path, omega.path, median.spectra.path, barcode.names, file=cNMF.result.file) print("finished writing all tables") } print("finished analysis script") # # modify GO.list if "pooled" # if (SEP) { # tmp.plus <- GO.list %>% mutate(Gene = paste0(GO.list$Gene,rep1.label), Pathway = paste0(GO.list$Pathway,rep1.label)) # tmp.no <- GO.list %>% mutate(Gene = paste0(GO.list$Gene,rep2.label), Pathway = paste0(GO.list$Pathway,rep2.label)) # GO.list <- rbind(tmp.no, tmp.plus) # } |
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 | library(conflicted) conflict_prefer("Position", "base") packages <- c("optparse","dplyr", "ggplot2", "reshape2", "ggrepel", "conflicted", "gplots", "org.Hs.eg.db") ## library(Seurat) xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), # make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210625_snakemake_output/top3000VariableGenes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), # make_option("--K.list", type="character", default="2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,19,21,23,25", help="K values available for analysis"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), # make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), # make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), # make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), # make_option("--raw.mtx.dir",type="character",default="stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), # make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), # the first lane: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.expandedMultiTargetGuide.RDS" # make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), # make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/barcodes.tsv", help="barcodes.tsv for all cells"), # make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), # make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), # make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), # make_option("--enhancer.fimo.threshold", type="character", default="1.0E-4", help="Enhancer fimo motif match threshold"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 ## ## mouse ENCODE adrenal data sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230117_snakemake_mouse_ENCODE_adrenal/figures/top2000VariableGenes/" ## opt$sampleName <- "mouse_ENCODE_adrenal" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230117_snakemake_mouse_ENCODE_adrenal/analysis/top2000VariableGenes" ## opt$K.val <- 60 ## ## mouse ENCODE heart data sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/figures/top2000VariableGenes/" ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes" ## opt$K.val <- 15 ## ## K562 gwps sdev ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$sampleName <- "WeissmanK562gwps" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$K.val <- 20 mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) mytheme <- theme_classic() + theme(axis.text = element_text(size = 7), axis.title = element_text(size = 8), plot.title = element_text(hjust = 0.5, face = "bold", size=10), axis.line = element_line(color = "black", size = 0.25), axis.ticks = element_line(color = "black", size = 0.25), legend.key.size = unit(10, units="pt"), legend.text = element_text(size=7), legend.title = element_text(size=8) ) SAMPLE=strsplit(opt$sampleName,",") %>% unlist() # STATIC.SAMPLE=c("Telo_no_IL1B_T200_1", "Telo_no_IL1B_T200_2", "Telo_plus_IL1B_T200_1", "Telo_plus_IL1B_T200_2", "no_IL1B", "plus_IL1B", "pooled") # DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir ## TMDIR=opt$topic.model.result.dir ## SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") ## FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") ## FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") message(FIGDIRTOP) ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) # SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr # create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ###################################################################### ## Process topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } else { warning(paste0(cNMF.result.file, " does not exist")) } ## End of data loading ## (patch up cNMF_analysis.R ENSGID to Gene conversion) ## gene mapping function map.ENSGID.SYMBOL <- function(topFeatures) { gene.type <- ifelse(sum(as.numeric(colnames(topFeatures) %in% "Gene")) > 0, ##(median.spectra.zscore.df) == sum(as.numeric(grepl("^ENS", median.spectra.zscore %>% rownames))), "Gene", "ENSGID") db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) if(gene.type == "Gene") { if (nrow(topFeatures) == sum(as.numeric(grepl("^ENS", topFeatures$Gene)))) { ## put median spectra zscore into ENSGID format for PoPS mapped.genes <- mapIds(get(db), keys=topFeatures$Gene, keytype = "ENSEMBL", column = "SYMBOL") topFeatures <- topFeatures %>% mutate(ENSGID = Gene) %>% as.data.frame %>% mutate(Gene = mapped.genes) na.index <- which(is.na(topFeatures$Gene)) if(length(na.index) > 0) topFeatures$Gene[na.index] <- topFeatures$ENSGID[na.index] } } else { mapped.genes <- mapIds(get(db), keys=topFeatures$ENSGID, keytype = "ENSEMBL", column = "SYMBOL") topFeatures <- topFeatures %>% as.data.frame %>% mutate(Gene = mapped.genes) } return(topFeatures) } ## end of ENSGID to Gene conversion ########################################################################## ## Plots ########################################################################## ## topic gene z-score list pdf(file=paste0(FIGDIRTOP,"top50GeneInTopics.zscore.pdf"), width=2.5, height=4.5) topFeatures.raw.weight <- theta.zscore %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene", variable.name="topic", value.name="scores") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col() + theme_minimal() p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("z-score (Specificity)") + ggtitle(paste(SAMPLE, ", Topic ", t, sep="")) + mytheme print(p) } dev.off() ########################################################################## ## Topic's top gene list, ranked by raw weight pdf(file=paste0(FIGDIRTOP,"top50GeneInTopics.rawWeight.pdf"), width=2.5, height=4.5) topFeatures <- theta %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(Gene), Score=topFeatures %>% subset(topic == t) %>% pull(scores)) p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col() + theme_minimal() p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("Raw Score (gene's weight in topic)") + ggtitle(paste(SAMPLE, ", Topic ", t, sep="")) + mytheme print(p) } dev.off() ########################################################################## ## raw program TPM list with annotataion pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.rawWeight.pdf"), width=2.5, height=3) topFeatures <- theta %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(Gene), Score=topFeatures %>% subset(topic == t) %>% pull(scores)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("Raw Score (gene's weight in topic)") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) print(p4) } dev.off() ########################################################################## ## raw program zscore list (top 10) (can potentially include annotation) pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.zscore.pdf"), width=2.5, height=3) topFeatures <- theta.zscore %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(Gene), Score=topFeatures %>% subset(topic == t) %>% pull(scores)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("z-score") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, "\nK = ", k, ", Topic ", t)) print(p4) } dev.off() ########################################################################## ## median spectra list (top 10) (can potentially include annotation) pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.median_spectra.zscore.pdf"), width=2.5, height=3) topFeatures <- median.spectra.zscore.df %>% as.data.frame() %>% group_by(ProgramID) %>% arrange(desc(median.spectra.zscore)) %>% slice(1:10) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { ProgramID.here <- paste0("K", k, "_", t) toPlot <- data.frame(Gene=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Gene), Score=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(median.spectra.zscore)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p7 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() p7 <- p7 + coord_flip() + xlab("Top 10 Genes") + ylab("z-score") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ",\nK = ", k, ", Program ", t, "\nMedian Spectra")) print(p7) } dev.off() ########################################################################## ## median spectra raw list (top 10) (can potentially include annotation) pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.median_spectra.raw.pdf"), width=2.5, height=3) topFeatures <- median.spectra %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene",value.name="Score", variable.name="ProgramID") %>% group_by(ProgramID) %>% arrange(desc(Score)) %>% slice(1:10) %>% mutate(ProgramID = paste0("K", k, "_", ProgramID)) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { ProgramID.here <- paste0("K", k, "_", t) toPlot <- data.frame(Gene=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Gene), Score=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Score)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p8 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() p8 <- p8 + coord_flip() + xlab("Top 10 Genes") + ylab("Weight in Program") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ",\nK = ", k, ", Program ", t, "\nMedian Spectra")) print(p8) } dev.off() ########################################################################## ## Topic's top gene list, ranked by median spectra zscore pdf(file=paste0(FIGDIRTOP,"top50GeneInProgram.median_spectra.zscore.pdf"), width=2.5, height=4.5) topFeatures <- median.spectra.zscore.df %>% as.data.frame() %>% group_by(ProgramID) %>% arrange(desc(median.spectra.zscore)) %>% slice(1:50) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { ProgramID.here <- paste0("K", k, "_", t) toPlot <- data.frame(Gene=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Gene), Score=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(median.spectra.zscore)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p9 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="gray30") + theme_minimal() p9 <- p9 + coord_flip() + xlab("Top 50 Genes") + ylab("z-score") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ",\nK = ", k, ", Program ", t, "\nMedian Spectra")) print(p9) } dev.off() ########################################################################## ## median spectra raw list (top 50) (can potentially include annotation) pdf(file=paste0(FIGDIRTOP,"top50GeneInProgram.median_spectra.raw.pdf"), width=2.5, height=4.5) topFeatures <- median.spectra %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene",value.name="Score", variable.name="ProgramID") %>% group_by(ProgramID) %>% arrange(desc(Score)) %>% slice(1:50) %>% mutate(ProgramID = paste0("K", k, "_", ProgramID)) %>% map.ENSGID.SYMBOL for ( t in 1:dim(theta)[2] ) { ProgramID.here <- paste0("K", k, "_", t) toPlot <- data.frame(Gene=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Gene), Score=topFeatures %>% subset(ProgramID == ProgramID.here) %>% pull(Score)) # %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" p10 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="gray30") + theme_minimal() p10 <- p10 + coord_flip() + xlab("Top 50 Genes") + ylab("Weight in Program") + mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ",\nK = ", k, ", Program ", t, "\nMedian Spectra")) print(p10) } dev.off() ########################################################################## ## topic Pearson correlation heatmap ## remove NA from theta.zscore tokeep <- (!is.na(theta.zscore)) %>% apply(1, sum) == k ## remove genes with NA ## why is there NA? d <- cor(theta.zscore[tokeep,], method="pearson") m <- as.matrix(d) ## Function for plotting heatmap # new version (adjusted font size) plotHeatmap <- function(mtx, labCol, title, margins=c(12,6), ...) { #original heatmap.2( mtx %>% t(), Rowv=T, Colv=T, trace='none', key=T, col=palette, labCol=labCol, margins=margins, cex.main=0.8, cexCol=4.8/sqrt(ncol(mtx)), cexRow=4.8/sqrt(ncol(mtx)), #4.8/sqrt(nrow(mtx)) ## cexCol=1/(ncol(mtx)^(1/3)), cexRow=1/(ncol(mtx)^(1/3)), #4.8/sqrt(nrow(mtx)) main=title, ... ) } pdf(file=paste0(FIGDIRTOP, "topic.Pearson.correlation.pdf")) plotHeatmap(m, labCol=rownames(m), margins=c(3,3), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 | import numpy as np import pandas as pd import os, errno import datetime import uuid import itertools import yaml import subprocess import scipy.sparse as sp from scipy.spatial.distance import squareform from sklearn.decomposition import non_negative_factorization from sklearn.cluster import KMeans from sklearn.metrics import silhouette_score from sklearn.utils import sparsefuncs from fastcluster import linkage from scipy.cluster.hierarchy import leaves_list import matplotlib.pyplot as plt import scanpy as sc def save_df_to_npz(obj, filename): np.savez_compressed(filename, data=obj.values, index=obj.index.values, columns=obj.columns.values) def save_df_to_text(obj, filename): obj.to_csv(filename, sep='\t') def load_df_from_npz(filename): with np.load(filename, allow_pickle=True) as f: obj = pd.DataFrame(**f) return obj def check_dir_exists(path): """ Checks if directory already exists or not and creates it if it doesn't """ try: os.makedirs(path) except OSError as exception: if exception.errno != errno.EEXIST: raise def worker_filter(iterable, worker_index, total_workers): return (p for i,p in enumerate(iterable) if (i-worker_index)%total_workers==0) def fast_euclidean(mat): D = mat.dot(mat.T) squared_norms = np.diag(D).copy() D *= -2.0 D += squared_norms.reshape((-1,1)) D += squared_norms.reshape((1,-1)) D = np.sqrt(D) D[D < 0] = 0 return squareform(D, checks=False) def fast_ols_all_cols(X, Y): pinv = np.linalg.pinv(X) beta = np.dot(pinv, Y) return(beta) def fast_ols_all_cols_df(X,Y): beta = fast_ols_all_cols(X, Y) beta = pd.DataFrame(beta, index=X.columns, columns=Y.columns) return(beta) def var_sparse_matrix(X): mean = np.array(X.mean(axis=0)).reshape(-1) Xcopy = X.copy() Xcopy.data **= 2 var = np.array(Xcopy.mean(axis=0)).reshape(-1) - (mean**2) return(var) def get_highvar_genes_sparse(expression, expected_fano_threshold=None, minimal_mean=0.5, numgenes=None): # Find high variance genes within those cells gene_mean = np.array(expression.mean(axis=0)).astype(float).reshape(-1) E2 = expression.copy(); E2.data **= 2; gene2_mean = np.array(E2.mean(axis=0)).reshape(-1) gene_var = pd.Series(gene2_mean - (gene_mean**2)) del(E2) gene_mean = pd.Series(gene_mean) gene_fano = gene_var / gene_mean # Find parameters for expected fano line top_genes = gene_mean.sort_values(ascending=False)[:20].index A = (np.sqrt(gene_var)/gene_mean)[top_genes].min() w_mean_low, w_mean_high = gene_mean.quantile([0.10, 0.90]) w_fano_low, w_fano_high = gene_fano.quantile([0.10, 0.90]) winsor_box = ((gene_fano > w_fano_low) & (gene_fano < w_fano_high) & (gene_mean > w_mean_low) & (gene_mean < w_mean_high)) fano_median = gene_fano[winsor_box].median() B = np.sqrt(fano_median) gene_expected_fano = (A**2)*gene_mean + (B**2) fano_ratio = (gene_fano/gene_expected_fano) # Identify high var genes if numgenes is not None: highvargenes = fano_ratio.sort_values(ascending=False).index[:numgenes] high_var_genes_ind = fano_ratio.index.isin(highvargenes) T=None else: if not expected_fano_threshold: T = (1. + gene_counts_fano[winsor_box].std()) else: T = expected_fano_threshold high_var_genes_ind = (fano_ratio > T) & (gene_counts_mean > minimal_mean) gene_counts_stats = pd.DataFrame({ 'mean': gene_mean, 'var': gene_var, 'fano': gene_fano, 'expected_fano': gene_expected_fano, 'high_var': high_var_genes_ind, 'fano_ratio': fano_ratio }) gene_fano_parameters = { 'A': A, 'B': B, 'T':T, 'minimal_mean': minimal_mean, } return(gene_counts_stats, gene_fano_parameters) def get_highvar_genes(input_counts, expected_fano_threshold=None, minimal_mean=0.5, numgenes=None): # Find high variance genes within those cells gene_counts_mean = pd.Series(input_counts.mean(axis=0).astype(float)) gene_counts_var = pd.Series(input_counts.var(ddof=0, axis=0).astype(float)) gene_counts_fano = pd.Series(gene_counts_var/gene_counts_mean) # Find parameters for expected fano line top_genes = gene_counts_mean.sort_values(ascending=False)[:20].index A = (np.sqrt(gene_counts_var)/gene_counts_mean)[top_genes].min() w_mean_low, w_mean_high = gene_counts_mean.quantile([0.10, 0.90]) w_fano_low, w_fano_high = gene_counts_fano.quantile([0.10, 0.90]) winsor_box = ((gene_counts_fano > w_fano_low) & (gene_counts_fano < w_fano_high) & (gene_counts_mean > w_mean_low) & (gene_counts_mean < w_mean_high)) fano_median = gene_counts_fano[winsor_box].median() B = np.sqrt(fano_median) gene_expected_fano = (A**2)*gene_counts_mean + (B**2) fano_ratio = (gene_counts_fano/gene_expected_fano) # Identify high var genes if numgenes is not None: highvargenes = fano_ratio.sort_values(ascending=False).index[:numgenes] high_var_genes_ind = fano_ratio.index.isin(highvargenes) T=None else: if not expected_fano_threshold: T = (1. + gene_counts_fano[winsor_box].std()) else: T = expected_fano_threshold high_var_genes_ind = (fano_ratio > T) & (gene_counts_mean > minimal_mean) gene_counts_stats = pd.DataFrame({ 'mean': gene_counts_mean, 'var': gene_counts_var, 'fano': gene_counts_fano, 'expected_fano': gene_expected_fano, 'high_var': high_var_genes_ind, 'fano_ratio': fano_ratio }) gene_fano_parameters = { 'A': A, 'B': B, 'T':T, 'minimal_mean': minimal_mean, } return(gene_counts_stats, gene_fano_parameters) def compute_tpm(input_counts): """ Default TPM normalization """ tpm = input_counts.copy() sc.pp.normalize_per_cell(tpm, counts_per_cell_after=1e6) return(tpm) class cNMF(): def __init__(self, output_dir=".", name=None): """ Parameters ---------- output_dir : path, optional (default=".") Output directory for analysis files. name : string, optional (default=None) A name for this analysis. Will be prefixed to all output files. If set to None, will be automatically generated from date (and random string). """ self.output_dir = output_dir if name is None: now = datetime.datetime.now() rand_hash = uuid.uuid4().hex[:6] name = '%s_%s' % (now.strftime("%Y_%m_%d"), rand_hash) self.name = name self.paths = None def _initialize_dirs(self): if self.paths is None: # Check that output directory exists, create it if needed. check_dir_exists(self.output_dir) check_dir_exists(os.path.join(self.output_dir, self.name)) check_dir_exists(os.path.join(self.output_dir, self.name, 'cnmf_tmp')) self.paths = { 'normalized_counts' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.norm_counts.h5ad'), 'nmf_replicate_parameters' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.nmf_params.df.npz'), 'nmf_run_parameters' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.nmf_idvrun_params.yaml'), 'nmf_genes_list' : os.path.join(self.output_dir, self.name, self.name+'.overdispersed_genes.txt'), 'tpm' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.tpm.h5ad'), 'tpm_stats' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.tpm_stats.df.npz'), 'iter_spectra' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.iter_%d.df.npz'), 'iter_usages' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.usages.k_%d.iter_%d.df.npz'), 'merged_spectra': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.merged.df.npz'), 'local_density_cache': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.local_density_cache.k_%d.merged.df.npz'), 'consensus_spectra': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.dt_%s.consensus.df.npz'), 'consensus_spectra__txt': os.path.join(self.output_dir, self.name, self.name+'.spectra.k_%d.dt_%s.consensus.txt'), 'consensus_usages': os.path.join(self.output_dir, self.name, 'cnmf_tmp',self.name+'.usages.k_%d.dt_%s.consensus.df.npz'), 'consensus_usages__txt': os.path.join(self.output_dir, self.name, self.name+'.usages.k_%d.dt_%s.consensus.txt'), 'consensus_stats': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.stats.k_%d.dt_%s.df.npz'), 'clustering_plot': os.path.join(self.output_dir, self.name, self.name+'.clustering.k_%d.dt_%s.png'), 'gene_spectra_score': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.gene_spectra_score.k_%d.dt_%s.df.npz'), 'gene_spectra_score__txt': os.path.join(self.output_dir, self.name, self.name+'.gene_spectra_score.k_%d.dt_%s.txt'), 'gene_spectra_tpm': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.gene_spectra_tpm.k_%d.dt_%s.df.npz'), 'gene_spectra_tpm__txt': os.path.join(self.output_dir, self.name, self.name+'.gene_spectra_tpm.k_%d.dt_%s.txt'), 'k_selection_plot' : os.path.join(self.output_dir, self.name, self.name+'.k_selection.png'), 'k_selection_stats' : os.path.join(self.output_dir, self.name, self.name+'.k_selection_stats.df.npz'), } def get_norm_counts(self, counts, tpm, high_variance_genes_filter = None, num_highvar_genes = None ): """ Parameters ---------- counts : anndata.AnnData Scanpy AnnData object (cells x genes) containing raw counts. Filtered such that no genes or cells with 0 counts tpm : anndata.AnnData Scanpy AnnData object (cells x genes) containing tpm normalized data matching counts high_variance_genes_filter : np.array, optional (default=None) A pre-specified list of genes considered to be high-variance. Only these genes will be used during factorization of the counts matrix. Must match the .var index of counts and tpm. If set to None, high-variance genes will be automatically computed, using the parameters below. num_highvar_genes : int, optional (default=None) Instead of providing an array of high-variance genes, identify this many most overdispersed genes for filtering Returns ------- normcounts : anndata.AnnData, shape (cells, num_highvar_genes) A counts matrix containing only the high variance genes and with columns (genes)normalized to unit variance """ if high_variance_genes_filter is None: ## Get list of high-var genes if one wasn't provided if sp.issparse(tpm.X): (gene_counts_stats, gene_fano_params) = get_highvar_genes_sparse(tpm.X, numgenes=num_highvar_genes) else: (gene_counts_stats, gene_fano_params) = get_highvar_genes(np.array(tpm.X), numgenes=num_highvar_genes) high_variance_genes_filter = list(tpm.var.index[gene_counts_stats.high_var.values]) ## Subset out high-variance genes norm_counts = counts[:, high_variance_genes_filter] ## Scale genes to unit variance if sp.issparse(tpm.X): sc.pp.scale(norm_counts, zero_center=False) if np.isnan(norm_counts.X.data).sum() > 0: print('Warning NaNs in normalized counts matrix') else: norm_counts.X /= norm_counts.X.std(axis=0, ddof=1) if np.isnan(norm_counts.X).sum().sum() > 0: print('Warning NaNs in normalized counts matrix') ## Save a \n-delimited list of the high-variance genes used for factorization open(self.paths['nmf_genes_list'], 'w').write('\n'.join(high_variance_genes_filter)) ## Check for any cells that have 0 counts of the overdispersed genes zerocells = norm_counts.X.sum(axis=1)==0 if zerocells.sum()>0: examples = norm_counts.obs.index[zerocells] print('Warning: %d cells have zero counts of overdispersed genes. E.g. %s' % (zerocells.sum(), examples[0])) print('Consensus step may not run when this is the case') return(norm_counts) def save_norm_counts(self, norm_counts): self._initialize_dirs() sc.write(self.paths['normalized_counts'], norm_counts) def get_nmf_iter_params(self, ks, n_iter = 100, random_state_seed = None, beta_loss = 'kullback-leibler'): """ Create a DataFrame with parameters for NMF iterations. Parameters ---------- ks : integer, or list-like. Number of topics (components) for factorization. Several values can be specified at the same time, which will be run independently. n_iter : integer, optional (defailt=100) Number of iterations for factorization. If several ``k`` are specified, this many iterations will be run for each value of ``k``. random_state_seed : int or None, optional (default=None) Seed for sklearn random state. """ if type(ks) is int: ks = [ks] # Remove any repeated k values, and order. k_list = sorted(set(list(ks))) n_runs = len(ks)* n_iter np.random.seed(seed=random_state_seed) nmf_seeds = np.random.randint(low=1, high=(2**32)-1, size=n_runs) replicate_params = [] for i, (k, r) in enumerate(itertools.product(k_list, range(n_iter))): replicate_params.append([k, r, nmf_seeds[i]]) replicate_params = pd.DataFrame(replicate_params, columns = ['n_components', 'iter', 'nmf_seed']) _nmf_kwargs = dict( alpha=0.0, l1_ratio=0.0, beta_loss=beta_loss, solver='mu', tol=1e-4, max_iter=1000, regularization=None, init='random' ) ## Coordinate descent is faster than multiplicative update but only works for frobenius if beta_loss == 'frobenius': _nmf_kwargs['solver'] = 'cd' return(replicate_params, _nmf_kwargs) def save_nmf_iter_params(self, replicate_params, run_params): self._initialize_dirs() save_df_to_npz(replicate_params, self.paths['nmf_replicate_parameters']) with open(self.paths['nmf_run_parameters'], 'w') as F: yaml.dump(run_params, F) def _nmf(self, X, nmf_kwargs): """ Parameters ---------- X : pandas.DataFrame, Normalized counts dataFrame to be factorized. nmf_kwargs : dict, Arguments to be passed to ``non_negative_factorization`` """ (usages, spectra, niter) = non_negative_factorization(X, **nmf_kwargs) return(spectra, usages) def run_nmf(self, worker_i=1, total_workers=1, ): """ Iteratively run NMF with prespecified parameters. Use the `worker_i` and `total_workers` parameters for parallelization. Generic kwargs for NMF are loaded from self.paths['nmf_run_parameters'], defaults below:: ``non_negative_factorization`` default arguments: alpha=0.0 l1_ratio=0.0 beta_loss='kullback-leibler' solver='mu' tol=1e-4, max_iter=200 regularization=None init='random' random_state, n_components are both set by the prespecified self.paths['nmf_replicate_parameters']. Parameters ---------- norm_counts : pandas.DataFrame, Normalized counts dataFrame to be factorized. (Output of ``normalize_counts``) run_params : pandas.DataFrame, Parameters for NMF iterations. (Output of ``prepare_nmf_iter_params``) """ self._initialize_dirs() run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) norm_counts = sc.read(self.paths['normalized_counts']) _nmf_kwargs = yaml.load(open(self.paths['nmf_run_parameters']), Loader=yaml.FullLoader) jobs_for_this_worker = worker_filter(range(len(run_params)), worker_i, total_workers) for idx in jobs_for_this_worker: p = run_params.iloc[idx, :] print('[Worker %d]. Starting task %d.' % (worker_i, idx)) _nmf_kwargs['random_state'] = p['nmf_seed'] _nmf_kwargs['n_components'] = p['n_components'] (spectra, usages) = self._nmf(norm_counts.X, _nmf_kwargs) spectra = pd.DataFrame(spectra, index=np.arange(1, _nmf_kwargs['n_components']+1), columns=norm_counts.var.index) save_df_to_npz(spectra, self.paths['iter_spectra'] % (p['n_components'], p['iter'])) def combine_nmf(self, k, remove_individual_iterations=False): run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) print('Combining factorizations for k=%d.'%k) self._initialize_dirs() combined_spectra = None n_iter = sum(run_params.n_components==k) run_params_subset = run_params[run_params.n_components==k].sort_values('iter') spectra_labels = [] for i,p in run_params_subset.iterrows(): spectra = load_df_from_npz(self.paths['iter_spectra'] % (p['n_components'], p['iter'])) if combined_spectra is None: combined_spectra = np.zeros((n_iter, k, spectra.shape[1])) combined_spectra[p['iter'], :, :] = spectra.values for t in range(k): spectra_labels.append('iter%d_topic%d'%(p['iter'], t+1)) combined_spectra = combined_spectra.reshape(-1, combined_spectra.shape[-1]) combined_spectra = pd.DataFrame(combined_spectra, columns=spectra.columns, index=spectra_labels) save_df_to_npz(combined_spectra, self.paths['merged_spectra']%k) return combined_spectra def consensus(self, k, density_threshold_str='0.5', local_neighborhood_size = 0.30,show_clustering = False, skip_density_and_return_after_stats = False, close_clustergram_fig=True): merged_spectra = load_df_from_npz(self.paths['merged_spectra']%k) norm_counts = sc.read(self.paths['normalized_counts']) ##here210830 if norm_counts.X.dtype != np.float64: norm_counts.X = norm_counts.X.astype(np.float64) if skip_density_and_return_after_stats: density_threshold_str = '2' density_threshold_repl = density_threshold_str.replace('.', '_') density_threshold = float(density_threshold_str) n_neighbors = int(local_neighborhood_size * merged_spectra.shape[0]/k) # Rescale topics such to length of 1. l2_spectra = (merged_spectra.T/np.sqrt((merged_spectra**2).sum(axis=1))).T if not skip_density_and_return_after_stats: # Compute the local density matrix (if not previously cached) topics_dist = None if os.path.isfile(self.paths['local_density_cache'] % k): local_density = load_df_from_npz(self.paths['local_density_cache'] % k) else: # first find the full distance matrix topics_dist = squareform(fast_euclidean(l2_spectra.values)) # partition based on the first n neighbors partitioning_order = np.argpartition(topics_dist, n_neighbors+1)[:, :n_neighbors+1] # find the mean over those n_neighbors (excluding self, which has a distance of 0) distance_to_nearest_neighbors = topics_dist[np.arange(topics_dist.shape[0])[:, None], partitioning_order] local_density = pd.DataFrame(distance_to_nearest_neighbors.sum(1)/(n_neighbors), columns=['local_density'], index=l2_spectra.index) save_df_to_npz(local_density, self.paths['local_density_cache'] % k) del(partitioning_order) del(distance_to_nearest_neighbors) density_filter = local_density.iloc[:, 0] < density_threshold l2_spectra = l2_spectra.loc[density_filter, :] kmeans_model = KMeans(n_clusters=k, n_init=10, random_state=1) kmeans_model.fit(l2_spectra) kmeans_cluster_labels = pd.Series(kmeans_model.labels_+1, index=l2_spectra.index) # Find median usage for each gene across cluster median_spectra = l2_spectra.groupby(kmeans_cluster_labels).median() # Normalize median spectra to probability distributions. median_spectra = (median_spectra.T/median_spectra.sum(1)).T # Compute the silhouette score stability = silhouette_score(l2_spectra.values, kmeans_cluster_labels, metric='euclidean') # Obtain the reconstructed count matrix by re-fitting the usage matrix and computing the dot product: usage.dot(spectra) refit_nmf_kwargs = yaml.load(open(self.paths['nmf_run_parameters']), Loader=yaml.FullLoader) refit_nmf_kwargs.update(dict( n_components = k, H = median_spectra.values, update_H = False )) # change refit_nmf_kwargs['H'] data type to match with norm_counts.X's refit_nmf_kwargs['H'] = refit_nmf_kwargs['H'].astype(norm_counts.X.dtype) ##here210830 _, rf_usages = self._nmf(norm_counts.X, nmf_kwargs=refit_nmf_kwargs) rf_usages = pd.DataFrame(rf_usages, index=norm_counts.obs.index, columns=median_spectra.index) rf_pred_norm_counts = rf_usages.dot(median_spectra) # Compute prediction error as a frobenius norm if sp.issparse(norm_counts.X): prediction_error = ((norm_counts.X.todense() - rf_pred_norm_counts)**2).sum().sum() else: prediction_error = ((norm_counts.X - rf_pred_norm_counts)**2).sum().sum() consensus_stats = pd.DataFrame([k, density_threshold, stability, prediction_error], index = ['k', 'local_density_threshold', 'stability', 'prediction_error'], columns = ['stats']) if skip_density_and_return_after_stats: return consensus_stats save_df_to_npz(median_spectra, self.paths['consensus_spectra']%(k, density_threshold_repl)) save_df_to_npz(rf_usages, self.paths['consensus_usages']%(k, density_threshold_repl)) save_df_to_npz(consensus_stats, self.paths['consensus_stats']%(k, density_threshold_repl)) save_df_to_text(median_spectra, self.paths['consensus_spectra__txt']%(k, density_threshold_repl)) save_df_to_text(rf_usages, self.paths['consensus_usages__txt']%(k, density_threshold_repl)) # Compute gene-scores for each GEP by regressing usage on Z-scores of TPM tpm = sc.read(self.paths['tpm']) tpm_stats = load_df_from_npz(self.paths['tpm_stats']) if sp.issparse(tpm.X): norm_tpm = (np.array(tpm.X.todense()) - tpm_stats['__mean'].values) / tpm_stats['__std'].values else: norm_tpm = (tpm.X - tpm_stats['__mean'].values) / tpm_stats['__std'].values # if norm_tpm.dtype != np.float64: # norm_tpm = norm_tpm.astype(np.float64) usage_coef = fast_ols_all_cols(rf_usages.values, norm_tpm) usage_coef = pd.DataFrame(usage_coef, index=rf_usages.columns, columns=tpm.var.index) save_df_to_npz(usage_coef, self.paths['gene_spectra_score']%(k, density_threshold_repl)) save_df_to_text(usage_coef, self.paths['gene_spectra_score__txt']%(k, density_threshold_repl)) # Convert spectra to TPM units, and obtain results for all genes by running last step of NMF # with usages fixed and TPM as the input matrix norm_usages = rf_usages.div(rf_usages.sum(axis=1), axis=0) refit_nmf_kwargs.update(dict( H = norm_usages.T.values, )) # Needed otherwise _nmf will crash because with inconsistent dtypes if tpm.X.dtype != np.float64: tpm.X = tpm.X.astype(np.float64) _, spectra_tpm = self._nmf(tpm.X.T, nmf_kwargs=refit_nmf_kwargs) spectra_tpm = pd.DataFrame(spectra_tpm.T, index=rf_usages.columns, columns=tpm.var.index) save_df_to_npz(spectra_tpm, self.paths['gene_spectra_tpm']%(k, density_threshold_repl)) save_df_to_text(spectra_tpm, self.paths['gene_spectra_tpm__txt']%(k, density_threshold_repl)) if show_clustering: if topics_dist is None: topics_dist = squareform(fast_euclidean(l2_spectra.values)) # (l2_spectra was already filtered using the density filter) else: # (but the previously computed topics_dist was not!) topics_dist = topics_dist[density_filter.values, :][:, density_filter.values] spectra_order = [] for cl in sorted(set(kmeans_cluster_labels)): cl_filter = kmeans_cluster_labels==cl if cl_filter.sum() > 1: cl_dist = squareform(topics_dist[cl_filter, :][:, cl_filter]) cl_dist[cl_dist < 0] = 0 #Rarely get floating point arithmetic issues cl_link = linkage(cl_dist, 'average') cl_leaves_order = leaves_list(cl_link) spectra_order += list(np.where(cl_filter)[0][cl_leaves_order]) else: ## Corner case where a component only has one element spectra_order += list(np.where(cl_filter)[0]) from matplotlib import gridspec import matplotlib.pyplot as plt width_ratios = [0.5, 9, 0.5, 4, 1] height_ratios = [0.5, 9] fig = plt.figure(figsize=(sum(width_ratios), sum(height_ratios))) gs = gridspec.GridSpec(len(height_ratios), len(width_ratios), fig, 0.01, 0.01, 0.98, 0.98, height_ratios=height_ratios, width_ratios=width_ratios, wspace=0, hspace=0) dist_ax = fig.add_subplot(gs[1,1], xscale='linear', yscale='linear', xticks=[], yticks=[],xlabel='', ylabel='', frameon=True) D = topics_dist[spectra_order, :][:, spectra_order] dist_im = dist_ax.imshow(D, interpolation='none', cmap='viridis', aspect='auto', rasterized=True) left_ax = fig.add_subplot(gs[1,0], xscale='linear', yscale='linear', xticks=[], yticks=[], xlabel='', ylabel='', frameon=True) left_ax.imshow(kmeans_cluster_labels.values[spectra_order].reshape(-1, 1), interpolation='none', cmap='Spectral', aspect='auto', rasterized=True) top_ax = fig.add_subplot(gs[0,1], xscale='linear', yscale='linear', xticks=[], yticks=[], xlabel='', ylabel='', frameon=True) top_ax.imshow(kmeans_cluster_labels.values[spectra_order].reshape(1, -1), interpolation='none', cmap='Spectral', aspect='auto', rasterized=True) hist_gs = gridspec.GridSpecFromSubplotSpec(3, 1, subplot_spec=gs[1, 3], wspace=0, hspace=0) hist_ax = fig.add_subplot(hist_gs[0,0], xscale='linear', yscale='linear', xlabel='', ylabel='', frameon=True, title='Local density histogram') hist_ax.hist(local_density.values, bins=np.linspace(0, 1, 50)) hist_ax.yaxis.tick_right() xlim = hist_ax.get_xlim() ylim = hist_ax.get_ylim() if density_threshold < xlim[1]: hist_ax.axvline(density_threshold, linestyle='--', color='k') hist_ax.text(density_threshold + 0.02, ylim[1] * 0.95, 'filtering\nthreshold\n\n', va='top') hist_ax.set_xlim(xlim) hist_ax.set_xlabel('Mean distance to k nearest neighbors\n\n%d/%d (%.0f%%) spectra above threshold\nwere removed prior to clustering'%(sum(~density_filter), len(density_filter), 100*(~density_filter).mean())) ## Add colorbar cbar_gs = gridspec.GridSpecFromSubplotSpec(8, 1, subplot_spec=hist_gs[1, 0], wspace=0, hspace=0) cbar_ax = fig.add_subplot(cbar_gs[4,0], xscale='linear', yscale='linear', xlabel='', ylabel='', frameon=True, title='Euclidean Distance') vmin = D.min().min() vmax = D.max().max() fig.colorbar(dist_im, cax=cbar_ax, ticks=np.linspace(vmin, vmax, 3), orientation='horizontal') #hist_ax.hist(local_density.values, bins=np.linspace(0, 1, 50)) #hist_ax.yaxis.tick_right() fig.savefig(self.paths['clustering_plot']%(k, density_threshold_repl), dpi=250) if close_clustergram_fig: plt.close(fig) def k_selection_plot(self, close_fig=True): ''' Borrowed from Alexandrov Et Al. 2013 Deciphering Mutational Signatures publication in Cell Reports ''' run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) stats = [] for k in sorted(set(run_params.n_components)): stats.append(self.consensus(k, skip_density_and_return_after_stats=True).stats) stats = pd.DataFrame(stats) stats.reset_index(drop = True, inplace = True) save_df_to_npz(stats, self.paths['k_selection_stats']) fig = plt.figure(figsize=(6, 4)) ax1 = fig.add_subplot(111) ax2 = ax1.twinx() ax1.plot(stats.k, stats.stability, 'o-', color='b') ax1.set_ylabel('Stability', color='b', fontsize=15) for tl in ax1.get_yticklabels(): tl.set_color('b') #ax1.set_xlabel('K', fontsize=15) ax2.plot(stats.k, stats.prediction_error, 'o-', color='r') ax2.set_ylabel('Error', color='r', fontsize=15) for tl in ax2.get_yticklabels(): tl.set_color('r') ax1.set_xlabel('Number of Components', fontsize=15) ax1.grid('on') plt.tight_layout() fig.savefig(self.paths['k_selection_plot'], dpi=250) if close_fig: plt.close(fig) if __name__=="__main__": """ Example commands for now: output_dir="/Users/averes/Projects/Melton/Notebooks/2018/07-2018/cnmf_test/" python cnmf.py prepare --output-dir $output_dir \ --name test --counts /Users/averes/Projects/Melton/Notebooks/2018/07-2018/cnmf_test/test_data.df.npz \ -k 6 7 8 9 --n-iter 5 python cnmf.py factorize --name test --output-dir $output_dir THis can be parallelized as such: python cnmf.py factorize --name test --output-dir $output_dir --total-workers 2 --worker-index WORKER_INDEX (where worker_index starts with 0) python cnmf.py combine --name test --output-dir $output_dir python cnmf.py consensus --name test --output-dir $output_dir """ import sys, argparse parser = argparse.ArgumentParser() parser.add_argument('command', type=str, choices=['prepare', 'factorize', 'combine', 'consensus', 'k_selection_plot']) parser.add_argument('--name', type=str, help='[all] Name for analysis. All output will be placed in [output-dir]/[name]/...', nargs='?', default='cNMF') parser.add_argument('--output-dir', type=str, help='[all] Output directory. All output will be placed in [output-dir]/[name]/...', nargs='?', default='.') parser.add_argument('-c', '--counts', type=str, help='[prepare] Input (cell x gene) counts matrix as df.npz or tab delimited text file') parser.add_argument('-k', '--components', type=int, help='[prepare] Number of components (k) for matrix factorization. Several can be specified with "-k 8 9 10"', nargs='+') parser.add_argument('-n', '--n-iter', type=int, help='[prepare] Numper of factorization replicates', default=100) parser.add_argument('--total-workers', type=int, help='[all] Total number of workers to distribute jobs to', default=1) parser.add_argument('--seed', type=int, help='[prepare] Seed for pseudorandom number generation', default=None) parser.add_argument('--genes-file', type=str, help='[prepare] File containing a list of genes to include, one gene per line. Must match column labels of counts matrix.', default=None) parser.add_argument('--numgenes', type=int, help='[prepare] Number of high variance genes to use for matrix factorization.', default=2000) parser.add_argument('--tpm', type=str, help='[prepare] Pre-computed (cell x gene) TPM values as df.npz or tab separated txt file. If not provided TPM will be calculated automatically', default=None) parser.add_argument('--beta-loss', type=str, choices=['frobenius', 'kullback-leibler', 'itakura-saito'], help='[prepare] Loss function for NMF.', default='frobenius') parser.add_argument('--densify', dest='densify', help='[prepare] Treat the input data as non-sparse', action='store_true', default=False) parser.add_argument('--worker-index', type=int, help='[factorize] Index of current worker (the first worker should have index 0)', default=0) parser.add_argument('--local-density-threshold', type=str, help='[consensus] Threshold for the local density filtering. This string must convert to a float >0 and <=2', default='0.5') parser.add_argument('--local-neighborhood-size', type=float, help='[consensus] Fraction of the number of replicates to use as nearest neighbors for local density filtering', default=0.30) parser.add_argument('--show-clustering', dest='show_clustering', help='[consensus] Produce a clustergram figure summarizing the spectra clustering', action='store_true') args = parser.parse_args() cnmf_obj = cNMF(output_dir=args.output_dir, name=args.name) cnmf_obj._initialize_dirs() if args.command == 'prepare': if args.counts.endswith('.h5ad'): input_counts = sc.read(args.counts) else: ## Load txt or compressed dataframe and convert to scanpy object if args.counts.endswith('.npz'): input_counts = load_df_from_npz(args.counts) else: input_counts = pd.read_csv(args.counts, sep='\t', index_col=0) if args.densify: input_counts = sc.AnnData(X=input_counts.values, obs=pd.DataFrame(index=input_counts.index), var=pd.DataFrame(index=input_counts.columns)) else: input_counts = sc.AnnData(X=sp.csr_matrix(input_counts.values), obs=pd.DataFrame(index=input_counts.index), var=pd.DataFrame(index=input_counts.columns)) if sp.issparse(input_counts.X) & args.densify: input_counts.X = np.array(input_counts.X.todense()) if args.tpm is None: tpm = compute_tpm(input_counts) sc.write(cnmf_obj.paths['tpm'], tpm) elif args.tpm.endswith('.h5ad'): subprocess.call('cp %s %s' % (args.tpm, cnmf_obj.paths['tpm']), shell=True) tpm = sc.read(cnmf_obj.paths['tpm']) else: if args.tpm.endswith('.npz'): tpm = load_df_from_npz(args.tpm) else: tpm = pd.read_csv(args.tpm, sep='\t', index_col=0) if args.densify: tpm = sc.AnnData(X=tpm.values, obs=pd.DataFrame(index=tpm.index), var=pd.DataFrame(index=tpm.columns)) else: tpm = sc.AnnData(X=sp.csr_matrix(tpm.values), obs=pd.DataFrame(index=tpm.index), var=pd.DataFrame(index=tpm.columns)) sc.write(cnmf_obj.paths['tpm'], tpm) if sp.issparse(tpm.X): gene_tpm_mean = np.array(tpm.X.mean(axis=0)).reshape(-1) gene_tpm_stddev = var_sparse_matrix(tpm.X)**.5 else: gene_tpm_mean = np.array(tpm.X.mean(axis=0)).reshape(-1) gene_tpm_stddev = np.array(tpm.X.std(axis=0, ddof=0)).reshape(-1) input_tpm_stats = pd.DataFrame([gene_tpm_mean, gene_tpm_stddev], index = ['__mean', '__std']).T save_df_to_npz(input_tpm_stats, cnmf_obj.paths['tpm_stats']) if args.genes_file is not None: highvargenes = open(args.genes_file).read().rstrip().split('\n') else: highvargenes = None norm_counts = cnmf_obj.get_norm_counts(input_counts, tpm, num_highvar_genes=args.numgenes, high_variance_genes_filter=highvargenes) if norm_counts.X.dtype != np.float64: norm_counts.X = norm_counts.X.astype(np.float64) cnmf_obj.save_norm_counts(norm_counts) (replicate_params, run_params) = cnmf_obj.get_nmf_iter_params(ks=args.components, n_iter=args.n_iter, random_state_seed=args.seed, beta_loss=args.beta_loss) cnmf_obj.save_nmf_iter_params(replicate_params, run_params) elif args.command == 'factorize': cnmf_obj.run_nmf(worker_i=args.worker_index, total_workers=args.total_workers) elif args.command == 'combine': run_params = load_df_from_npz(cnmf_obj.paths['nmf_replicate_parameters']) if type(args.components) is int: ks = [args.components] elif args.components is None: ks = sorted(set(run_params.n_components)) else: ks = args.components for k in ks: cnmf_obj.combine_nmf(k) elif args.command == 'consensus': run_params = load_df_from_npz(cnmf_obj.paths['nmf_replicate_parameters']) if type(args.components) is int: ks = [args.components] elif args.components is None: ks = sorted(set(run_params.n_components)) else: ks = args.components for k in ks: merged_spectra = load_df_from_npz(cnmf_obj.paths['merged_spectra']%k) cnmf_obj.consensus(k, args.local_density_threshold, args.local_neighborhood_size, args.show_clustering) elif args.command == 'k_selection_plot': cnmf_obj.k_selection_plot() |
1
of
cNMF/cnmf.modified.py
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 | import numpy as np import pandas as pd import os, errno import datetime import uuid import itertools import yaml import subprocess import scipy.sparse as sp from scipy.spatial.distance import squareform from sklearn.decomposition import non_negative_factorization from sklearn.cluster import KMeans from sklearn.metrics import silhouette_score from sklearn.utils import sparsefuncs from fastcluster import linkage from scipy.cluster.hierarchy import leaves_list import matplotlib.pyplot as plt import scanpy as sc def save_df_to_npz(obj, filename): np.savez_compressed(filename, data=obj.values, index=obj.index.values, columns=obj.columns.values) def save_df_to_text(obj, filename): obj.to_csv(filename, sep='\t') def load_df_from_npz(filename): with np.load(filename, allow_pickle=True) as f: obj = pd.DataFrame(**f) return obj def check_dir_exists(path): """ Checks if directory already exists or not and creates it if it doesn't """ try: os.makedirs(path) except OSError as exception: if exception.errno != errno.EEXIST: raise def worker_filter(iterable, worker_index, total_workers): return (p for i,p in enumerate(iterable) if (i-worker_index)%total_workers==0) def fast_euclidean(mat): D = mat.dot(mat.T) squared_norms = np.diag(D).copy() D *= -2.0 D += squared_norms.reshape((-1,1)) D += squared_norms.reshape((1,-1)) D = np.sqrt(D) D[D < 0] = 0 return squareform(D, checks=False) def fast_ols_all_cols(X, Y): pinv = np.linalg.pinv(X) beta = np.dot(pinv, Y) return(beta) def fast_ols_all_cols_df(X,Y): beta = fast_ols_all_cols(X, Y) beta = pd.DataFrame(beta, index=X.columns, columns=Y.columns) return(beta) def var_sparse_matrix(X): mean = np.array(X.mean(axis=0)).reshape(-1) Xcopy = X.copy() Xcopy.data **= 2 var = np.array(Xcopy.mean(axis=0)).reshape(-1) - (mean**2) return(var) def get_highvar_genes_sparse(expression, expected_fano_threshold=None, minimal_mean=0.5, numgenes=None): # Find high variance genes within those cells gene_mean = np.array(expression.mean(axis=0)).astype(float).reshape(-1) E2 = expression.copy(); E2.data **= 2; gene2_mean = np.array(E2.mean(axis=0)).reshape(-1) gene_var = pd.Series(gene2_mean - (gene_mean**2)) del(E2) gene_mean = pd.Series(gene_mean) gene_fano = gene_var / gene_mean # Find parameters for expected fano line top_genes = gene_mean.sort_values(ascending=False)[:20].index A = (np.sqrt(gene_var)/gene_mean)[top_genes].min() w_mean_low, w_mean_high = gene_mean.quantile([0.10, 0.90]) w_fano_low, w_fano_high = gene_fano.quantile([0.10, 0.90]) winsor_box = ((gene_fano > w_fano_low) & (gene_fano < w_fano_high) & (gene_mean > w_mean_low) & (gene_mean < w_mean_high)) fano_median = gene_fano[winsor_box].median() B = np.sqrt(fano_median) gene_expected_fano = (A**2)*gene_mean + (B**2) fano_ratio = (gene_fano/gene_expected_fano) # Identify high var genes if numgenes is not None: highvargenes = fano_ratio.sort_values(ascending=False).index[:numgenes] high_var_genes_ind = fano_ratio.index.isin(highvargenes) T=None else: if not expected_fano_threshold: T = (1. + gene_counts_fano[winsor_box].std()) else: T = expected_fano_threshold high_var_genes_ind = (fano_ratio > T) & (gene_counts_mean > minimal_mean) gene_counts_stats = pd.DataFrame({ 'mean': gene_mean, 'var': gene_var, 'fano': gene_fano, 'expected_fano': gene_expected_fano, 'high_var': high_var_genes_ind, 'fano_ratio': fano_ratio }) gene_fano_parameters = { 'A': A, 'B': B, 'T':T, 'minimal_mean': minimal_mean, } return(gene_counts_stats, gene_fano_parameters) def get_highvar_genes(input_counts, expected_fano_threshold=None, minimal_mean=0.5, numgenes=None): # Find high variance genes within those cells gene_counts_mean = pd.Series(input_counts.mean(axis=0).astype(float)) gene_counts_var = pd.Series(input_counts.var(ddof=0, axis=0).astype(float)) gene_counts_fano = pd.Series(gene_counts_var/gene_counts_mean) # Find parameters for expected fano line top_genes = gene_counts_mean.sort_values(ascending=False)[:20].index A = (np.sqrt(gene_counts_var)/gene_counts_mean)[top_genes].min() w_mean_low, w_mean_high = gene_counts_mean.quantile([0.10, 0.90]) w_fano_low, w_fano_high = gene_counts_fano.quantile([0.10, 0.90]) winsor_box = ((gene_counts_fano > w_fano_low) & (gene_counts_fano < w_fano_high) & (gene_counts_mean > w_mean_low) & (gene_counts_mean < w_mean_high)) fano_median = gene_counts_fano[winsor_box].median() B = np.sqrt(fano_median) gene_expected_fano = (A**2)*gene_counts_mean + (B**2) fano_ratio = (gene_counts_fano/gene_expected_fano) # Identify high var genes if numgenes is not None: highvargenes = fano_ratio.sort_values(ascending=False).index[:numgenes] high_var_genes_ind = fano_ratio.index.isin(highvargenes) T=None else: if not expected_fano_threshold: T = (1. + gene_counts_fano[winsor_box].std()) else: T = expected_fano_threshold high_var_genes_ind = (fano_ratio > T) & (gene_counts_mean > minimal_mean) gene_counts_stats = pd.DataFrame({ 'mean': gene_counts_mean, 'var': gene_counts_var, 'fano': gene_counts_fano, 'expected_fano': gene_expected_fano, 'high_var': high_var_genes_ind, 'fano_ratio': fano_ratio }) gene_fano_parameters = { 'A': A, 'B': B, 'T':T, 'minimal_mean': minimal_mean, } return(gene_counts_stats, gene_fano_parameters) def compute_tpm(input_counts): """ Default TPM normalization """ tpm = input_counts.copy() sc.pp.normalize_per_cell(tpm, counts_per_cell_after=1e6) return(tpm) class cNMF(): def __init__(self, output_dir=".", name=None): """ Parameters ---------- output_dir : path, optional (default=".") Output directory for analysis files. name : string, optional (default=None) A name for this analysis. Will be prefixed to all output files. If set to None, will be automatically generated from date (and random string). """ self.output_dir = output_dir if name is None: now = datetime.datetime.now() rand_hash = uuid.uuid4().hex[:6] name = '%s_%s' % (now.strftime("%Y_%m_%d"), rand_hash) self.name = name self.paths = None def _initialize_dirs(self): if self.paths is None: # Check that output directory exists, create it if needed. check_dir_exists(self.output_dir) check_dir_exists(os.path.join(self.output_dir, self.name)) check_dir_exists(os.path.join(self.output_dir, self.name, 'cnmf_tmp')) self.paths = { 'normalized_counts' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.norm_counts.h5ad'), 'nmf_replicate_parameters' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.nmf_params.df.npz'), 'nmf_run_parameters' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.nmf_idvrun_params.yaml'), 'nmf_genes_list' : os.path.join(self.output_dir, self.name, self.name+'.overdispersed_genes.txt'), 'tpm' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.tpm.h5ad'), 'tpm_stats' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.tpm_stats.df.npz'), 'iter_spectra' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.iter_%d.df.npz'), 'iter_usages' : os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.usages.k_%d.iter_%d.df.npz'), 'merged_spectra': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.merged.df.npz'), 'local_density_cache': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.local_density_cache.k_%d.merged.df.npz'), 'consensus_spectra': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.spectra.k_%d.dt_%s.consensus.df.npz'), 'consensus_spectra__txt': os.path.join(self.output_dir, self.name, self.name+'.spectra.k_%d.dt_%s.consensus.txt'), 'consensus_usages': os.path.join(self.output_dir, self.name, 'cnmf_tmp',self.name+'.usages.k_%d.dt_%s.consensus.df.npz'), 'consensus_usages__txt': os.path.join(self.output_dir, self.name, self.name+'.usages.k_%d.dt_%s.consensus.txt'), 'consensus_stats': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.stats.k_%d.dt_%s.df.npz'), 'clustering_plot': os.path.join(self.output_dir, self.name, self.name+'.clustering.k_%d.dt_%s.png'), 'gene_spectra_score': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.gene_spectra_score.k_%d.dt_%s.df.npz'), 'gene_spectra_score__txt': os.path.join(self.output_dir, self.name, self.name+'.gene_spectra_score.k_%d.dt_%s.txt'), 'gene_spectra_tpm': os.path.join(self.output_dir, self.name, 'cnmf_tmp', self.name+'.gene_spectra_tpm.k_%d.dt_%s.df.npz'), 'gene_spectra_tpm__txt': os.path.join(self.output_dir, self.name, self.name+'.gene_spectra_tpm.k_%d.dt_%s.txt'), 'k_selection_plot' : os.path.join(self.output_dir, self.name, self.name+'.k_selection.png'), 'k_selection_stats' : os.path.join(self.output_dir, self.name, self.name+'.k_selection_stats.df.npz'), } def get_norm_counts(self, counts, tpm, high_variance_genes_filter = None, num_highvar_genes = None ): """ Parameters ---------- counts : anndata.AnnData Scanpy AnnData object (cells x genes) containing raw counts. Filtered such that no genes or cells with 0 counts tpm : anndata.AnnData Scanpy AnnData object (cells x genes) containing tpm normalized data matching counts high_variance_genes_filter : np.array, optional (default=None) A pre-specified list of genes considered to be high-variance. Only these genes will be used during factorization of the counts matrix. Must match the .var index of counts and tpm. If set to None, high-variance genes will be automatically computed, using the parameters below. num_highvar_genes : int, optional (default=None) Instead of providing an array of high-variance genes, identify this many most overdispersed genes for filtering Returns ------- normcounts : anndata.AnnData, shape (cells, num_highvar_genes) A counts matrix containing only the high variance genes and with columns (genes)normalized to unit variance """ if high_variance_genes_filter is None: ## Get list of high-var genes if one wasn't provided if sp.issparse(tpm.X): (gene_counts_stats, gene_fano_params) = get_highvar_genes_sparse(tpm.X, numgenes=num_highvar_genes) else: (gene_counts_stats, gene_fano_params) = get_highvar_genes(np.array(tpm.X), numgenes=num_highvar_genes) high_variance_genes_filter = list(tpm.var.index[gene_counts_stats.high_var.values]) ## Subset out high-variance genes norm_counts = counts[:, high_variance_genes_filter] ## Scale genes to unit variance if sp.issparse(tpm.X): sc.pp.scale(norm_counts, zero_center=False) if np.isnan(norm_counts.X.data).sum() > 0: print('Warning NaNs in normalized counts matrix') else: norm_counts.X /= norm_counts.X.std(axis=0, ddof=1) if np.isnan(norm_counts.X).sum().sum() > 0: print('Warning NaNs in normalized counts matrix') ## Save a \n-delimited list of the high-variance genes used for factorization open(self.paths['nmf_genes_list'], 'w').write('\n'.join(high_variance_genes_filter)) ## Check for any cells that have 0 counts of the overdispersed genes zerocells = norm_counts.X.sum(axis=1)==0 if zerocells.sum()>0: examples = norm_counts.obs.index[zerocells] print('Warning: %d cells have zero counts of overdispersed genes. E.g. %s' % (zerocells.sum(), examples[0])) print('Consensus step may not run when this is the case') return(norm_counts) def save_norm_counts(self, norm_counts): self._initialize_dirs() sc.write(self.paths['normalized_counts'], norm_counts) def get_nmf_iter_params(self, ks, n_iter = 100, random_state_seed = None, beta_loss = 'kullback-leibler'): """ Create a DataFrame with parameters for NMF iterations. Parameters ---------- ks : integer, or list-like. Number of topics (components) for factorization. Several values can be specified at the same time, which will be run independently. n_iter : integer, optional (defailt=100) Number of iterations for factorization. If several ``k`` are specified, this many iterations will be run for each value of ``k``. random_state_seed : int or None, optional (default=None) Seed for sklearn random state. """ if type(ks) is int: ks = [ks] # Remove any repeated k values, and order. k_list = sorted(set(list(ks))) n_runs = len(ks)* n_iter np.random.seed(seed=random_state_seed) nmf_seeds = np.random.randint(low=1, high=(2**32)-1, size=n_runs) replicate_params = [] for i, (k, r) in enumerate(itertools.product(k_list, range(n_iter))): replicate_params.append([k, r, nmf_seeds[i]]) replicate_params = pd.DataFrame(replicate_params, columns = ['n_components', 'iter', 'nmf_seed']) _nmf_kwargs = dict( alpha=0.0, l1_ratio=0.0, beta_loss=beta_loss, solver='mu', tol=1e-4, max_iter=1000, regularization=None, init='random' ) ## Coordinate descent is faster than multiplicative update but only works for frobenius if beta_loss == 'frobenius': _nmf_kwargs['solver'] = 'cd' return(replicate_params, _nmf_kwargs) def save_nmf_iter_params(self, replicate_params, run_params): self._initialize_dirs() save_df_to_npz(replicate_params, self.paths['nmf_replicate_parameters']) with open(self.paths['nmf_run_parameters'], 'w') as F: yaml.dump(run_params, F) def _nmf(self, X, nmf_kwargs): """ Parameters ---------- X : pandas.DataFrame, Normalized counts dataFrame to be factorized. nmf_kwargs : dict, Arguments to be passed to ``non_negative_factorization`` """ (usages, spectra, niter) = non_negative_factorization(X, **nmf_kwargs) return(spectra, usages) def run_nmf(self, worker_i=1, total_workers=1, ): """ Iteratively run NMF with prespecified parameters. Use the `worker_i` and `total_workers` parameters for parallelization. Generic kwargs for NMF are loaded from self.paths['nmf_run_parameters'], defaults below:: ``non_negative_factorization`` default arguments: alpha=0.0 l1_ratio=0.0 beta_loss='kullback-leibler' solver='mu' tol=1e-4, max_iter=200 regularization=None init='random' random_state, n_components are both set by the prespecified self.paths['nmf_replicate_parameters']. Parameters ---------- norm_counts : pandas.DataFrame, Normalized counts dataFrame to be factorized. (Output of ``normalize_counts``) run_params : pandas.DataFrame, Parameters for NMF iterations. (Output of ``prepare_nmf_iter_params``) """ self._initialize_dirs() run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) norm_counts = sc.read(self.paths['normalized_counts']) _nmf_kwargs = yaml.load(open(self.paths['nmf_run_parameters']), Loader=yaml.FullLoader) jobs_for_this_worker = worker_filter(range(len(run_params)), worker_i, total_workers) for idx in jobs_for_this_worker: p = run_params.iloc[idx, :] print('[Worker %d]. Starting task %d.' % (worker_i, idx)) _nmf_kwargs['random_state'] = p['nmf_seed'] _nmf_kwargs['n_components'] = p['n_components'] (spectra, usages) = self._nmf(norm_counts.X, _nmf_kwargs) spectra = pd.DataFrame(spectra, index=np.arange(1, _nmf_kwargs['n_components']+1), columns=norm_counts.var.index) save_df_to_npz(spectra, self.paths['iter_spectra'] % (p['n_components'], p['iter'])) def combine_nmf(self, k, remove_individual_iterations=False): run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) print('Combining factorizations for k=%d.'%k) self._initialize_dirs() combined_spectra = None n_iter = sum(run_params.n_components==k) run_params_subset = run_params[run_params.n_components==k].sort_values('iter') spectra_labels = [] for i,p in run_params_subset.iterrows(): spectra = load_df_from_npz(self.paths['iter_spectra'] % (p['n_components'], p['iter'])) if combined_spectra is None: combined_spectra = np.zeros((n_iter, k, spectra.shape[1])) combined_spectra[p['iter'], :, :] = spectra.values for t in range(k): spectra_labels.append('iter%d_topic%d'%(p['iter'], t+1)) combined_spectra = combined_spectra.reshape(-1, combined_spectra.shape[-1]) combined_spectra = pd.DataFrame(combined_spectra, columns=spectra.columns, index=spectra_labels) save_df_to_npz(combined_spectra, self.paths['merged_spectra']%k) return combined_spectra def consensus(self, k, density_threshold_str='0.5', local_neighborhood_size = 0.30,show_clustering = False, skip_density_and_return_after_stats = False, close_clustergram_fig=True): merged_spectra = load_df_from_npz(self.paths['merged_spectra']%k) norm_counts = sc.read(self.paths['normalized_counts']) if skip_density_and_return_after_stats: density_threshold_str = '2' density_threshold_repl = density_threshold_str.replace('.', '_') density_threshold = float(density_threshold_str) n_neighbors = int(local_neighborhood_size * merged_spectra.shape[0]/k) # Rescale topics such to length of 1. l2_spectra = (merged_spectra.T/np.sqrt((merged_spectra**2).sum(axis=1))).T if not skip_density_and_return_after_stats: # Compute the local density matrix (if not previously cached) topics_dist = None if os.path.isfile(self.paths['local_density_cache'] % k): local_density = load_df_from_npz(self.paths['local_density_cache'] % k) else: # first find the full distance matrix topics_dist = squareform(fast_euclidean(l2_spectra.values)) # partition based on the first n neighbors partitioning_order = np.argpartition(topics_dist, n_neighbors+1)[:, :n_neighbors+1] # find the mean over those n_neighbors (excluding self, which has a distance of 0) distance_to_nearest_neighbors = topics_dist[np.arange(topics_dist.shape[0])[:, None], partitioning_order] local_density = pd.DataFrame(distance_to_nearest_neighbors.sum(1)/(n_neighbors), columns=['local_density'], index=l2_spectra.index) save_df_to_npz(local_density, self.paths['local_density_cache'] % k) del(partitioning_order) del(distance_to_nearest_neighbors) density_filter = local_density.iloc[:, 0] < density_threshold l2_spectra = l2_spectra.loc[density_filter, :] kmeans_model = KMeans(n_clusters=k, n_init=10, random_state=1) kmeans_model.fit(l2_spectra) kmeans_cluster_labels = pd.Series(kmeans_model.labels_+1, index=l2_spectra.index) # Find median usage for each gene across cluster median_spectra = l2_spectra.groupby(kmeans_cluster_labels).median() # Normalize median spectra to probability distributions. median_spectra = (median_spectra.T/median_spectra.sum(1)).T # Compute the silhouette score stability = silhouette_score(l2_spectra.values, kmeans_cluster_labels, metric='euclidean') # Obtain the reconstructed count matrix by re-fitting the usage matrix and computing the dot product: usage.dot(spectra) refit_nmf_kwargs = yaml.load(open(self.paths['nmf_run_parameters']), Loader=yaml.FullLoader) refit_nmf_kwargs.update(dict( n_components = k, H = median_spectra.values, update_H = False )) _, rf_usages = self._nmf(norm_counts.X, nmf_kwargs=refit_nmf_kwargs) rf_usages = pd.DataFrame(rf_usages, index=norm_counts.obs.index, columns=median_spectra.index) rf_pred_norm_counts = rf_usages.dot(median_spectra) # Compute prediction error as a frobenius norm if sp.issparse(norm_counts.X): prediction_error = ((norm_counts.X.todense() - rf_pred_norm_counts)**2).sum().sum() else: prediction_error = ((norm_counts.X - rf_pred_norm_counts)**2).sum().sum() consensus_stats = pd.DataFrame([k, density_threshold, stability, prediction_error], index = ['k', 'local_density_threshold', 'stability', 'prediction_error'], columns = ['stats']) if skip_density_and_return_after_stats: return consensus_stats save_df_to_npz(median_spectra, self.paths['consensus_spectra']%(k, density_threshold_repl)) save_df_to_npz(rf_usages, self.paths['consensus_usages']%(k, density_threshold_repl)) save_df_to_npz(consensus_stats, self.paths['consensus_stats']%(k, density_threshold_repl)) save_df_to_text(median_spectra, self.paths['consensus_spectra__txt']%(k, density_threshold_repl)) save_df_to_text(rf_usages, self.paths['consensus_usages__txt']%(k, density_threshold_repl)) # Compute gene-scores for each GEP by regressing usage on Z-scores of TPM tpm = sc.read(self.paths['tpm']) tpm_stats = load_df_from_npz(self.paths['tpm_stats']) if sp.issparse(tpm.X): norm_tpm = (np.array(tpm.X.todense()) - tpm_stats['__mean'].values) / tpm_stats['__std'].values else: norm_tpm = (tpm.X - tpm_stats['__mean'].values) / tpm_stats['__std'].values if norm_tpm.dtype != np.float64: norm_tpm = norm_tpm.astype(np.float64) usage_coef = fast_ols_all_cols(rf_usages.values, norm_tpm) usage_coef = pd.DataFrame(usage_coef, index=rf_usages.columns, columns=tpm.var.index) save_df_to_npz(usage_coef, self.paths['gene_spectra_score']%(k, density_threshold_repl)) save_df_to_text(usage_coef, self.paths['gene_spectra_score__txt']%(k, density_threshold_repl)) # Convert spectra to TPM units, and obtain results for all genes by running last step of NMF # with usages fixed and TPM as the input matrix norm_usages = rf_usages.div(rf_usages.sum(axis=1), axis=0) refit_nmf_kwargs.update(dict( H = norm_usages.T.values, )) # Needed otherwise _nmf will crash because with inconsistent dtypes if tpm.X.dtype != np.float64: tpm.X = tpm.X.astype(np.float64) _, spectra_tpm = self._nmf(tpm.X.T, nmf_kwargs=refit_nmf_kwargs) spectra_tpm = pd.DataFrame(spectra_tpm.T, index=rf_usages.columns, columns=tpm.var.index) save_df_to_npz(spectra_tpm, self.paths['gene_spectra_tpm']%(k, density_threshold_repl)) save_df_to_text(spectra_tpm, self.paths['gene_spectra_tpm__txt']%(k, density_threshold_repl)) if show_clustering: if topics_dist is None: topics_dist = squareform(fast_euclidean(l2_spectra.values)) # (l2_spectra was already filtered using the density filter) else: # (but the previously computed topics_dist was not!) topics_dist = topics_dist[density_filter.values, :][:, density_filter.values] spectra_order = [] for cl in sorted(set(kmeans_cluster_labels)): cl_filter = kmeans_cluster_labels==cl if cl_filter.sum() > 1: cl_dist = squareform(topics_dist[cl_filter, :][:, cl_filter]) cl_dist[cl_dist < 0] = 0 #Rarely get floating point arithmetic issues cl_link = linkage(cl_dist, 'average') cl_leaves_order = leaves_list(cl_link) spectra_order += list(np.where(cl_filter)[0][cl_leaves_order]) else: ## Corner case where a component only has one element spectra_order += list(np.where(cl_filter)[0]) from matplotlib import gridspec import matplotlib.pyplot as plt width_ratios = [0.5, 9, 0.5, 4, 1] height_ratios = [0.5, 9] fig = plt.figure(figsize=(sum(width_ratios), sum(height_ratios))) gs = gridspec.GridSpec(len(height_ratios), len(width_ratios), fig, 0.01, 0.01, 0.98, 0.98, height_ratios=height_ratios, width_ratios=width_ratios, wspace=0, hspace=0) dist_ax = fig.add_subplot(gs[1,1], xscale='linear', yscale='linear', xticks=[], yticks=[],xlabel='', ylabel='', frameon=True) D = topics_dist[spectra_order, :][:, spectra_order] dist_im = dist_ax.imshow(D, interpolation='none', cmap='viridis', aspect='auto', rasterized=True) left_ax = fig.add_subplot(gs[1,0], xscale='linear', yscale='linear', xticks=[], yticks=[], xlabel='', ylabel='', frameon=True) left_ax.imshow(kmeans_cluster_labels.values[spectra_order].reshape(-1, 1), interpolation='none', cmap='Spectral', aspect='auto', rasterized=True) top_ax = fig.add_subplot(gs[0,1], xscale='linear', yscale='linear', xticks=[], yticks=[], xlabel='', ylabel='', frameon=True) top_ax.imshow(kmeans_cluster_labels.values[spectra_order].reshape(1, -1), interpolation='none', cmap='Spectral', aspect='auto', rasterized=True) hist_gs = gridspec.GridSpecFromSubplotSpec(3, 1, subplot_spec=gs[1, 3], wspace=0, hspace=0) hist_ax = fig.add_subplot(hist_gs[0,0], xscale='linear', yscale='linear', xlabel='', ylabel='', frameon=True, title='Local density histogram') hist_ax.hist(local_density.values, bins=np.linspace(0, 1, 50)) hist_ax.yaxis.tick_right() xlim = hist_ax.get_xlim() ylim = hist_ax.get_ylim() if density_threshold < xlim[1]: hist_ax.axvline(density_threshold, linestyle='--', color='k') hist_ax.text(density_threshold + 0.02, ylim[1] * 0.95, 'filtering\nthreshold\n\n', va='top') hist_ax.set_xlim(xlim) hist_ax.set_xlabel('Mean distance to k nearest neighbors\n\n%d/%d (%.0f%%) spectra above threshold\nwere removed prior to clustering'%(sum(~density_filter), len(density_filter), 100*(~density_filter).mean())) ## Add colorbar cbar_gs = gridspec.GridSpecFromSubplotSpec(8, 1, subplot_spec=hist_gs[1, 0], wspace=0, hspace=0) cbar_ax = fig.add_subplot(cbar_gs[4,0], xscale='linear', yscale='linear', xlabel='', ylabel='', frameon=True, title='Euclidean Distance') vmin = D.min().min() vmax = D.max().max() fig.colorbar(dist_im, cax=cbar_ax, ticks=np.linspace(vmin, vmax, 3), orientation='horizontal') #hist_ax.hist(local_density.values, bins=np.linspace(0, 1, 50)) #hist_ax.yaxis.tick_right() fig.savefig(self.paths['clustering_plot']%(k, density_threshold_repl), dpi=250) if close_clustergram_fig: plt.close(fig) def k_selection_plot(self, close_fig=True): ''' Borrowed from Alexandrov Et Al. 2013 Deciphering Mutational Signatures publication in Cell Reports ''' run_params = load_df_from_npz(self.paths['nmf_replicate_parameters']) stats = [] for k in sorted(set(run_params.n_components)): stats.append(self.consensus(k, skip_density_and_return_after_stats=True).stats) stats = pd.DataFrame(stats) stats.reset_index(drop = True, inplace = True) save_df_to_npz(stats, self.paths['k_selection_stats']) fig = plt.figure(figsize=(6, 4)) ax1 = fig.add_subplot(111) ax2 = ax1.twinx() ax1.plot(stats.k, stats.stability, 'o-', color='b') ax1.set_ylabel('Stability', color='b', fontsize=15) for tl in ax1.get_yticklabels(): tl.set_color('b') #ax1.set_xlabel('K', fontsize=15) ax2.plot(stats.k, stats.prediction_error, 'o-', color='r') ax2.set_ylabel('Error', color='r', fontsize=15) for tl in ax2.get_yticklabels(): tl.set_color('r') ax1.set_xlabel('Number of Components', fontsize=15) ax1.grid('on') plt.tight_layout() fig.savefig(self.paths['k_selection_plot'], dpi=250) if close_fig: plt.close(fig) if __name__=="__main__": """ Example commands for now: output_dir="/Users/averes/Projects/Melton/Notebooks/2018/07-2018/cnmf_test/" python cnmf.py prepare --output-dir $output_dir \ --name test --counts /Users/averes/Projects/Melton/Notebooks/2018/07-2018/cnmf_test/test_data.df.npz \ -k 6 7 8 9 --n-iter 5 python cnmf.py factorize --name test --output-dir $output_dir THis can be parallelized as such: python cnmf.py factorize --name test --output-dir $output_dir --total-workers 2 --worker-index WORKER_INDEX (where worker_index starts with 0) python cnmf.py combine --name test --output-dir $output_dir python cnmf.py consensus --name test --output-dir $output_dir """ import sys, argparse parser = argparse.ArgumentParser() parser.add_argument('command', type=str, choices=['prepare', 'factorize', 'combine', 'consensus', 'k_selection_plot']) parser.add_argument('--name', type=str, help='[all] Name for analysis. All output will be placed in [output-dir]/[name]/...', nargs='?', default='cNMF') parser.add_argument('--output-dir', type=str, help='[all] Output directory. All output will be placed in [output-dir]/[name]/...', nargs='?', default='.') parser.add_argument('-c', '--counts', type=str, help='[prepare] Input (cell x gene) counts matrix as df.npz or tab delimited text file') parser.add_argument('-k', '--components', type=int, help='[prepare] Numper of components (k) for matrix factorization. Several can be specified with "-k 8 9 10"', nargs='+') parser.add_argument('-n', '--n-iter', type=int, help='[prepare] Numper of factorization replicates', default=100) parser.add_argument('--total-workers', type=int, help='[all] Total number of workers to distribute jobs to', default=1) parser.add_argument('--seed', type=int, help='[prepare] Seed for pseudorandom number generation', default=None) parser.add_argument('--genes-file', type=str, help='[prepare] File containing a list of genes to include, one gene per line. Must match column labels of counts matrix.', default=None) parser.add_argument('--numgenes', type=int, help='[prepare] Number of high variance genes to use for matrix factorization.', default=2000) parser.add_argument('--tpm', type=str, help='[prepare] Pre-computed (cell x gene) TPM values as df.npz or tab separated txt file. If not provided TPM will be calculated automatically', default=None) parser.add_argument('--beta-loss', type=str, choices=['frobenius', 'kullback-leibler', 'itakura-saito'], help='[prepare] Loss function for NMF.', default='frobenius') parser.add_argument('--densify', dest='densify', help='[prepare] Treat the input data as non-sparse', action='store_true', default=False) parser.add_argument('--worker-index', type=int, help='[factorize] Index of current worker (the first worker should have index 0)', default=0) parser.add_argument('--local-density-threshold', type=str, help='[consensus] Threshold for the local density filtering. This string must convert to a float >0 and <=2', default='0.5') parser.add_argument('--local-neighborhood-size', type=float, help='[consensus] Fraction of the number of replicates to use as nearest neighbors for local density filtering', default=0.30) parser.add_argument('--show-clustering', dest='show_clustering', help='[consensus] Produce a clustergram figure summarizing the spectra clustering', action='store_true') args = parser.parse_args() cnmf_obj = cNMF(output_dir=args.output_dir, name=args.name) cnmf_obj._initialize_dirs() if args.command == 'prepare': if args.counts.endswith('.h5ad'): input_counts = sc.read(args.counts) else: ## Load txt or compressed dataframe and convert to scanpy object if args.counts.endswith('.npz'): input_counts = load_df_from_npz(args.counts) else: input_counts = pd.read_csv(args.counts, sep='\t', index_col=0) if args.densify: input_counts = sc.AnnData(X=input_counts.values, obs=pd.DataFrame(index=input_counts.index), var=pd.DataFrame(index=input_counts.columns)) else: input_counts = sc.AnnData(X=sp.csr_matrix(input_counts.values), obs=pd.DataFrame(index=input_counts.index), var=pd.DataFrame(index=input_counts.columns)) if sp.issparse(input_counts.X) & args.densify: input_counts.X = np.array(input_counts.X.todense()) if args.tpm is None: tpm = compute_tpm(input_counts) sc.write(cnmf_obj.paths['tpm'], tpm) elif args.tpm.endswith('.h5ad'): subprocess.call('cp %s %s' % (args.tpm, cnmf_obj.paths['tpm']), shell=True) tpm = sc.read(cnmf_obj.paths['tpm']) else: if args.tpm.endswith('.npz'): tpm = load_df_from_npz(args.tpm) else: tpm = pd.read_csv(args.tpm, sep='\t', index_col=0) if args.densify: tpm = sc.AnnData(X=tpm.values, obs=pd.DataFrame(index=tpm.index), var=pd.DataFrame(index=tpm.columns)) else: tpm = sc.AnnData(X=sp.csr_matrix(tpm.values), obs=pd.DataFrame(index=tpm.index), var=pd.DataFrame(index=tpm.columns)) sc.write(cnmf_obj.paths['tpm'], tpm) if sp.issparse(tpm.X): gene_tpm_mean = np.array(tpm.X.mean(axis=0)).reshape(-1) gene_tpm_stddev = var_sparse_matrix(tpm.X)**.5 else: gene_tpm_mean = np.array(tpm.X.mean(axis=0)).reshape(-1) gene_tpm_stddev = np.array(tpm.X.std(axis=0, ddof=0)).reshape(-1) input_tpm_stats = pd.DataFrame([gene_tpm_mean, gene_tpm_stddev], index = ['__mean', '__std']).T save_df_to_npz(input_tpm_stats, cnmf_obj.paths['tpm_stats']) if args.genes_file is not None: highvargenes = open(args.genes_file).read().rstrip().split('\n') else: highvargenes = None norm_counts = cnmf_obj.get_norm_counts(input_counts, tpm, num_highvar_genes=args.numgenes, high_variance_genes_filter=highvargenes) if norm_counts.X.dtype != np.float64: norm_counts.X = norm_counts.X.astype(np.float64) cnmf_obj.save_norm_counts(norm_counts) (replicate_params, run_params) = cnmf_obj.get_nmf_iter_params(ks=args.components, n_iter=args.n_iter, random_state_seed=args.seed, beta_loss=args.beta_loss) cnmf_obj.save_nmf_iter_params(replicate_params, run_params) elif args.command == 'factorize': cnmf_obj.run_nmf(worker_i=args.worker_index, total_workers=args.total_workers) elif args.command == 'combine': run_params = load_df_from_npz(cnmf_obj.paths['nmf_replicate_parameters']) if type(args.components) is int: ks = [args.components] elif args.components is None: ks = sorted(set(run_params.n_components)) else: ks = args.components for k in ks: cnmf_obj.combine_nmf(k) elif args.command == 'consensus': run_params = load_df_from_npz(cnmf_obj.paths['nmf_replicate_parameters']) if type(args.components) is int: ks = [args.components] elif args.components is None: ks = sorted(set(run_params.n_components)) else: ks = args.components for k in ks: merged_spectra = load_df_from_npz(cnmf_obj.paths['merged_spectra']%k) cnmf_obj.consensus(k, args.local_density_threshold, args.local_neighborhood_size, args.show_clustering) elif args.command == 'k_selection_plot': cnmf_obj.k_selection_plot() |
1
of
cNMF/cnmf.py
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 | suppressPackageStartupMessages(library(dplyr)) suppressPackageStartupMessages(library(ggplot2)) suppressPackageStartupMessages(library(ggpubr)) suppressPackageStartupMessages(library(data.table)) suppressPackageStartupMessages(library(tidyr)) suppressPackageStartupMessages(library(readxl)) suppressPackageStartupMessages(library(ggrepel)) suppressPackageStartupMessages(library(optparse)) suppressPackageStartupMessages(library(gplots)) suppressPackageStartupMessages(library(cowplot)) ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/2kG.library/acrossK/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK/", help="Output directory"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--p.adj.threshold", type="numeric", default=0.1, help="Threshold for fdr and adjusted p-value"), make_option("--aggregated.data", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis//all_genes/Perturb_2kG_dup4/acrossK/aggregated.outputs.findK.perturb-seq.RData") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev for 2n1.99x singlets ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/figures/all_genes/Perturb_2kG_dup4/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/outputs/all_genes/Perturb_2kG_dup4/acrossK/" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/acrossK/aggregated.outputs.findK.perturb-seq.RData" ## opt$sampleName <- "Perturb_2kG_dup4" ## ## for all genes (in sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/2kG.library/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK/" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK//aggregated.outputs.findK.perturb-seq.RData" ## ## ## for testing cNMF_ pipeline ## ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/2kG.library/acrossK/" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_2min/acrossK/aggregated.outputs.findK.RData" ## ## for testing cNMF_pipeline with FT010_fresh_4min ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/FT010_fresh_4min/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_4min/acrossK/" ## opt$sampleName <- "FT010_fresh_4min" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_4min/acrossK/aggregated.outputs.findK.RData" ## ## for testing findK_plots for scRNAseq_2kG_11AMDox_1 ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/figures/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/" ## opt$sampleName <- "scRNAseq_2kG_11AMDox_1" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/aggregated.outputs.findK.RData" ## ## for testing findK_plots for control only cells ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/2kG.library.ctrl.only/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/acrossK/" ## opt$sampleName <- "2kG.library.ctrl.only" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/acrossK/aggregated.outputs.findK.RData" ## ## for testing findK_plots for perturb-seq only data for control only cells ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/acrossK/aggregated.outputs.findK.perturb-seq.RData" ## ## K562 gwps 2k overdispersed genes ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/acrossK/" ## opt$sampleName <- "WeissmanK562gwps" ## opt$p.adj.threshold <- 0.1 ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/acrossK/aggregated.outputs.findK.perturb-seq.RData" ## Directories and Constants SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DATADIR=opt$datadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir FIGDIR=opt$figdir check.dir <- c(OUTDIR, FIGDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) fdr.thr <- opt$p.adj.threshold ## load("/Volumes/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/analysis/no_IL1B/aggregated.outputs.findK.RData") ## load("/Volumes/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/differential_expression/210526_SeuratDE/outputs/no_IL1B/de.markers.RData") mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) mytheme <- theme_classic() + theme(axis.text = element_text(size = 5), axis.title = element_text(size = 6), plot.title = element_text(hjust = 0.5, face = "bold", size=7), axis.line = element_line(color = "black", size = 0.25), axis.ticks = element_line(color = "black", size = 0.25), legend.title = element_text(size=6), legend.text = element_text(size=6)) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## Load data load(opt$aggregated.data) ## MAST.df, all.test.df, all.enhancer.ttest.df, all.promoter.ttest.df ## ref.table <- read_xlsx(opt$reference.table, sheet="2000_gene_library_annotated") ## process mast data MAST.original <- MAST.df <- MAST.df %>% mutate(ProgramID = paste0("K", K, "_", gsub("topic_", "", primerid))) %>% mutate(perturbation = gsub("MESDC1", "TLNRD1", perturbation)) ## for antiparallel genes, GeneA-and-GeneB, keep {GeneA, GeneB}_multiTarget and remove {GeneA, GeneB} perturbations if(grepl("2kG.library", SAMPLE)) { antiparallel.perturbation <- MAST.df %>% subset(grepl("multiTarget", perturbation)) %>% pull(perturbation) %>% unique %>% gsub("_multiTarget", "", .) MAST.df <- MAST.df %>% subset(!(perturbation %in% antiparallel.perturbation)) } MAST.df <- MAST.df %>% subset(zlm.model.name == "batch.correction") %>% group_by(zlm.model.name, K) %>% mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method="fdr")) %>% group_by(ProgramID) %>% arrange(desc(coef)) %>% mutate(coef_rank = 1:n()) %>% as.data.frame sig.MAST.df <- MAST.df %>% subset(fdr.across.ptb < fdr.thr) ################################################## ## plots fig.file.name <- paste0(FIGDIR, "/percent.batch.topics") batch.percent.df$batch.thr <- as.character(batch.percent.df$batch.thr) ## percent of topics correlated with batch over K pdf(paste0(fig.file.name, ".pdf"), width=3, height=2) p <- batch.percent.df %>% ggplot(aes(x = K, y = percent.correlated, color = batch.thr)) + geom_line(size=0.5) + geom_point(size=0.5) + mytheme + ggtitle(paste0(SAMPLE, " Percent of Programs Correlated with Batch")) + ## scale_x_continuous(breaks = batch.percent.df$K %>% unique) + scale_y_continuous(name = "% Programs Correlated with Batch", labels = scales::percent) + scale_color_discrete(name = "Pearson correlation") print(p) ggsave(paste0(fig.file.name, ".eps")) dev.off() ## MAST DE topics results ## metrics: ## # programs ## # unique programs ## fraction of significant programs MAST.program.summary.df <- sig.MAST.df %>% select(K, ProgramID) %>% unique %>% group_by(K) %>% summarize(nPrograms = n()) %>% mutate(fractionPrograms = nPrograms / K) %>% as.data.frame MAST.ptb.summary.df <- sig.MAST.df %>% select(K, perturbation) %>% unique %>% group_by(K) %>% summarize(nPerturbations = n()) %>% as.data.frame MAST.ptb.program.pair.summary.df <- sig.MAST.df %>% select(K, ProgramID, perturbation) %>% unique %>% group_by(K) %>% summarize(nPerturbationProgramPairs = n()) %>% mutate(averagePerturbationProgramPairs = nPerturbationProgramPairs / K) %>% as.data.frame MAST.summary.df <- merge(MAST.program.summary.df, MAST.ptb.summary.df, by="K", all=T) %>% merge(MAST.ptb.program.pair.summary.df, by="K", all=T) ## function to produce find K plot based on MAST result plotMAST <- function(toplot, MAST.metric, MAST.metric.label) { p <- toplot %>% ggplot(aes(x=K, y=get(MAST.metric))) + geom_point(size=0.5) + geom_line(size=0.5) + mytheme + xlab("K") + ylab(paste0(MAST.metric.label)) print(p) return(p) } toplot <- MAST.summary.df MAST.metrics <- MAST.summary.df %>% select(-K) %>% colnames MAST.metric.labels <- c("# Significant\nPrograms", "Fraction of\nSignificant\nPrograms", "# Regulators", "# Regulator x\n Program Pairs", "Average #\nRegulator x\nProgram Pairs") plotFilename <- paste0(FIGDIR, "MAST") plot.list <- list() pdf(paste0(plotFilename, ".pdf"), width=3, height=3) for(i in 1:length(MAST.metrics)) { MAST.metric = MAST.metrics[i] MAST.metric.label = MAST.metric.labels[i] plot.list[[MAST.metric]] <- plotMAST(toplot, MAST.metric, MAST.metric.label) eval(parse(text = paste0("p.", MAST.metric, " <- p"))) } dev.off() plotFilename <- paste0(FIGDIR, "All_MAST") pdf(paste0(plotFilename, ".pdf"), width=2, height=4) p <- plot_grid(plotlist = plot.list, nrow = length(plot.list), align="v", axis="lr") p <- annotate_figure(p, top = text_grob("Statistical test by MAST", size=8)) print(p) ggsave(paste0(plotFilename, ".eps")) dev.off() ## Wilcoxon Test results ## ## Update the files for 2kG library ## ## Intersecting EdgeR and Topic Model Results ## threshold <- opt$p.adj.threshold ## # % of perturbations with a large number of DE genes in the per-gene analysis that ALSO have significantly DE topics (p-value 0.005, logFC 1.2) ## strict.DE.all <- read_excel(path="/Users/helenkang/Documents/EngreitzLab/Pertube-seq_Analysis/200_gene_lib_EdgeR_DE.xlsx", sheet="EdgeR_DE_p.001_lfc1.2") ## lenient.DE.all <- read_excel(path="/Users/helenkang/Documents/EngreitzLab/Pertube-seq_Analysis/200_gene_lib_EdgeR_DE.xlsx", sheet="EdgeR_DE_p.01_lfc1.15") ## strict.DE <- strict.DE.all %>% select(Target...13, `NoI_#DE_genes`, `PlusI_#DE_genes`) %>% `colnames<-`(c("Gene", "NoI.num.genes", "PlusI.num.genes")) %>% mutate(cutoff.type = "strict", cutoff.detail = "p < 0.001, logFC > 1.2") ## lenient.DE <- lenient.DE.all %>% select(Target...13, NoI_DE...14, PlusI_DE...15) %>% `colnames<-`(c("Gene", "NoI.num.genes", "PlusI.num.genes")) %>% mutate(cutoff.type = "lenient", cutoff.detail = "p < 0.01, logFC > 1.15") ## edgeR.DE <- rbind(strict.DE, lenient.DE) ## perturbation.test.stat.df <- all.test.df %>% subset(adjusted.p.value < threshold) %>% select(K, Gene, test.type, Topic) %>% group_by(K, Gene, test.type) %>% summarize(topic.count = n()) ## edgeR.test.df <- merge(edgeR.DE, perturbation.test.stat.df, by="Gene") ## MSigDB Pathway Enrichment ## Number of GO enrichment per model ## Notes 210518 ## also output pdf ## plot top enriched MSigDB pathways for each topic and K ## also consider eps file ## make sure that p.adjust is caluclated over all topics for a model ## plot number of GO enrichment on raw score ranking per K ## par(mar = c(4, 4, .1, .1)) threshold = opt$p.adj.threshold |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 | suppressPackageStartupMessages(library(dplyr)) suppressPackageStartupMessages(library(ggplot2)) suppressPackageStartupMessages(library(ggpubr)) suppressPackageStartupMessages(library(data.table)) suppressPackageStartupMessages(library(tidyr)) suppressPackageStartupMessages(library(readxl)) suppressPackageStartupMessages(library(ggrepel)) suppressPackageStartupMessages(library(optparse)) suppressPackageStartupMessages(library(gplots)) suppressPackageStartupMessages(library(cowplot)) suppressPackageStartupMessages(library(ggpubr)) ## annotate arranged figure ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/2kG.library/acrossK/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK/", help="Output directory"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--p.adj.threshold", type="numeric", default=0.1, help="Threshold for fdr and adjusted p-value"), make_option("--aggregated.data", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK/aggregated.outputs.findK.RData") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## for overdispersedGenes 220617 ## opt$figdi <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/figures/top2000VariableGenes/2kG.library_overdispersedGenes/acrossK/" ## ## for all genes 210707 folder ## ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/2kG.library/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/acrossK/" ## opt$sampleName <- "2kG.library" ## ## for all genes (in sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/figures/all_genes/Perturb_2kG_dup4/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/acrossK/" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/acrossK//aggregated.outputs.findK.RData" ## opt$sampleName <- "Perturb_2kG_dup4" ## ## ## for testing cNMF_ pipeline ## ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/2kG.library/acrossK/" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_2min/acrossK/aggregated.outputs.findK.RData" ## ## for testing cNMF_pipeline with FT010_fresh_4min ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/FT010_fresh_4min/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_4min/acrossK/" ## opt$sampleName <- "FT010_fresh_4min" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/all_genes/FT010_fresh_4min/acrossK/aggregated.outputs.findK.RData" ## ## for testing findK_plots for scRNAseq_2kG_11AMDox_1 ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/figures/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/" ## opt$sampleName <- "scRNAseq_2kG_11AMDox_1" ## opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/acrossK/aggregated.outputs.findK.RData" # ## for testing findK_plots for control only cells # opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/2kG.library.ctrl.only/acrossK/" # opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/acrossK/" # opt$sampleName <- "2kG.library.ctrl.only" # opt$aggregated.data <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/acrossK/aggregated.outputs.findK.RData" ## ## K562 gwps 2k overdispersed genes ## opt$figdir <- "" ## Directories and Constants SAMPLE=strsplit(opt$sampleName,",") %>% unlist() threshold <- opt$p.adj.threshold DATADIR=opt$datadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir FIGDIR=opt$figdir check.dir <- c(OUTDIR, FIGDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## load("/Volumes/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2105_findK/analysis/no_IL1B/aggregated.outputs.findK.RData") ## load("/Volumes/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/differential_expression/210526_SeuratDE/outputs/no_IL1B/de.markers.RData") mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) mytheme <- theme_classic() + theme(axis.text = element_text(size = 5), axis.title = element_text(size = 6), plot.title = element_text(hjust = 0.5, face = "bold", size=7), axis.line = element_line(color = "black", size = 0.25), axis.ticks = element_line(color = "black", size = 0.25)) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) palette.log2FC = colorRampPalette(c("blue", "white", "red"))(n = 100) palette.pos = colorRampPalette(c("white", "red"))(n = 100) adjusted.p.thr <- 0.05 fdr.thr <- 0.05 ## Load data load(opt$aggregated.data) ## all.fdr.df, all.test.df, enhancer.fisher.df, count.by.GWAS, fgsea.results.df, promoter.fisher.df ## ref.table <- read_xlsx(opt$reference.table, sheet="2000_gene_library_annotated") ## ## Update the files for 2kG library ## ## Intersecting EdgeR and Topic Model Results ## threshold <- opt$p.adj.threshold ## # % of perturbations with a large number of DE genes in the per-gene analysis that ALSO have significantly DE topics (p-value 0.005, logFC 1.2) ## strict.DE.all <- read_excel(path="/Users/helenkang/Documents/EngreitzLab/Pertube-seq_Analysis/200_gene_lib_EdgeR_DE.xlsx", sheet="EdgeR_DE_p.001_lfc1.2") ## lenient.DE.all <- read_excel(path="/Users/helenkang/Documents/EngreitzLab/Pertube-seq_Analysis/200_gene_lib_EdgeR_DE.xlsx", sheet="EdgeR_DE_p.01_lfc1.15") ## strict.DE <- strict.DE.all %>% select(Target...13, `NoI_#DE_genes`, `PlusI_#DE_genes`) %>% `colnames<-`(c("Gene", "NoI.num.genes", "PlusI.num.genes")) %>% mutate(cutoff.type = "strict", cutoff.detail = "p < 0.001, logFC > 1.2") ## lenient.DE <- lenient.DE.all %>% select(Target...13, NoI_DE...14, PlusI_DE...15) %>% `colnames<-`(c("Gene", "NoI.num.genes", "PlusI.num.genes")) %>% mutate(cutoff.type = "lenient", cutoff.detail = "p < 0.01, logFC > 1.15") ## edgeR.DE <- rbind(strict.DE, lenient.DE) ## perturbation.test.stat.df <- all.test.df %>% subset(adjusted.p.value < threshold) %>% select(K, Gene, test.type, Topic) %>% group_by(K, Gene, test.type) %>% summarize(topic.count = n()) ## edgeR.test.df <- merge(edgeR.DE, perturbation.test.stat.df, by="Gene") ## MSigDB Pathway Enrichment ## Number of GO enrichment per model ## Notes 210518 ## also output pdf ## plot top enriched MSigDB pathways for each topic and K ## also consider eps file ## make sure that p.adjust is caluclated over all topics for a model ########################################################################################## ## GSEA plots ################################################## ## labels and parameters for the for loops ranking.types <- c("zscore", "raw") GSEA.types <- c("GOEnrichment", "ByWeightGSEA", "GSEA") GSEA.type.labels <- c("GO Term Enrichment\n(Top 300 Program Gene by Hypergeometric Test)", "Gene Set Enrichment\n(All Genes)", "Gene Set Enrichment\n(Top 300 Program Gene by Hypergeomteric Test)") ################################################## ## function to output plot in the same format plotGSEA <- function(toplot, nPathwayMetric, nPathwayMetricLabel) { p <- toplot %>% ggplot(aes(x=K, y=get(nPathwayMetric))) + geom_line(size=0.5) + geom_point(size=0.5) + mytheme + xlab("K") + ylab(nPathwayMetricLabel) #+ #ggtitle(paste0(GSEA.type.label)) + ## scale_x_continuous("K", labels = as.character(K), breaks = K) print(p) return(p) } ################################################## for (GSEA.type.i in 1:length(GSEA.types)) { GSEA.type <- GSEA.types[GSEA.type.i] GSEA.type.label <- GSEA.type.labels[GSEA.type.i] ################################################## ## process the GSEA data here tmp.df <- get(paste0("clusterProfiler.", GSEA.type, ".df")) %>% subset(p.adjust < fdr.thr) %>% group_by(K, type) %>% mutate(nPathways = n()) %>% select(K, type, nPathways, ID, Description) %>% unique %>% mutate(nUniquePathways = n(), normalizedNPathways = nPathways / K, normalizedNUniquePathways = nUniquePathways / K) %>% as.data.frame summary.df <- tmp.df %>% select(-ID, -Description) %>% unique %>% `rownames<-`(paste0(.$type, "_K", .$K)) K <- summary.df %>% pull(K) %>% unique() # K for x tick labels ################################################## for (ranking.type in ranking.types) { toplot <- summary.df %>% subset(type == ranking.type) ## subset to selected program gene ranking type (zscore or raw) nPathwayMetrics <- toplot %>% select(-K, -type) %>% colnames nPathwayMetricLabels <- c("# Total Pathways", "# Unique Pathways", "Average Pathways\nper Program", "Aveage Unique Pathways\nper Program") plotFilename <- paste0(FIGDIR, GSEA.type, "_", ranking.type) pdf(paste0(plotFilename, ".pdf"), width=3, height=3) plot.list <- list() ## create a list to store all metrics for (i in 1:length(nPathwayMetrics)) { ## make a line plot nPathwayMetric <- nPathwayMetrics[i] nPathwayMetricLabel <- nPathwayMetricLabels[i] plot.list[[nPathwayMetric]] <- plotGSEA(toplot, nPathwayMetric, nPathwayMetricLabel) } p <- plot_grid(plotlist = plot.list, nrow=length(plot.list), align="v", axis="lr") p <- annotate_figure(p, top = text_grob(paste0(GSEA.type.label, "\n", ranking.type), size=8)) print(p) eval(parse(text = paste0("p.", GSEA.type, ".", ranking.type, " <- p"))) dev.off() } } ## combine all plots in one panel p.all.GSEA <- plot_grid(p.GOEnrichment.zscore, p.GOEnrichment.raw, p.GSEA.zscore, p.GSEA.raw, p.ByWeightGSEA.zscore, p.ByWeightGSEA.raw, nrow = 3, ncol = 2, align="hv", axis="tblr") plotFilename <- paste0(FIGDIR, "/All_GSEA") pdf(paste0(plotFilename, ".pdf"), width=6, height=10) print(p.all.GSEA) ggsave(paste0(plotFilename, ".eps")) dev.off() ################################################## ## End of GSEA plots ################################################## ## Notes: ## why is there a sharp change from K=19 to K=21? ## Fraction of topics that has at least one significant (metric) versus K num.program.genes <- 300 enrichment.thr <- 1 ## plot number of TF enriched per K for {enhancers, promoters} for (ep.type in c("promoter", "enhancer")) { ep.type.label <- ifelse(ep.type == "promoter", "Promoter", "Enhancer") pdf(file=paste0(FIGDIR, "/TF.motif.enrichment.", ep.type, ".fdr.thr", as.character(fdr.thr), ".pdf"), width=3, height=3) ## promoters toplot <- get(paste0("all.", ep.type, ".ttest.df")) %>% mutate(significant = two.sided.p.adjust < fdr.thr & enrichment > enrichment.thr) %>% group_by(K) %>% summarise(total=significant %>% as.numeric %>% sum) %>% mutate(average.per.topic = total / K) %>% as.data.frame ## K <- toplot %>% pull(K) %>% unique() # K for x tick labels title <- paste0("Transcription Factors Enriched in \n", ep.type.label, " of Program Genes ") ## total number of significant TF plots p1 <- toplot %>% ggplot(aes(x=K, y=total)) + geom_line(size=0.5) + geom_point(size=0.5) + ## p1 <- toplot %>% ggplot(aes(x=K, y=total)) + geom_col(color="gray35") + xlab("K") + ylab(paste0("# Transcription Factors\n (FDR < ", fdr.thr, ")")) + mytheme + ## scale_x_continuous("K", labels = as.character(K), breaks = K) + ggtitle(title) print(p1) p2 <- toplot %>% ggplot(aes(x=K, y=average.per.topic)) + geom_line(size=0.5) + geom_point(size=0.5) + ## p2 <- toplot %>% ggplot(aes(x=K, y=average.per.topic)) + geom_col(width=0.5, color="gray35") + xlab("K") + ylab(paste0("Average # Transcription Factors per Program\n(FDR < ", fdr.thr, ")")) + mytheme + ## scale_x_continuous("K", labels = as.character(K), breaks = K) + ggtitle(paste0("Transcription Factors Enriched in \nPromoter of the Top ", num.program.genes, " Genes of Each Program")) print(p2) ## plot number of unique TF enriched per K for promoters toplot.unique <- get(paste0("all.", ep.type, ".ttest.df")) %>% mutate(significant = two.sided.p.adjust < fdr.thr & enrichment > enrichment.thr) %>% select(K, motif, significant) %>% unique() %>% group_by(K) %>% summarise(total=significant %>% as.numeric %>% sum) %>% mutate(average.per.topic = total / K) %>% as.data.frame K <- toplot.unique %>% pull(K) %>% unique() # K for x tick labels ## total number of significant TF plots p3 <- toplot.unique %>% ggplot(aes(x=K, y=total)) + geom_line(size=0.5) + geom_point(size=0.5) + xlab("K") + ylab(paste0("Number of Transcription Factors with adjusted p-value < ", fdr.thr)) + mytheme #+ ## scale_x_continuous("K", labels = as.character(K), breaks = K) print(p3) p4 <- toplot.unique %>% ggplot(aes(x=K, y=average.per.topic)) + geom_line(size=0.5) + geom_point(size=0.5) + xlab("K") + ylab(paste0("Average Number of Unique Transcription Factors per Program \n with adjusted p-value < ", fdr.thr)) + mytheme #+ ## scale_x_continuous("K", labels = as.character(K), breaks = K) + ggtitle(paste0("Unique Transcription Factors Enriched in \nPromoters of the Top 100 Genes of Each Program")) print(p4) p <- plot_grid(p1 + ggtitle("") + ylab("# TFs"), p2 + ggtitle("") + ylab("# TF per\nProgram"), p3 + ggtitle("") + ylab("# Unique TFs"), p4 + ggtitle("") + ylab("# Unique TF\nper Program"), nrow=4, align = "hv", axis="tblr") ## title = ep.type.label eval(parse(text = paste0("p.", ep.type, " <- annotate_figure(p, top = text_grob(label=title, face='bold', size=8))"))) print(get(paste0("p.", ep.type))) dev.off() } ## put together all motif enrichment plot in one panel p.all.TFMotifEnrichment <- plot_grid(p.promoter, p.enhancer, nrow=1) plotFilename <- paste0(FIGDIR, "/All_TFMotifEnrichment") pdf(paste0(plotFilename, ".pdf"), width=4, height=4) print(p.all.TFMotifEnrichment) ggsave(paste0(plotFilename, ".eps")) dev.off() ## ## cluster theta.zscore across topics ## ## old 211115 ## theta.zscore.df.wide <- theta.zscore.df %>% mutate(K_Factor = paste0("K",K,"_",Factor), Gene = rownames(.)) %>% select(Gene, weight, K_Factor) %>% spread(key = "K_Factor", value = "weight") ## write.table(theta.zscore.df.wide, paste0(OUTDIR, "topic.zscore.Pearson.corr.txt"), row.names=F, quote=F, sep="\t") ## theta.zscore.df.wide.mtx <- theta.zscore.df.wide %>% `rownames<-`(.$Gene) %>% select(-Gene) %>% as.matrix() ## d <- cor(theta.zscore.df.wide.mtx, method="pearson") ## m <- as.matrix(d) d <- cor(theta.zscore.df, method="pearson") m <- as.matrix(d) ## Function for plotting heatmap # new version (adjusted font size) plotHeatmap <- function(mtx, labCol, title, margins=c(12,6), ...) { #original heatmap.2( mtx %>% t(), Rowv=T, Colv=T, trace='none', key=T, col=palette, labCol=labCol, margins=margins, cex.main=0.8, cexCol=0.01 * ncol(mtx), cexRow=0.01 * ncol(mtx), #4.8/sqrt(nrow(mtx)) ## cexCol=1/(ncol(mtx)^(1/3)), cexRow=1/(ncol(mtx)^(1/3)), #4.8/sqrt(nrow(mtx)) main=title, ... ) } pdf(file=paste0(FIGDIR,"/cluster.topic.zscore.by.Pearson.corr.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.zscore.by.Pearson.corr.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() ## higher values of K index <- which(theta.zscore.df %>% colnames() %>% strsplit(split="_") %>% sapply("[[",1) %>% gsub("K","",.) >= 30) d <- cor(theta.zscore.df[index,index], method="pearson") m <- as.matrix(d) pdf(file=paste0(FIGDIR,"/cluster.topic.zscore.by.Pearson.corr.K30_higher.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.zscore.by.Pearson.corr.K30_higher.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() ## set correlation < 0.5 to zero to expand the color range m[m<0.5] <- 0 pdf(file=paste0(FIGDIR,"/cluster.topic.zscore.by.Pearson.corr.K30_higher.threshold_cor_0.5.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.zscore.by.Pearson.corr.K30_higher.threshold_cor_0.5.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic zscore clustering by Pearson Correlation")) dev.off() ## topic defined by raw weights d <- cor(theta.raw.df, method="pearson") m <- as.matrix(d) pdf(file=paste0(FIGDIR,"/cluster.topic.raw.by.Pearson.corr.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.raw.by.Pearson.corr.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() ## higher values of K index <- which(theta.raw.df %>% colnames() %>% strsplit(split="_") %>% sapply("[[",1) %>% gsub("K","",.) >= 30) d <- cor(theta.raw.df[index,index], method="pearson") m <- as.matrix(d) pdf(file=paste0(FIGDIR,"/cluster.topic.raw.by.Pearson.corr.K30_higher.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.raw.by.Pearson.corr.K30_higher.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() ## set correlation < 0.5 to zero to expand the color range m[m<0.5] <- 0 pdf(file=paste0(FIGDIR,"/cluster.topic.raw.by.Pearson.corr.K30_higher.threshold_cor_0.5.pdf"), width=30, height=30) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() png(file=paste0(FIGDIR, "/cluster.topic.raw.by.Pearson.corr.K30_higher.threshold_cor_0.5.png"), width=3000, height=3000) plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, topic raw weight clustering by Pearson Correlation")) dev.off() ########################################################################################## ## Variance Explained Plots ########################################################################################## toplot <- varianceExplainedByModel.df p <- toplot %>% ggplot(aes(x=K, y=Total)) + geom_point(size=0.5) + geom_line(size=0.5) + mytheme + xlab("K") + ylab("Fraction of Variance\nExplained by the Model") filename <- paste0(FIGDIR, "/variance.explained.by.model") pdf(paste0(filename, ".pdf"), width=3, height=1.5) print(p) dev.off() |
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 | packages <- c("optparse","dplyr", "ggplot2", "reshape2", "ggrepel", "conflicted", "Seurat", "SeuratObject") ## library(Seurat) xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/figures/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), make_option("--inputSeuratObject", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2105_FT005_Analysis/outputs/FT005_gex/withUMAP.SeuratObject.RDS", help="Path to the Seurat Object"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), # make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210625_snakemake_output/top3000VariableGenes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), # make_option("--K.list", type="character", default="2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,19,21,23,25", help="K values available for analysis"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), # make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), # make_option("--raw.mtx.dir",type="character",default="stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), # make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), # the first lane: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.expandedMultiTargetGuide.RDS" # make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), # make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/barcodes.tsv", help="barcodes.tsv for all cells"), # make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), # make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), # make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), # make_option("--enhancer.fimo.threshold", type="character", default="1.0E-4", help="Enhancer fimo motif match threshold"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 ## opt$inputSeuratObject <- paste0(opt$datadir,"/", SAMPLE, "_SeuratObjectUMAP.rds") mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) SAMPLE=strsplit(opt$sampleName,",") %>% unlist() # STATIC.SAMPLE=c("Telo_no_IL1B_T200_1", "Telo_no_IL1B_T200_2", "Telo_plus_IL1B_T200_1", "Telo_plus_IL1B_T200_2", "no_IL1B", "plus_IL1B", "pooled") # DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir ## TMDIR=opt$topic.model.result.dir ## SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr ## ## directories for factor motif enrichment ## FILENAME=opt$filename ## ## modify motif.enhancer.background input directory ##HERE: perhaps do a for loop for all the desired thresholds (use strsplit on enhancer.fimo.threshold) ## opt$motif.enhancer.background <- paste0(opt$motif.enhancer.background, opt$enhancer.fimo.threshold, "/fimo.formatted.tsv") # create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE, FGSEADIR, FGSEAFIG) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## selected.gene <- c("EDN1", "NOS3", "TP53", "GOSR2", "CDKN1A") # ABC genes # gene.set <- c("INPP5B", "SF3A3", "SERPINH1", "NR2C1", "FGD6", "VEZT", "SMAD3", "AAGAB", "GOSR2", "ATP5G1", "ANGPTL4", "SRBD1", "PRKCE", "DAGLB") # ABC_0.015_CAD_pp.1_genes #200 gene library # # cell cycle genes # ## need to update these for 2kG library # gene.list.three.groups <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/ptbd.genes_three.groups.txt"), header=T, stringsAsFactors=F) # enhancer.set <- gene.list.three.groups$Gene[grep("E_at_", gene.list.three.groups$Gene)] # CAD.focus.gene.set <- gene.list.three.groups %>% subset(Group=="CAD_focus") %>% pull(Gene) %>% append(enhancer.set) # EC.pos.ctrl.gene.set <- gene.list.three.groups %>% subset(Group=="EC_pos._ctrls") %>% pull(Gene) # cell.count.thr <- opt$cell.count.thr # greater than this number, filter to keep the guides with greater than this number of cells # guide.count.thr <- opt$guide.count.thr # greater than this number, filter to keep the perturbations with greater than this number of guides # guide.design = read.delim(file=paste0(DATADIR, "/200607_ECPerturbSeqMiniPool.design.txt"), header=T, stringsAsFactors = F) # ## add GO pathway log2FC # GO <- read.delim(file=paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.table.brief.txt"), header=T, check.names=FALSE) # GO.list <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/GO.Pathway.list.brief.txt", header=T, check.names=F) # colnames(GO)[1] <- "Gene" # colnames(GO.list)[1] <- "Gene" # ## load all sample, K, topic's top 100 genes (by TopFeatures() KL-score measure) # ## allGeneKtopic100 <- read.delim(paste0(TMDIR, "no.plus.pooled.top100.topicStats.txt"), header=T) # # load non-expressed control gene list # non.expressed.genes <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/non.expressed.ctrl.genes.txt", header=F, stringsAsFactors=F) %>% unlist %>% as.character() %>% sort() # # perturbation type list # gene.set.type.df <- data.frame(Gene=guide.design %>% pull(guideSet) %>% unique(), # type=rep("other", guide.design %>% pull(guideSet) %>% unique() %>% length())) # gene.set.type.df$Gene <- gene.set.type.df$Gene %>% as.character() # gene.set.type.df$type <- gene.set.type.df$type %>% as.character() # gene.set.type.df$type[which(gene.set.type.df$Gene %in% non.expressed.genes)] <- "non-expressed" # gene.set.type.df$type[which(gene.set.type.df$Gene %in% CAD.focus.gene.set)] <- "CAD focus" # gene.set.type.df$type[grepl("^safe|^negative", gene.set.type.df$Gene)] <- "negative-control" # gene.set.type.df$Gene[which(gene.set.type.df$Gene == "negative_control")] <- "negative-control" # gene.set.type.df$Gene[which(gene.set.type.df$Gene == "safe_targeting")] <- "safe-targeting" # # gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # gene.set.type.df.200 <- gene.set.type.df # # reference table # ref.table <- read_xlsx(opt$reference.table, sheet="2000_gene_library_annotated") # gene.set.type.df <- ref.table %>% select(Symbol, `Class(es)`) %>% `colnames<-`(c("Gene", "type")) # gene.set.type.df$type[grepl("EC_ctrls", gene.set.type.df$type)] <- "EC_ctrls" # gene.set.type.df$type[grepl("NonExpressed", gene.set.type.df$type)] <- "non-expressed" # gene.set.type.df$type[grepl("abc.015", gene.set.type.df$type)] <- "ABC" # gene.set.type.df <- rbind(gene.set.type.df, c("negative-control", "negative-control"), c("safe-targeting", "safe-targeting")) # non.expressed.genes <- gene.set.type.df %>% subset(type == "non-expressed") %>% pull(Gene) # # ABC genes # gene.set <- gene.set.type.df %>% subset(grepl("ABC", type)) %>% pull(Gene) # ## add GWAS classification # modified.ref.table <- ref.table %>% mutate(GWAS.classification="") # CAD.index <- which(grepl("CAD_Loci",ref.table$`Class(es)`)) # EC_ctrls.index <- which(grepl("^EC_ctrls",ref.table$`Class(es)`)) # ABC_linked.index <- which(grepl("MIG_etc",ref.table$`Class(es)`)) # IBD.index <- which(grepl("Non-CAD_loci_IBD",ref.table$`Class(es)`)) # non.expressed.index <- which(grepl("NonExpressed",ref.table$`Class(es)`)) # poorly.annotated.9p21.index <- which(grepl("9p21",ref.table$`Class(es)`)) # # length(CAD.index) + length(EC_ctrls.index) + length(ABC_linked.index) + length(IBD.index) + length(non.expressed.index) + length(poorly.annotated.9p21.index) # modified.ref.table$GWAS.classification[ABC_linked.index] <- "ABC" # modified.ref.table$GWAS.classification[IBD.index] <- "IBD" # modified.ref.table$GWAS.classification[non.expressed.index] <- "NonExpressed" # modified.ref.table$GWAS.classification[poorly.annotated.9p21.index] <- "9p21.poorly.annotated" # modified.ref.table$GWAS.classification[EC_ctrls.index] <- "EC_ctrls" # modified.ref.table$GWAS.classification[CAD.index] <- "CAD" # modified.ref.table <- modified.ref.table %>% group_by(GWAS.classification) %>% mutate(gene.count.per.GWAS.category = n()) # ref.table <- modified.ref.table # ## add TSS distance to SNP # modified.ref.table <- ref.table %>% mutate(TSS.dist.to.SNP = abs(`TSS v. SNP loc`)) # not.in.SNP.index <- which(is.na(modified.ref.table$`TSS v. SNP loc`)) # modified.ref.table$TSS.dist.to.SNP[not.in.SNP.index] <- NA # ref.table <- modified.ref.table %>% ungroup() # ## add closest gene to top GWAS loci ranking # modified.ref.table <- ref.table %>% # group_by(`Top SNP ID`) %>% # per SNP metrics # arrange(abs(`TSS v. SNP loc`)) %>% # mutate(TSS.v.SNP.ranking = 1:n(), # total.gene.in.this.loci = n()) %>% ungroup() %>% # group_by(`Top SNP ID`, GWAS.classification) %>% # per SNP per GWAS class (CAD, IBD, NonExpressed, ABC, 9p21.poorly.annotated) # arrange(abs(`TSS v. SNP loc`)) %>% # mutate(TSS.v.SNP.ranking.in.GWAS.category = 1:n(), # total.gene.in.this.loci.in.GWAS.category = n()) %>% ungroup() # not.in.SNP.index <- which(is.na(modified.ref.table$`TSS v. SNP loc`)) # modified.ref.table$TSS.v.SNP.ranking.in.GWAS.category[not.in.SNP.index] <- NA # modified.ref.table$TSS.v.SNP.ranking[not.in.SNP.index] <- NA # ref.table <- modified.ref.table # ## add gene count per distance ranking per GWAS loci # modified.ref.table <- ref.table # modified.ref.table <- modified.ref.table %>% # group_by(TSS.v.SNP.ranking) %>% # per ranking, not considering which GWAS category the gene is from # mutate(total.TSS.v.SNP.ranking.count = n()) %>% ungroup() %>% # group_by(GWAS.classification, TSS.v.SNP.ranking.in.GWAS.category) %>% # per GWAS category and per ranking # mutate(total.TSS.v.SNP.ranking.count.per.GWAS.classification = n()) %>% ungroup() # not.in.SNP.index <- which(is.na(modified.ref.table$TSS.v.SNP.ranking)) # modified.ref.table$total.TSS.v.SNP.ranking.count[not.in.SNP.index] <- NA # modified.ref.table$total.TSS.v.SNP.ranking.count.per.GWAS.classification[not.in.SNP.index] <- NA # ref.table <- modified.ref.table # write.table(ref.table, file=paste0(opt$datadir, "/ref.table.txt"), row.names=F, quote=F, sep="\t") # ## ref.table ranking count summary table # ref.table.gene.to.SNP.dist.ranking.count.summary.allGWAS <- ref.table %>% select(TSS.v.SNP.ranking, total.TSS.v.SNP.ranking.count) %>% mutate(GWAS.classification="all") %>% unique() # ref.table.gene.to.SNP.dist.ranking.count.summary.indGWAS <- ref.table %>% select(TSS.v.SNP.ranking.in.GWAS.category, total.TSS.v.SNP.ranking.count.per.GWAS.classification, GWAS.classification) %>% `colnames<-`(c("TSS.v.SNP.ranking", "total.TSS.v.SNP.ranking.count", "GWAS.classification")) %>% unique() # ref.table.gene.to.SNP.dist.ranking.count.summary <- rbind(ref.table.gene.to.SNP.dist.ranking.count.summary.allGWAS, ref.table.gene.to.SNP.dist.ranking.count.summary.indGWAS) # ref.table.summary.na.index <- which(is.na(ref.table.gene.to.SNP.dist.ranking.count.summary$TSS.v.SNP.ranking)) # ref.table.gene.to.SNP.dist.ranking.count.summary <- ref.table.gene.to.SNP.dist.ranking.count.summary[-ref.table.summary.na.index,] # rm(ref.table.summary.na.index) # # convert enhancer SNP rs number to enhancer target gene name # need 2kG library version # enh.snp.to.gene <- read.delim(paste0(DATADIR, "/enhancer.SNP.to.gene.name.txt"), header=T, stringsAsFactors = F) %>% mutate(Enhancer_name=gsub("_","-", Enhancer_name)) # # gene corresponding pathway # gene.def.pathways <- read_excel(paste0(DATADIR,"topic.gene.definition.pathways.xlsx"), sheet="Gene_Pathway") # ## Gavin's new list # gene.classes.ranked <- read.table(paste0(opt$datadir, "Gene_Classes_Ranked_for_CAD_n_EC.txt"), header=T, stringsAsFactors = F) # summaries <- read.delim(paste0(opt$datadir, "Gene_Summaries_n_Classes.txt"), sep="\t", header=T, stringsAsFactors = F) # gene.summaries <- read_xlsx(paste0(opt$datadir, "Gene_Summaries.xlsx"), sheet="uniprot_summaries") # ## Perturbation name and 10X gene name conversion table # ptb.10X.name.conversion <- read_xlsx(paste0(opt$datadir, "Perturbation 10X names.xlsx")) # ## EdgeR log2fcs and p-values # log2fc.edgeR <- read.table(paste0(opt$datadir, "/EdgeR/ALL_log2fcs_dup4_s4n3.99x.txt"), header=T, stringsAsFactors=F) # p.value.edgeR <- read.table(paste0(opt$datadir, "/EdgeR/ALL_Pvalues_dup4_s4n3.99x.txt"), header=T, stringsAsFactors=F) # print("loaded all prerequisite data") ###################################################################### ## Process topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } else { warning(paste0(cNMF.result.file, " does not exist")) } ## load ann.omega file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.",SUBSCRIPT,".RData") print(file.name) if(file.exists((file.name))) { print(paste0("loading ",file.name)) load(file.name) } # ## load statistical test data # toSave.features <- read.delim(paste0(OUTDIRSAMPLE, "/topic.KL.score_K", k, ".dt_", DENSITY.THRESHOLD, ".txt"), header=T, stringsAsFactors=F) # all.test <- read.delim(file=paste0(OUTDIRSAMPLE, "/all.test.", SUBSCRIPT, ".txt"), header=T, stringsAsFactors=F) ##here # realPvals.df <- read.delim(file=paste0(OUTDIRSAMPLE, "/all.expressed.genes.pval.fdr.",SUBSCRIPT,".txt"), header=T, stringsAsFactors=F) # all.test.guide.w <- all.test %>% subset(test.type==opt$test.type) # realPvals.df.guide.w <- realPvals.df %>% subset(test.type==opt$test.type) # fdr.thr <- 0.1 # topFeatures.raw.weight <- theta.zscore %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene", variable.name="topic", value.name="scores") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) ## load Seurat Object with UMAP # ## load full matrix # file.name=paste0(opt$raw.mtx.RDS.dir, "_modified.multiTargetGuide.cell.names.RDS") # fc.file.name=paste0(opt$raw.mtx.RDS.dir, "_FC_modified.multiTargetGuide.cell.names.RDS") # log2fc.file.name=paste0(opt$raw.mtx.RDS.dir, "_log2FC_modified.multiTargetGuide.cell.names.RDS") # if(file.exists(file.name) & file.exists(fc.file.name) & file.exists(log2fc.file.name)) { # print(paste0("Loading ", file.name)) # X.full = readRDS(file.name) # print(paste0("Loading ", fc.file.name)) # fc.X.full = readRDS(fc.file.name) # print(paste0("Loading ", log2fc.file.name)) # log2fc.X.full = readRDS(log2fc.file.name)##here210813 # } else { # warning(paste0("full scRNA-seq matrix does not exist")) # } # ## modify X.full # ## old 210819 # tokeep.index <- which(rownames(X.full) %in% ann.omega.filtered$long.CBC) ## need to expand X.full to match ann.omega.filtered due to guides that target two promoters # fc.X.full <- fc.X.full[tokeep.index,] # X.full <- X.full[tokeep.index,] # colnames(X.full) <- colnames(fc.X.full) <- colnames(fc.X.full) %>% strsplit(., split=":") %>% sapply("[[",1) # ## ## adjust colnames, remove ENSG number # ann.X.full.filtered <- X.full # ## add back ENSG names? # ## tmp <- colnames(ann.X.full.filtered) %>% strsplit(., split=":") %>% sapply("[[",1) # ## tmpp <- data.frame(table(tmp)) %>% subset(Freq > 1) # keep row names that have duplicated gene names but different ENSG names # ## tmp.copy <- tmp # ## tmp.copy[grepl(paste0(tmpp$tmp,collapse="|"),tmp)] <- colnames(ann.X.full.filtered)[grepl(paste0(tmpp$tmp,collapse="|"), colnames(ann.X.full.filtered))] # ## colnames(ann.X.full.filtered) <- tmp.copy # the above section takes a while # ## end of old 210819 # ## get ctrl log2 transformed expression # tokeep.index <- which(grepl("control|targeting",rownames(X.full))) # ## ctrl.X <- ann.X.full.filtered %>% subset(grepl("control|targeting",Gene)) # ctrl.X <- X.full[tokeep.index,] # fc.ctrl.X <- fc.X.full[tokeep.index,] # ## get ctrl topic weights # ctrl.ann.omega <- ann.omega.filtered %>% subset(grepl("control|targeting",Gene)) %>% `rownames<-`(.$long.CBC) # X.gene.names <- rownames(X.full) %>% strsplit(., split=":") %>% sapply("[[",1) %>% gsub("_multiTarget|-TSS2","",.) ## load motif enrichment results # file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.factorMotifEnrichment.",SUBSCRIPT.SHORT,".RData") # print(file.name) # if(file.exists((file.name))) { # load(file.name) # print(paste0("loading ", file.name)) # } # motif.enrichment.variables <- c("all.enhancer.fisher.df", "all.promoter.fisher.df", # "promoter.wide", "enhancer.wide", "promoter.wide.binary", "enhancer.wide.binary", # "enhancer.wide.10en6", "enhancer.wide.binary.10en6", "all.enhancer.fisher.df.10en6", # "promoter.wide.10en6", "promoter.wide.binary.10en6", "all.promoter.fisher.df.10en6", # "all.promoter.ttest.df", "all.promoter.ttest.df.10en6", "all.enhancer.ttest.df", "all.enhancer.ttest.df.10en6") # motif.enrichment.variables.missing <- (!(motif.enrichment.variables %in% ls())) %>% as.numeric %>% sum # if ( motif.enrichment.variables.missing > 0 ) { # warning(paste0(motif.enrichment.variables[!(motif.enrichment.variables %in% ls())], " not available")) # } # ## load count.by.GWAS # count.by.GWAS <- read.delim(file=paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt"), header=T, stringsAsFactors = F) # count.by.GWAS.withTopic <- read.delim(file=paste0(OUTDIRSAMPLE,"/count.by.GWAS.classes.withTopic_p.adj.",p.value.thr %>% as.character,"_",SUBSCRIPT,".txt"), header=T, stringsAsFactors=F) # ## load UMAP data # file.name <- paste0(OUTDIRSAMPLE, "gene.score.SeuratObject.RDS") # if(file.exists(file.name)) { # s.gene.score <- readRDS(file.name) # } else { # warning(paste0(file.name, " does not exist")) # } ## full UMAP file.name <- opt$inputSeuratObject # file.name <- paste0(opt$datadir,"/", SAMPLE, "_SeuratObjectUMAP.rds") ## todo: change to calcUMAP output ## subset.file.name <- paste0(opt$datadir,"/", SAMPLE, "_subset_SeuratObjectUMAP.rds") options(future.globals.maxSize=1000*1024^2) s <- readRDS(file.name) ## s@meta.data <- s@meta.data[,-which(grepl("topic",colnames(s@meta.data)))] s <- AddMetaData(s, metadata = omega, col.name = paste0("K",k,"_",colnames(omega))) ## End of data loading ########################################################################## ## Plots # ########################################################################## # ## topic gene z-score list # pdf(file=paste0(FIGDIRTOP,"top50GeneInTopics.zscore.pdf"), width=4, height=6) # topFeatures.raw.weight <- theta.zscore %>% as.data.frame() %>% mutate(Gene=rownames(.)) %>% melt(id.vars="Gene", variable.name="topic", value.name="scores") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) # for ( t in 1:dim(theta)[2] ) { # toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), # Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) # p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100) ) + geom_col() + theme_minimal() # p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("z-score (Specificity)") + ggtitle(paste(SAMPLE, ", Topic ", t, sep="")) + mytheme # print(p) # } # dev.off() ## ########################################################################## ## ## top expressed genes per topic by KL specificity score list ## pdf(file=paste0(FIGDIRTOP, "topGeneInTopics.KL.pdf"), width=4, height=6) ## topFeatures <- toSave.features %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) ## for ( t in 1:dim(theta)[2] ) { ## toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), ## Score=topFeatures %>% subset(topic == t) %>% pull(scores)) ## p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100) ) + geom_col() + theme_minimal() ## p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("KL score (gene specific to this topic)") + ggtitle(paste(SAMPLE, ", K = ", k, ", Topic ", t, sep="")) + mytheme ## print(p) ## } ## dev.off() # ########################################################################## # ## Topic's top gene list, ranked by raw weight # pdf(file=paste0(FIGDIRTOP,"top50GeneInTopics.rawWeight.pdf"), width=4, height=6) # topFeatures <- theta %>% as.data.frame() %>% mutate(genes=rownames(.)) %>% melt(id.vars="genes",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:50) # for ( t in 1:dim(theta)[2] ) { # toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), # Score=topFeatures %>% subset(topic == t) %>% pull(scores)) # p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100) ) + geom_col() + theme_minimal() # p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("Raw Score (gene's weight in topic)") + ggtitle(paste(SAMPLE, ", Topic ", t, sep="")) + mytheme # print(p) # } # dev.off() ## ########################################################################## ## ## KL score list with annotataion ## pdf(file=paste0(FIGDIRTOP,"topGeneInTopics.annotated.KL.pdf"), width=4.5, height=5) ## topFeatures <- toSave.features %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) ## for ( t in 1:dim(theta)[2] ) { ## toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), ## Score=topFeatures %>% subset(topic == t) %>% pull(scores)) %>% ## merge(., gene.def.pathways, by="Gene", all.x=T) ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" ## p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100, fill=Pathway) ) + geom_col(width=0.5) + theme_minimal() ## p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("KL score (gene specific to this topic)") + ## mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) ## print(p4) ## } ## dev.off() # ########################################################################## # ## raw program TPM list with annotataion # pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.TPM.pdf"), width=4.5, height=5) # topFeatures <- theta %>% as.data.frame() %>% mutate(genes=rownames(.)) %>% melt(id.vars="genes",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) # for ( t in 1:dim(theta)[2] ) { # toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), # Score=topFeatures %>% subset(topic == t) %>% pull(scores)) # %>% # ## merge(., gene.def.pathways, by="Gene", all.x=T) # ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*1000000) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() # p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("Raw Weight (in TPM)") + # mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) # print(p4) # } # dev.off() # ########################################################################## # ## raw program zscore list (top 10) (can potentially include annotation) # pdf(file=paste0(FIGDIRTOP,"top10GeneInTopics.zscore.pdf"), width=4.5, height=5) # topFeatures <- theta.zscore %>% as.data.frame() %>% mutate(genes=rownames(.)) %>% melt(id.vars="genes",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) # for ( t in 1:dim(theta)[2] ) { # toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), # Score=topFeatures %>% subset(topic == t) %>% pull(scores)) # %>% # ## merge(., gene.def.pathways, by="Gene", all.x=T) # ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() # p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("z-score") + # mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) # print(p4) # } # dev.off() ## ########################################################################## ## ## raw program zscore list without annotation ## pdf(file=paste0(FIGDIRTOP,"topGeneInTopics.shortList.zscore.pdf"), width=3.5, height=4) ## topFeatures <- theta.zscore %>% as.data.frame() %>% mutate(genes=rownames(.)) %>% melt(id.vars="genes",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) ## for ( t in 1:dim(theta)[2] ) { ## toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), ## Score=topFeatures %>% subset(topic == t) %>% pull(scores)) ## p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(width=0.5, fill="#38b4f7") + theme_minimal() ## p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("z-score") + ## mytheme + theme(legend.position="bottom", legend.direction="vertical", text=element_text(size=16), plot.title=element_text(size=12)) + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) ## print(p4) ## } ## dev.off() # ########################################################################## # ## Perturbation zscore list with annotataion # pdf(file=paste0(FIGDIRTOP,"Perturbation_zscore.annotated.pdf"), width=4.5, height=5) # topFeatures <- ptb.zscore %>% as.data.frame() %>% mutate(genes=rownames(.)) %>% melt(id.vars="genes",value.name="scores", variable.name="topic") %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) %>% mutate(topic = gsub("topic_","", topic)) # for ( t in 1:dim(theta)[2] ) { # toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), # Score=topFeatures %>% subset(topic == t) %>% pull(scores)) %>% # merge(., gene.def.pathways, by="Gene", all.x=T) # toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score, fill=Pathway) ) + geom_col(width=0.5) + theme_minimal() # p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("Perturbation z-score") + # mytheme + theme(legend.position="bottom", legend.direction="vertical") + ggtitle(paste0(SAMPLE, ", K = ", k, ", Topic ", t)) # print(p4) # } # dev.off() # pdf(file=paste0(FIGDIRTOP, "Perturbation.zscore.sig.list.pdf"), width=6, height=6) # ptb.zscore.long <- ptb.zscore %>% as.data.frame %>% mutate(Gene=rownames(.)) %>% melt(id.vars = "Gene", value.name = "perturbation.zscore", variable.name = "Topic") # for ( topic in colnames(ptb.zscore) ) { # t <- gsub("topic_","",topic) # toPlot.all.test <- all.test %>% subset(test.type=="per.cell.wilcoxon" & Topic==topic) # toPlot.fdr <- realPvals.df %>% subset(test.type=="per.cell.wilcoxon" & Topic == topic) %>% select(Gene,fdr)##here210809 # # assemble toPlot # toPlot <- ptb.zscore.long %>% subset(Topic == topic) %>% merge(.,toPlot.all.test,by=c("Gene","Topic"), all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene", all.x=T) %>% ##here210809 # ## merge(.,gene.def.pathways, by="Gene", all.x=T) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(perturbation.zscore)) %>% slice(1:25) # toPlot.bottom <- toPlot %>% arrange(perturbation.zscore) %>% slice(1:25) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(grepl("CAD", type), "red", # ifelse(type=="non-expressed", "gray", # ifelse(type=="EC_ctrls", "blue", "black")))) %>% # mutate(color=ifelse(is.na(type), "black", color)) # ## colors <- toPlot.extreme$color[order(toPlot.extreme %>% arrange(desc(perturbation.zscore)) %>% pull(color))] # toPlot.extreme <- toPlot.extreme %>% arrange(perturbation.zscore) # ## add gene distance to CAD # toPlot.extreme$ann.Gene <- factor(toPlot.extreme$ann.Gene, levels = toPlot.extreme$ann.Gene) # p <- toPlot.extreme %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=ann.Gene, y=perturbation.zscore, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("Perturbation z-score") + ggtitle(paste(SAMPLE, " perturbations, ", topic)) + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none", axis.text.y = element_text(colour = toPlot.extreme$color)) # print(p) # } # dev.off() # pdf(file=paste0(FIGDIRTOP, "Perturbation.zscore.sig.shortList.pdf"), width=6, height=6) # ptb.zscore.long <- ptb.zscore %>% as.data.frame %>% mutate(Gene=rownames(.)) %>% melt(id.vars = "Gene", value.name = "perturbation.zscore", variable.name = "Topic") # for ( topic in colnames(ptb.zscore) ) { # t <- gsub("topic_","",topic) # toPlot.all.test <- all.test %>% subset(test.type=="per.cell.wilcoxon" & Topic==topic) # toPlot.fdr <- realPvals.df %>% subset(test.type=="per.cell.wilcoxon" & Topic == topic) %>% select(Gene,fdr)##here210809 # # assemble toPlot # toPlot <- ptb.zscore.long %>% subset(Topic == topic) %>% merge(.,toPlot.all.test,by=c("Gene","Topic"), all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene", all.x=T) %>% ##here210809 # ## merge(.,gene.def.pathways, by="Gene", all.x=T) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(perturbation.zscore)) %>% slice(1:10) # toPlot.bottom <- toPlot %>% arrange(perturbation.zscore) %>% slice(1:10) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(grepl("CAD", type), "red", # ifelse(type=="non-expressed", "gray", # ifelse(type=="EC_ctrls", "blue", "black")))) %>% # mutate(color=ifelse(is.na(type), "black", color)) # ## colors <- toPlot.extreme$color[order(toPlot.extreme %>% arrange(desc(perturbation.zscore)) %>% pull(color))] # toPlot.extreme <- toPlot.extreme %>% arrange(perturbation.zscore) # ## add gene distance to CAD # toPlot.extreme$ann.Gene <- factor(toPlot.extreme$ann.Gene, levels = toPlot.extreme$ann.Gene) # p <- toPlot.extreme %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=ann.Gene, y=perturbation.zscore, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("Perturbation z-score") + ggtitle(paste(SAMPLE, " perturbations, ", topic)) + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none", axis.text.y = element_text(colour = toPlot.extreme$color)) # print(p) # } # dev.off() # ## volcano plots # volcano.plot <- function(toplot, ep.type, ranking.type, label.type="") { # if( label.type == "pos") { # label <- toplot %>% subset(-log10(p.adjust) > 1 & enrichment.log2fc > 0) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) # } else { # label <- toplot %>% subset(-log10(p.adjust) > 1) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) # } # t <- gsub("topic_", "", toplot$topic[1]) # p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(p.adjust))) + geom_point(size=0.5) + mytheme + # ggtitle(paste0(SAMPLE[1], " Topic ", t, " Top 100 ", ranking.type," ", ifelse(ep.type=="promoter", "Promoter", "Enhancer"), " Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(adjusted p-value)") + # geom_text_repel(data=label, box.padding = 0.5, # aes(label=motif.toshow), size=5, # color="black") + theme(text=element_text(size=16), axis.title=element_text(size=16), axis.text=element_text(size=16), plot.title=element_text(size=14)) # print(p) # p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(p.value))) + geom_point(size=0.5) + mytheme + # ggtitle(paste0(SAMPLE[1], " Topic ", t, " Top 100 ", ranking.type," ", ifelse(ep.type=="promoter", "Promoter", "Enhancer"), " Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(p-value)") + # geom_text_repel(data=label, box.padding = 0.5, # aes(label=motif.toshow), size=5, # color="black") + theme(text=element_text(size=16), axis.title=element_text(size=16), axis.text=element_text(size=16), plot.title=element_text(size=14)) # return(p) # } # ## function for all volcano plots # all.volcano.plots <- function(all.fisher.df, ep.type, ranking.type, label.type="") { # for ( t in 1:k ){ # toplot <- all.fisher.df %>% subset(topic==paste0("topic_",t)) # volcano.plot(toplot, ep.type, ranking.type, label.type) %>% print() # } # } # if(opt$subsample.type!="ctrl") { # ########################################################################## # ## q-q plot # pdf(file=paste0(FIGDIRTOP,"p-value.qqplot.pdf"), width=8, height=8) # for (test.name in unique(all.test$test.type)) { # toPlot <- all.test %>% subset(test.type==test.name) # toPlot$p.value <- -1*log10(toPlot$p.value) # min.value <- min(toPlot %>% subset(gene.type=="expressed") %>% select(p.value) %>% unlist() %>% as.numeric(), # toPlot %>% subset(gene.type=="non-expressed") %>% select(p.value) %>% unlist() %>% as.numeric()) # max.value <- 10 # toPlot$p.value[toPlot$p.value < 10^-10] <- 10^-10 # exp.gene.count <- toPlot %>% subset(gene.type=="expressed") %>% select(Gene) %>% unique() %>% unlist() %>% length() # nonexp.gene.count <- toPlot %>% subset(gene.type=="non-expressed") %>% select(Gene) %>% unique() %>% unlist() %>% length() # match.df <- toPlot %>% subset(adjusted.p.value < 0.1 & gene.type=="expressed") # my.qqplot(y = toPlot %>% subset(gene.type=="expressed") %>% select(p.value) %>% unlist() %>% as.numeric(), # x = toPlot %>% subset(gene.type=="non-expressed") %>% select(p.value) %>% unlist() %>% as.numeric(), # xlimit = c(min.value, max.value), ylimit = c(min.value, max.value), # ylab = paste0("Expressed Genes (-log10 p-value, n=[", exp.gene.count, " genes])"), # xlab = paste0("Control - Non-Expressed Genes (-log10 p-value, n=[", nonexp.gene.count, " genes])"), # main = paste0(SAMPLE, ", K = ", k, ", ", test.name), # match=T, match.y=T, match.df=match.df # ) # } # dev.off() # ########################################################################## # ## empirical fdr vs p.adjust (BH) plot # pdf(paste0(FIGDIRTOP,"empirical.fdr.vs.p.adjust.pdf"), width=8, height=6) # for (j in 1:length(unique(realPvals.df$test.type))) { # test.name<-unique(all.test$test.type)[j] # toPlot <- realPvals.df %>% subset(test.type==test.name) # p <- toPlot %>% ggplot(aes(x=adjusted.p.value, y=fdr)) + geom_point(size=0.1) + geom_abline(color="red") + # mytheme + xlab("Adjusted p-value") + ylab("Empirical False Discovery Rate") + ggtitle(paste0(SAMPLE, ", ", test.name)) + coord_fixed() # print(p) # } # dev.off() # ########################################################################## # ## empirical FDR heatmaps # for(fdr.method in c("empirical.fdr", "p.adjust")) { # pdf(file=paste0(FIGDIRTOP, fdr.method, ".sig.ptbd.gene_fill.log2fc_heatmap.pdf"), width=12, height=6) # for (emp.fdr.thr in c(0.05, 0.1, 0.25)) { # for (current.test.type in realPvals.df$test.type %>% unique()) { # if(fdr.method == "empirical.fdr") { # test.list <- realPvals.df %>% subset(test.type==current.test.type) %>% mutate(adjusted.p.value=fdr) # } else { # test.list <- realPvals.df %>% subset(test.type==current.test.type) # } # genes.toInclude <- test.list %>% subset(adjusted.p.value < emp.fdr.thr) %>% pull(Gene) %>% unique() # toPlot <- gene.score %>% subset(., rownames(gene.score) %in% genes.toInclude) # toPlot <- add.snp.gene.info(toPlot, type="rownames") # # plot heatmap # cols <- rep('black', nrow(toPlot)) # #turn red the specified rows in tf # cols[row.names(toPlot) %in% (gene.set.type.df %>% subset(grepl("EC_ctrls", type)) %>% pull(Gene))] <- "blue" # cols[row.names(toPlot) %in% (gene.set.type.df %>% subset(grepl("CAD_Loci_all", type)) %>% pull(Gene))] <- "red" # # cols[row.names(toPlot) %in% gene.set] <- "red" # rownames(toPlot)[row.names(toPlot) %in% gene.set] <- paste0("[ ", rownames(toPlot)[which(row.names(toPlot) %in% gene.set)], " ]") # if(nrow(toPlot) > 1) { # ## plotHeatmap( toPlot, cellNote=NULL, rownames(toPlot), title=title, colCol=cols) # toHighlight.asterisk <- merge.score.with.test(gene.score, test.list, test.col.name="p.value", p.value.thr=1, adj.p.value.thr=emp.fdr.thr, fill.all=T, fill="", overlay=T)$score.mtx # title=paste0(SAMPLE,", K = ", k, ", ", current.test.type, ", \n", ifelse(fdr.method == "empirical.fdr", "empirical fdr", "BH adjusted p-value"), " < ", emp.fdr.thr, ", number of significant genes = ", nrow(toPlot)) # plotHeatmap( toPlot, cellNote=toHighlight.asterisk, rownames(toPlot), title=title, colCol=cols) # toHighlight.value <- merge.score.with.test(gene.score, test.list, test.col.name="p.value", p.value.thr=1, adj.p.value.thr=emp.fdr.thr, fill.all=T, fill="", overlay=F, num.thr = 1)$score.mtx # plotHeatmap( toPlot, cellNote=toHighlight.value, rownames(toPlot), title=title, colCol=cols) # toHighlight.value <- merge.score.with.test(gene.score, test.list, test.col.name="p.value", p.value.thr=1, adj.p.value.thr=emp.fdr.thr, fill.all=T, fill="", overlay=F)$score.mtx # plotHeatmap( toPlot, cellNote=toHighlight.value, rownames(toPlot), title=title, colCol=cols) # } # } # } # dev.off() # } # ## ########################################################################## # ## full heatmap # pdf(file = paste0(FIGDIRTOP,"Gene.full.heatmap.pdf"), width=36, height=6) # current.test.type <- "per.guide.wilcoxon" # toPlot <- gene.score # toPlot <- add.snp.gene.info(toPlot, type="rownames") # cols <- rep('black', nrow(toPlot)) # #turn red the specified rows in tf # cols <- rep('black', nrow(toPlot)) # #turn red the specified rows in tf # cols[row.names(toPlot) %in% (gene.set.type.df %>% subset(grepl("EC_ctrls", type)) %>% pull(Gene))] <- "blue" # cols[row.names(toPlot) %in% (gene.set.type.df %>% subset(grepl("CAD_Loci_all", type)) %>% pull(Gene))] <- "red" # rownames(toPlot)[row.names(toPlot) %in% gene.set] <- paste0("[ ", rownames(toPlot)[which(row.names(toPlot) %in% gene.set)], " ]") # title=paste0(SAMPLE,", K = ", k) # plotHeatmap(toPlot, cellNote=NULL, rownames(toPlot), title=title, colCol=cols) # for(emp.fdr.thr in c(0.05, 0.1, 0.25)){ # # add asterisks for significant perturbation/topic # # test.list <- realPvals.df %>% subset(test.type==current.test.type) %>% mutate(adjusted.p.value=fdr) # # toHighlight.asterisk <- merge.score.with.test(gene.score, test.list, test.col.name="p.value", p.value.thr=1, adj.p.value.thr=emp.fdr.thr, fill.all=T, fill="", overlay=T)$score.mtx # # title=paste0(SAMPLE,", K = ", k, ", ", current.test.type, ", \nempirical fdr < ", emp.fdr.thr) # # plotHeatmap(toPlot, cellNote=toHighlight.asterisk, rownames(toPlot), title=title, colCol=cols) # } # dev.off() # ## ########################################################################## # ## Lists of genes or perturbations to understand topics # all.test.guide.w <- all.test %>% subset(test.type=="per.guide.wilcoxon") # realPvals.df.guide.w <- realPvals.df %>% subset(test.type=="per.guide.wilcoxon") # # make a plot for each topic, refer to TopFeatures list code # # Genes with the lowest empirical FDR # pdf(file=paste0(FIGDIRTOP, "top.perturbation.list_empirical.fdr.pdf"), width=4, height=6) # for ( topic in realPvals.df.guide.w$Topic %>% unique() ) { # toPlot <- realPvals.df.guide.w %>% subset(Topic == topic) %>% arrange(fdr) %>% slice(1:50) %>% mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) # p <- toPlot %>% ggplot(aes(x=reorder(Gene, -fdr), y=fdr) ) + geom_col() + theme_minimal() + # coord_flip() + xlab("Top 50 Gene (Perturbation)") + ylab("Empirical FDR") + ggtitle(paste(SAMPLE, " perturbations, ", topic)) # print(p) # } # dev.off() # } # ########################################################################## # ## most extreme log2FC (use omega) # ## # add ABC to gene.set.type.df for this particular plot # ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # for (test.type.here in c("per.cell.wilcoxon", "per.guide.wilcoxon")) { # all.test.subset <- all.test %>% subset(test.type==test.type.here) # realPvals.df.subset <- realPvals.df %>% subset(test.type==test.type.here) # pdf(file=paste0(FIGDIRTOP, "top.perturbation.list_log2FC_highlight.", test.type.here, ".pdf"), width=4, height=6) # for ( topic in colnames(gene.score) ) { # toPlot.all.test <- all.test.subset %>% subset(Topic==topic) # toPlot.fdr <- realPvals.df.subset %>% subset(Topic == topic) %>% select(Gene,fdr) # ## assemble toPlot # toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene") %>% # mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) # colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(fdr)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(log2FC)) %>% slice(1:25) # toPlot.bottom <- toPlot %>% arrange(log2FC) %>% slice(1:25) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(grepl("CAD",type), "red", # ifelse(type=="non-expressed", "grey", # ifelse(type=="other", "blue", "black")))) %>% # mutate(Gene = ifelse(type=="ABC", paste0("[ ", Gene, " ]"), Gene)) # colors <- toPlot.extreme$color[order(toPlot.extreme %>% arrange(desc(log2FC)) %>% pull(color))] # p <- toPlot.extreme %>% arrange(desc(log2FC)) %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=reorder(Gene, log2FC), y=log2FC, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("log2 Fold Change") + ggtitle(paste(SAMPLE, " perturbations, ", topic)) + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none", axis.text.y = element_text(colour = colors)) # print(p)#here # } # dev.off() # } # ########################################################################## # ## most extreme log2FC (use omega) short list # ## # add ABC to gene.set.type.df for this particular plot # ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # for (test.type.here in c("per.cell.wilcoxon", "per.guide.wilcoxon")) { # all.test.subset <- all.test %>% subset(test.type==test.type.here) # realPvals.df.subset <- realPvals.df %>% subset(test.type==test.type.here) # pdf(file=paste0(FIGDIRTOP, "top.perturbation.shortList_log2FC_highlight.", test.type.here, ".pdf"), width=4, height=4) # for ( topic in colnames(gene.score) ) { # toPlot.all.test <- all.test.subset %>% subset(Topic==topic) # toPlot.fdr <- realPvals.df.subset %>% subset(Topic == topic) %>% select(Gene,fdr) # # assemble toPlot # toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene") %>% # mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) # colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(fdr)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(log2FC)) %>% slice(1:11) # toPlot.bottom <- toPlot %>% arrange(log2FC) %>% slice(1:11) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(grepl("CAD",type), "red", # ifelse(type=="non-expressed", "grey", # ifelse(type=="other", "blue", "black")))) # colors <- toPlot.extreme$color[order(toPlot.extreme %>% arrange(desc(log2FC)) %>% pull(color))] # p <- toPlot.extreme %>% arrange(desc(log2FC)) %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=reorder(Gene, log2FC), y=log2FC, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("log2 Fold Change") + ggtitle(paste(SAMPLE, " perturbations, ", topic)) + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none", text=element_text(size=14), plot.title=element_text(size=12)) # print(p)#here # } # dev.off() # } # ########################################################################## # ## enrichment by GWAS classification # pdf(file=paste0(FIGDIRTOP,"_GWAS.class.enrichment.pdf")) # test.condition.names <- count.by.GWAS$test.type %>% unique() # for( i in 1:length(test.condition.names) ){ # test.condition <- test.condition.names[i] # toPlot <- count.by.GWAS %>% subset(test.type==test.condition) %>% select(GWAS.class.enrichment,GWAS.classification,gene.count.per.GWAS.category) %>% unique() # p <- toPlot %>% ggplot(aes(x=GWAS.classification, y=GWAS.class.enrichment)) + geom_bar(stat="identity",width=0.5,fill = "#38b4f7") + mytheme + # ggtitle(paste0("Fraction of genes with significant topics,", test.condition)) + xlab("GWAS Classification") + ylab("Fraction of Perturbations") # print(p) # p <- toPlot %>% ggplot(aes(x=GWAS.classification, y=gene.count.per.GWAS.category)) + geom_bar(stat="identity",width=0.5) + mytheme + # ggtitle(paste0("Number of genes with significant topics,", test.condition)) + xlab("GWAS Classification") + ylab("Number of Perturbations") # print(p) # } # dev.off() # ########################################################################## # ## GWAS gene ranking plot # pdf(file=paste0(FIGDIRTOP,"_GWAS.gene.rank.barplot.pdf"), width=8, height=4) # test.condition.names <- count.by.GWAS$test.type %>% unique() # GWAS.type.list <- c("CAD","IBD") # p.list <- vector("list", length(test.condition.names) * length(GWAS.type.list)) # for( i in 1:length(test.condition.names) ){ # for (GWAS.type.index in 1:length(GWAS.type.list)) { # test.condition <- test.condition.names[i] # GWAS.type <- GWAS.type.list[GWAS.type.index] # ## select the columns and rows we need for this plot # toPlot.tmp <- count.by.GWAS %>% subset(test.type==test.condition & GWAS.classification==GWAS.type) %>% select(TSS.v.SNP.ranking.in.GWAS.category, passed.filter.ranking.count, GWAS.classification,total.TSS.v.SNP.ranking.count.per.GWAS.classification) %>% unique() # toPlot <- toPlot.tmp %>% mutate(ranking.fraction = passed.filter.ranking.count / total.TSS.v.SNP.ranking.count.per.GWAS.classification) # p <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.fraction)) + geom_bar(stat="identity", width=0.5) + mytheme + # ggtitle(paste0("Fraction of ", GWAS.type, " genes per distance ranking \n with significant topics, ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Fraction of Perturbations") # print(p) # ## concatenate all genes ranked as 5+ # closest.ranking.boolean <- toPlot.tmp$TSS.v.SNP.ranking.in.GWAS.category < 5 # toPlot.closer.genes <- toPlot.tmp[which(closest.ranking.boolean),] %>% as.data.frame() # toPlot.farther.genes <- toPlot.tmp[which(!closest.ranking.boolean),] %>% ungroup() %>% select(-GWAS.classification, -TSS.v.SNP.ranking.in.GWAS.category) %>% apply(2,sum) %>% t() %>% as.data.frame() %>% mutate(TSS.v.SNP.ranking.in.GWAS.category = "5+", .before = passed.filter.ranking.count) %>% mutate(GWAS.classification = GWAS.type, .after = passed.filter.ranking.count) # toPlot <- rbind(toPlot.closer.genes, toPlot.farther.genes) %>% mutate(ranking.fraction = passed.filter.ranking.count / total.TSS.v.SNP.ranking.count.per.GWAS.classification) # p <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.fraction)) + geom_bar(stat="identity", width=0.5) + mytheme + # ggtitle(paste0("Fraction of ", GWAS.type, " genes per distance ranking \n with significant topics, ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Fraction of Perturbations") # print(p) # p.list[[(i-1)*2 + GWAS.type.index]] <- p # toPlot <- toPlot %>% select(-ranking.fraction) %>% melt(id.vars=c("TSS.v.SNP.ranking.in.GWAS.category","GWAS.classification"), variable.name="ranking.type", value.name="ranking.count") # toPlot$GWAS.classification <- factor(toPlot$GWAS.classification) # toPlot$ranking.type <- factor(toPlot$ranking.type) # p <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.count, fill=ranking.type)) + geom_bar(position="dodge",stat="identity",width=0.5) + mytheme + # ggtitle(paste0("Number of ", GWAS.type, " genes with significant topics, ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Number of Perturbations") + scale_fill_manual(values=wes_palette(n=length(unique(toPlot$ranking.type)), name="Darjeeling2"), name = "Type", labels = c("Significant Genes", "All Genes")) # print(p) # ## plot 4 # toPlot <- count.by.GWAS %>% subset(test.type==test.condition & GWAS.classification==GWAS.type) %>% select(TSS.v.SNP.ranking.in.GWAS.category, GWAS.classification, passed.filter.ranking.count, total.TSS.v.SNP.ranking.count.per.GWAS.classification) %>% unique() %>% melt(id.vars=c("TSS.v.SNP.ranking.in.GWAS.category","GWAS.classification"), variable.name="ranking.type", value.name="ranking.count") # p <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.count, fill=ranking.type)) + geom_bar(position="dodge",stat="identity",width=0.5) + mytheme + # ggtitle(paste0("Number of ", GWAS.type, " genes with significant topics, ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Number of Perturbations") + scale_fill_manual(values=wes_palette(n=length(unique(toPlot$ranking.type)), name="Darjeeling2"), name = "Type", labels = c("Significant Genes", "All Genes")) # print(p) # } # } # dev.off() # ########################################################################## # ## GWAS gene ranking by topic ## debug # pdf(file=paste0(FIGDIRTOP,"_GWAS.gene.rank.barplot.by.topic.pdf"), width=8, height=4) ##todo:210812: # test.condition.names <- count.by.GWAS.withTopic$test.type %>% unique() # for (GWAS.type.index in 1:length(GWAS.type.list)) { # GWAS.type <- GWAS.type.list[GWAS.type.index] # for( i in 1:length(test.condition.names) ){ # test.condition <- test.condition.names[i] # toPlot.allTopics <- count.by.GWAS.withTopic %>% subset(test.type==test.condition & GWAS.classification == GWAS.type) %>% # ungroup() %>% # select(Topic, # TSS.v.SNP.ranking.in.GWAS.category, # passed.filter.ranking.count, # GWAS.classification, # total.TSS.v.SNP.ranking.count.per.GWAS.classification # ) %>% unique() # toPlot.allTopics$GWAS.classification <- factor(toPlot.allTopics$GWAS.classification) # for( t in sort(unique(toPlot.allTopics$Topic %>% gsub("topic_","",.))) ) { # toPlot.tmp <- toPlot.allTopics %>% subset(Topic == paste0("topic_",t)) # toPlot <- toPlot.tmp %>% mutate(ranking.fraction = passed.filter.ranking.count / total.TSS.v.SNP.ranking.count.per.GWAS.classification) # p1 <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.fraction)) + geom_bar(stat="identity", width=0.5) + mytheme + # ggtitle(paste0("Fraction of ", GWAS.type, " significant genes per distance ranking \n in topic ", t, ", ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Fraction of Perturbations") # ## print(p1) # ## concatenate all genes ranked as 5+ # closest.ranking.boolean <- toPlot.tmp$TSS.v.SNP.ranking.in.GWAS.category < 5 # toPlot.closer.genes <- toPlot.tmp[which(closest.ranking.boolean),] %>% as.data.frame() %>% select(-Topic) # toPlot.farther.genes <- toPlot.tmp[which(!closest.ranking.boolean),] %>% ungroup() %>% select(-Topic, -GWAS.classification, -TSS.v.SNP.ranking.in.GWAS.category) %>% apply(2,sum) %>% t() %>% as.data.frame() %>% mutate(TSS.v.SNP.ranking.in.GWAS.category = "5+", .before = passed.filter.ranking.count) %>% mutate(GWAS.classification = GWAS.type, .after = passed.filter.ranking.count) # toPlot <- rbind(toPlot.closer.genes, toPlot.farther.genes) %>% mutate(ranking.fraction = passed.filter.ranking.count / total.TSS.v.SNP.ranking.count.per.GWAS.classification) # p2 <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.fraction)) + geom_bar(stat="identity", width=0.5) + mytheme + # ggtitle(paste0("Fraction of ", GWAS.type, " significant genes per distance ranking \n in topic ", t, ", ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Fraction of Perturbations") # ## print(p2) # toPlot <- toPlot %>% select(-ranking.fraction) %>% melt(id.vars=c("TSS.v.SNP.ranking.in.GWAS.category","GWAS.classification"), variable.name="ranking.type", value.name="ranking.count") # toPlot$GWAS.classification <- factor(toPlot$GWAS.classification) # toPlot$ranking.type <- factor(toPlot$ranking.type) # p3 <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.count, fill=ranking.type)) + geom_bar(position="dodge",stat="identity",width=0.5) + mytheme + # ggtitle(paste0("Number of ", GWAS.type, " genes DE in topic ", t, ", ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Number of Perturbations") + scale_fill_manual(values=wes_palette(n=length(unique(toPlot$ranking.type)), name="Darjeeling2"), name = "Type", labels = c("Significant Genes", "All Genes")) # ## print(p3) # ## plot 4 # toPlot <- count.by.GWAS.withTopic %>% ungroup() %>% subset(Topic == paste0("topic_",t) & test.type==test.condition & GWAS.classification==GWAS.type) %>% select(TSS.v.SNP.ranking.in.GWAS.category, GWAS.classification, passed.filter.ranking.count, total.TSS.v.SNP.ranking.count.per.GWAS.classification) %>% unique() %>% melt(id.vars=c("TSS.v.SNP.ranking.in.GWAS.category","GWAS.classification"), variable.name="ranking.type", value.name="ranking.count") # p4 <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking.in.GWAS.category, y=ranking.count, fill=ranking.type)) + geom_bar(position="dodge",stat="identity",width=0.5) + mytheme + # ggtitle(paste0("Number of ", GWAS.type, " significant genes in topic ", t, ", ", test.condition)) + xlab("Rank of Distance to the Closest SNP") + ylab("Number of Perturbations") + scale_fill_manual(values=wes_palette(n=length(unique(toPlot$ranking.type)), name="Darjeeling2"), name = "Type", labels = c("Significant Genes", "All Genes")) # ## print(p4) # p <- ggarrange(p4 + ggtitle("") + xlab(""), p1 + ggtitle("") + xlab(""), p3 + ggtitle("") + xlab(""), p2 + ggtitle("") + xlab(""), nrow=2, ncol=2, common.legend=T) # p <- annotate_figure(p, top = text_grob(paste0("Significant ", GWAS.type, " Genes in Topic ", t, ", ", test.condition), face="bold", size=14), # bottom = "Rank of Distance to the Closest SNP") # print(p) # ## toPlot <- toPlot %>% subset(Topic==paste0("topic_",t) & test.type = test.condition) # ## toPlot$GWAS.classification <- factor(toPlot$GWAS.classification) # ## p <- toPlot %>% ggplot(aes(x=TSS.v.SNP.ranking, y=ranking.count, fill=GWAS.classification)) + geom_bar(position="dodge",stat="identity",width=0.5) + mytheme + # ## ggtitle(paste0("Number of genes with significant topic ", t, ", ", test.condition)) + xlab("Distance Ranking to the Closest SNP") + ylab("Number of Perturbations") + scale_fill_manual(values=wes_palette(n=length(unique(toPlot$GWAS.classification)), name="Darjeeling2")) + # ## theme(legend.title = element_blank()) # ## print(p) # } # } # } # dev.off() ## ## GWAS ranking by distance in kb ## pdf(file=paste0(FIGDIRTOP,"_GWAS.gene.distance.by.bp.cdf.pdf"), width=8, height=4) ## test.condition.names <- count.by.GWAS$test.type %>% unique() ## GWAS.type.list <- c("CAD","IBD") ## toPlot.list <- vector("list", length(test.condition.names) * length(GWAS.type.list)) ## for( i in 1:length(test.condition.names) ){ ## for (GWAS.type.index in 1:length(GWAS.type.list)) { ## test.condition <- test.condition.names[i] ## GWAS.type <- GWAS.type.list[GWAS.type.index] ## toPlot <- count.by.GWAS %>% subset(test.type==test.condition & GWAS.classification==GWAS.type) %>% select(TSS.dist.to.SNP, GWAS.classification) ## toPlot$GWAS.classification <- factor(toPlot$GWAS.classification) ## p <- toPlot %>% ggplot(aes(x=TSS.dist.to.SNP)) + stat_ecdf() + mytheme + ## ggtitle(paste0(GWAS.type, " genes with significant topics, ", test.condition)) + xlab("Distance to the Closest SNP (in bp)") + ylab("Fraction of Significant Perturbed Genes") ## print(p) ## toPlot.list[[(i-1)*2 + GWAS.type.index]] <- toPlot %>% mutate(test.type=test.condition) ## } ## } ## dev.off() ## Need to double check test.type ## toPlot.all <- do.call(rbind, toPlot.list) %>% mutate(GWAS.classification = paste0(GWAS.classification, ".significant")) %>% rbind(ref.table %>% ungroup() %>% subset(GWAS.classification %in% GWAS.type.list) %>% select(TSS.dist.to.SNP, GWAS.classification) %>% mutate(test.type = "None")) ## for( i in 1:length(test.condition.names) ){ ## test.condition <- test.condition.names[i] ## toPlot <- toPlot.all %>% subset(test.type %in% c("None", test.condition)) ## p <- toPlot %>% ggplot(aes(x=TSS.dist.to.SNP, color = GWAS.classification)) + stat_ecdf() + mytheme + ## xlim(0,1000000) + ## ggtitle(paste0("Test by ", test.condition)) + xlab("Distance to the Closest SNP (in bp)") + ylab("Fraction of Perturbed Genes") ## print(p) ## heatmap of ptb correlation for factor values##here210810 ## heatmap of ptb correlation for factor values threshold with BH adjusted.p.value < 0.1 ## heatmap of factor correlation by expressed gene raw weights ## heatmap of factor correlation by expressed gene zscore # ########################################################################## # ## UMAP based on factor log2FC values (to see how perturbations cluster) # pdf(file=paste0(FIGDIRTOP,"perturbation.UMAP.based.on.factor.log2FC.pdf")) # DimPlot(s.gene.score, reduction = "umap", label=TRUE) %>% print() # dev.off() # ########################################################################## # ## motif enrichment plot # ep.names <- c("enhancer", "promoter") # for (ep.type in ep.names) { # pdf(file=paste0(FIGDIRTOP,"zscore.",ep.type,".motif.enrichment.pdf")) # all.volcano.plots(get(paste0("all.",ep.type,".fisher.df")), ep.type, ranking.type="z-score")##here210812 # dev.off() # pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.enrichment_motif.thr.10e-6.pdf")) # all.volcano.plots(get(paste0("all.",ep.type,".fisher.df.10en6")), ep.type, ranking.type="z-score")##here210812 # dev.off() # pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.enrichment.by.count.ttest.pdf")) # all.volcano.plots(get(paste0("all.",ep.type,".ttest.df")) %>% subset(top.gene.mean != 0 & !grepl("X.NA.",motif)), ep.type, ranking.type="z-score") # dev.off() # pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.enrichment.by.count.ttest_motif.thr.10e-6.pdf")) # all.volcano.plots(get(paste0("all.",ep.type,".ttest.df.10en6")) %>% subset(top.gene.mean != 0 & !grepl("X.NA.",motif)), ep.type, ranking.type="z-score") # dev.off() # pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.enrichment.by.count.ttest.labelPos.pdf"), width=6, height=6) # all.volcano.plots(get(paste0("all.",ep.type,".ttest.df")) %>% subset(top.gene.mean != 0 & !grepl("X.NA.",motif)), ep.type, ranking.type="z-score", label.type="pos") # dev.off() # pdf(file=paste0(FIGDIRTOP, "zscore.",ep.type,".motif.enrichment.by.count.ttest_motif.thr.10e-6.labelPos.pdf")) # all.volcano.plots(get(paste0("all.",ep.type,".ttest.df.10en6")) %>% subset(top.gene.mean != 0 & !grepl("X.NA.",motif)), ep.type, ranking.type="z-score", label.type="pos") # dev.off() # } ## ########################################################################## ## ## GSEA ## ## check if files already exist ## check.file <- paste0(FGSEADIR,"/fgsea_all_pathways_df_", c("raw.score", "z.score"), "_", SUBSCRIPT, ".RData") ## if(file.exists(check.file) %>% as.numeric() %>% sum() == length(check.file) & !opt$recompute ) { ## for (i in 1:length(check.file)) { ## print(paste0("Loading ", check.file[i])) ## load(check.file[i]) ## } ## } else {warning(paste0(check.file, " does not exist")) ## } ## bracket for checking files ## pdf(paste0(FGSEAFIG, "/msigdb.all.sig.pathways.pdf")) ## p <- toplot %>% ggplot(aes(x=topic, y=sig.pathway.count)) + geom_bar(stat="identity", fill ="#38b4f7") + mytheme + ## ggtitle(paste0(SAMPLE, " msigdb all, GSEA, number of significant pathway per topic")) + xlab("Topic") + ylab("Number of Significant Pathways") ## print(p) ## dev.off() # ########################################################################## # ## Pairwise Pearson correlation for perturbation's topic expression # d <- cor(gene.score %>% t(), method="pearson") # m <- as.matrix(d) # pdf(file=paste0(FIGDIRTOP,"cluster.ptb.by.Pearson.corr.pdf"), width=75, height=75) # plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, K=", k, ", ", SAMPLE, " topic clustering by Pearson correlation")) # dev.off() # png(file=paste0(FIGDIRTOP, "cluster.ptb.by.Pearson.corr.png"), width=1500, height=1500) # plotHeatmap(m, labCol=rownames(m), margins=c(12,12), title=paste0("cNMF, K=", k, ", ", SAMPLE, " topic clustering by Pearson correlation")) # dev.off() ###################################################################### ## Seurat UMAPs ## plot UMAPs pdf(paste0(FIGDIRTOP, "Factor.Expression.UMAP.pdf")) DimPlot(s, reduction = "umap", label=TRUE) %>% print() meta.data.names <- colnames(s[[]]) plot.features <- paste0("K",k,"_",colnames(omega)) for (feature.name in plot.features) { feature.vec <- s@meta.data %>% select(all_of(feature.name)) if(feature.vec[1,1] %>% is.numeric()) { # numeric FeaturePlot(s, reduction = "umap", features=feature.name) %>% print() } else { # discrete values / categories Idents(s) <- s@meta.data %>% select(feature.name) DimPlot(s, reduction = "umap", label=TRUE) %>% print() } } dev.off() ## ## 10x lane sample label UMAP ## checking for batch effects ## CBC.sample <- colnames(s) %>% gsub("^.*-", "", .) %>% gsub("scRNAseq_2kG_","",.) ## CBC.sample.short <- CBC.sample %>% gsub("_.*$","",.) ## s <- AddMetaData(s, CBC.sample, col.name="sample.label.10X") ## s <- AddMetaData(s, CBC.sample.short, col.name="sample.label.short") ## pdf(file=paste0(FIGDIRTOP, "10X.lane.sample.label.UMAP.pdf"), width=10, height=6) ## Idents(s) <- s$sample.label.10X ## DimPlot(s, reduction = "umap", label=F, group.by="sample.label.10X") %>% print() ## Idents(s) <- s$sample.label.short ## DimPlot(s, reduction = "umap", label=F) %>% print() ## DimPlot(s, reduction = "umap", label=F, split.by="sample.label.short") %>% print() ## dev.off() # ########################################################################## # ##### SUMMARY PLOTS ##### # ########################################################################## # ## functions # get.average.ptb.gene.expression.based.on.ctrl <- function(ptb, ptb.df, ctrl.df, mode="per.guide") { # # ptb: column that has gene expression or topic weight, (e.g. "SWAP70" or "topic_4") # if (mode=="per.guide") { # first average by guide then average by perturbation # e.name <- ptb.df %>% rownames() %>% gsub("_multiTarget|-TSS2","",.) %>% strsplit(., split=":") %>% sapply("[[", 1) %>% unique() # ptb.gene.column.index <- which(grepl(paste0("^",ptb,"$"), colnames(ptb.df))) # expression.of.ptb <- ptb.df[,ptb.gene.column.index] %>% as.data.frame() %>% `rownames<-`(rownames(ptb.df)) %>% `colnames<-`(ptb) %>% mutate(long.CBC=rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), remove=F, sep=":") %>% group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene=e.name) # ## expression.of.ptb <- ptb.df %>% select(c(all_of(ptb), colnames(guideCounts))) %>% # ## group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene=e.name) # expression.ctrl.ptb <- ctrl.df[,ptb.gene.column.index] %>% as.data.frame() %>% `rownames<-`(rownames(ctrl.df)) %>% `colnames<-`(ptb) %>% mutate(long.CBC=rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), remove=F, sep=":") %>% group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene="Control")# %>% select(c(all_of(ptb), colnames(guideCounts))) %>% group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene="Control") # toCalculate <- rbind(expression.of.ptb, expression.ctrl.ptb) # toCalculate$Gene.expression <- toCalculate$Gene.expression / (toCalculate %>% subset(Gene=="Control") %>% pull(Gene.expression) %>% mean()) * 100 # toCalculate <- toCalculate %>% group_by(Gene) %>% mutate(Gene = paste0(gsub("topic_","Topic ", Gene), "\n(n=", n(), ")")) # toPlot <- toCalculate %>% group_by(Gene) %>% summarise(mean.expression=mean(Gene.expression), error.bar=1.96 * sd(Gene.expression)/sqrt(n()), count=n()) # 1.96 * sd(vals)/sqrt(length(vals)) # } else { # directly average by cell # e.name <- ptb.df %>% rownames() %>% gsub("_multiTarget|-TSS2","",.) %>% strsplit(., split=":") %>% sapply("[[", 1) %>% unique() # # e.name <- ptb.df$Gene %>% unique() # ptb.gene.column.index <- which(grepl(paste0("^",ptb,"$"), colnames(ptb.df))) # expression.of.ptb <- ptb.df[,ptb.gene.column.index] %>% as.data.frame() %>% `rownames<-`(rownames(ptb.df)) %>% `colnames<-`(ptb) %>% mutate(long.CBC=rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), remove=F, sep=":") %>% mutate(Gene=e.name) # expression.ctrl.ptb <- ctrl.df[,ptb.gene.column.index] %>% as.data.frame() %>% `rownames<-`(rownames(ctrl.df)) %>% `colnames<-`(ptb) %>% mutate(long.CBC=rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), remove=F, sep=":") %>% mutate(Gene="Control") # ## expression.of.ptb <- ptb.df %>% select(c(all_of(ptb), colnames(guideCounts))) %>% mutate(Gene=e.name) # ## expression.ctrl.ptb <- ctrl.df %>% select(c(all_of(ptb), colnames(guideCounts))) %>% mutate(Gene="Control") # toCalculate <- rbind(expression.of.ptb, expression.ctrl.ptb) %>% select(all_of(ptb),Guide,Gene) %>% mutate(Gene.expression = get(ptb)) # toCalculate$Gene.expression <- toCalculate$Gene.expression / (toCalculate %>% subset(Gene=="Control") %>% pull(Gene.expression) %>% mean()) * 100 # toCalculate <- toCalculate %>% group_by(Gene) %>% mutate(Gene = paste0(gsub("topic_","Topic ", Gene), "\n(n=", n(), ")")) # toPlot <- toCalculate %>% group_by(Gene) %>% summarise(mean.expression=mean(Gene.expression), error.bar=1.96 * sd(Gene.expression)/sqrt(n()), count=n()) # # 1.96 * sd(vals)/sqrt(length(vals)) # } # return(list(toCalculate=toCalculate, toPlot=toPlot)) # } # lm_eqn <- function(df){ # m <- lm(y ~ x, df); # eq <- substitute(italic(y) == a + b %.% italic(x)*","~~italic(r)^2~"="~r2, # list(a = format(unname(coef(m)[1]), digits = 2), # b = format(unname(coef(m)[2]), digits = 2), # r2 = format(summary(m)$r.squared, digits = 3))) # as.character(as.expression(eq)); # } # lm_eqn_manual <- function(a, b){ # eq <- substitute(italic(y) == p + q %.% italic(x), # list(p = format(unname(a), digits = 2), # q = format(unname(b), digits = 2))) # as.character(as.expression(eq)) %>% return() # } # normalize.by.ctrl.avg <- function(ptb, ptb.df, ctrl.df, mode="per.guide"){ # e.name <- ptb.df$Gene %>% gsub("_multiTarget|-TSS2","",.) %>% strsplit(., split=":") %>% sapply("[[", 1) %>% unique() # ptb.col.index <- which(grepl(paste0("^",ptb,"$"), colnames(ctrl.df))) # ctrl.df.tmp <- ctrl.df[,ptb.col.index] %>% as.data.frame() %>% `rownames<-`(rownames(ctrl.df)) %>% `colnames<-`(ptb) # ctrl.df.tmp <- ctrl.df.tmp %>% mutate(long.CBC=rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), sep=":", remove=F) # if (mode=="per.guide") { # first average by guide then average by perturbation # expression.ctrl.ptb.tmp <- ctrl.df.tmp %>% group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene="Control") # ## expression.ctrl.ptb.tmp <- ctrl.df %>% select(c(all_of(ptb), colnames(guideCounts))) %>% # ## group_by(Guide) %>% summarise(Gene.expression=mean(get(ptb))) %>% mutate(Gene="Control") # # ptb.df with mean as % control and errorbar # expression.of.ptb <- ptb.df %>% # mutate(Gene.expression = get(ptb) / (expression.ctrl.ptb.tmp$Gene.expression %>% mean()) * 100) %>% # group_by(Guide) %>% # summarise(Gene.error.bar = 1.96 * sd(Gene.expression)/sqrt(n()), # Gene.expression=mean(Gene.expression)) %>% mutate(Gene=e.name) # summarise or mutate? # need to keep CBC # # control with error bar # expression.ctrl.ptb <- ctrl.df.tmp %>% # mutate(Gene.expression = get(ptb) / (expression.ctrl.ptb.tmp$Gene.expression %>% mean()) * 100) %>% # group_by(Guide) %>% # summarise(Gene.error.bar = 1.96 * sd(Gene.expression)/sqrt(n()), # Gene.expression=mean(Gene.expression)) %>% mutate(Gene="Control") # toCalculate <- rbind(expression.of.ptb, expression.ctrl.ptb) # } else { # directly average by cell # expression.ctrl.ptb <- ctrl.df.tmp %>% mutate(Gene="Control") # expression.of.ptb <- ptb.df %>% mutate(Gene=e.name) # toCalculate <- rbind(expression.of.ptb %>% select(all_of(ptb),Guide,Gene,long.CBC), expression.ctrl.ptb %>% select(all_of(ptb),Guide,Gene,long.CBC)) %>% mutate(Gene.expression = get(ptb)) # } # toCalculate <- toCalculate %>% group_by(Gene) %>% mutate(Gene.count = paste0(gsub("topic_","Topic ", Gene), " (n=", n(), ")")) # return(toCalculate) # } # giant.summary.plot <- function(ptb, ptb.expressed.name, expressed.gene, mode.selection, enhancer=F) { # # toPlot.per.guide.DE.test.wilcox <- per.guide.DE.test.wilcox %>% subset(Topic==ptb) # # if (enhancer) { # # array <- ann.X.full.filtered$Gene[grep("E_at_", ann.X.full.filtered$Gene)] %>% unique() # # e.name <- array[grep(ptb,array)] # # } else { # # e.name <- ptb # # } # # tmp <- ptb # "E_at_" "-no" # # ptb <- expressed.gene # "GOSR2" # # e.name <- tmp # all.test.guide.w <- all.test %>% subset(test.type==paste0(mode.selection,".wilcoxon")) # realPvals.df.guide.w <- realPvals.df %>% subset(test.type==paste0(mode.selection,"per.guide.wilcoxon")) # #TODO: add rep1rep2 to ctrl.X %>% subset() # # for SEP=T, subset ctrl.X to the right sample's control # if(SEP) { # label.here <- strsplit(ptb, split="-") %>% unlist() %>% nth(2) %>% paste0("-",.) # ctrl.X.here <- ctrl.X %>% subset(grepl(label.here,Gene)) # ctrl.ann.omega.here <- ctrl.ann.omega %>% subset(grepl(label.here,Gene)) # } else { # ctrl.X.here <- ctrl.X # ctrl.ann.omega.here <- ctrl.ann.omega # } # subset.index <- which(X.gene.names == ptb) # ## ann.X.full.filtered.df <- # toPlot.list <- get.average.ptb.gene.expression.based.on.ctrl(expressed.gene, ann.X.full.filtered[subset.index,], ctrl.X.here, mode=mode.selection) # toPlot1 <- toPlot.list[["toPlot"]] # order.toPlot.Gene <- function(df) { # df$Gene <- factor(x=df$Gene, levels=df$Gene[c(which(grepl("^Control", df$Gene)) %>% min(), which(!grepl("^Control", df$Gene))%>% min())]) # return(df) # } # order.toPlot <- function(df, column) { # array <- df %>% pull(all_of(column)) # df <- df %>% mutate(!!column:=factor(x=array, levels=array[c(which(grepl("^Control", array)) %>% min(), which(!grepl("^Control", array))%>% min())])) # return(df) # } # toPlot1 <- order.toPlot(toPlot1, column="Gene") # KD.per.guide <- toPlot.list[["toCalculate"]] # colnames(KD.per.guide)[colnames(KD.per.guide)=="Gene.expression"] <- "mean.expression" # # plot 5 data # subset.index <- which(X.gene.names == ptb) # ## ptb.gene.col.index <- which(grepl(ptb.expressed.name, colnames(ann.X.full.filtered))) # ptb.gene.col.index <- which(colnames(ann.X.full.filtered) == ptb.expressed.name) # ann.X.ptb <- ann.X.full.filtered[subset.index,ptb.gene.col.index] %>% as.data.frame() %>% `colnames<-`(ptb.expressed.name) %>% mutate(long.CBC = rownames(.)) %>% separate(., col=long.CBC, into=c("Gene", "Guide", "CBC"), sep=":", remove=F)# select ptb expression column # ## ann.X.ptb <- ann.X.full.filtered %>% subset(Gene==ptb) %>% select(colnames(guideCounts), all_of(expressed.gene)) # select ptb expression column # normalized.ann.X.ptb.by.guide <- normalize.by.ctrl.avg(expressed.gene, ann.X.ptb, ctrl.X.here, mode.selection) ##210823:debug expressed.gene or ptb # normalized.ann.X.ptb.by.guide[is.na(normalized.ann.X.ptb.by.guide)] <- 0 # # topic # ptb.omega.filtered <- ann.omega.filtered %>% subset(Gene == ptb) # if (dim(ptb.omega.filtered)[1] > 0){ # if the guide didn't cause a fitness effect # if(!dir.exists(paste0(FIGDIRSAMPLE,"/gene.by.topic/"))) dir.create(paste0(FIGDIRSAMPLE,"/gene.by.topic/"), recursive=T) # if(!enhancer) pdf(file=paste0(FIGDIRSAMPLE,"/gene.by.topic/",ptb, ".", mode.selection, ".dt_", DENSITY.THRESHOLD, ".pdf"), width=16, height=8) # # Do gRNAs at the promoter reduce gene expression? # # old plot # p1 <- toPlot1 %>% ggplot(aes(x=Gene,y=mean.expression)) + geom_bar(stat='identity', width=0.5, fill="#38b4f7") + # geom_errorbar(data = toPlot1, aes(x=Gene, ymin=mean.expression-error.bar, ymax=mean.expression+error.bar), width=.15) + # # ylim(min(KD.per.guide$mean.expression - 25, 0), max(KD.per.guide$mean.expression + 50, 150)) + # ylab(paste0(expressed.gene, " RNA Expression\n(% vs control)")) + xlab("Guides") + mytheme + # # scale_y_continuous(limits = c(0, max(KD.per.guide$mean.expression + 50, 150)), breaks = round(seq(0, max(KD.per.guide$mean.expression + 50, 150), 20),20) ) + # geom_hline(yintercept = 100, color="grey", linetype="dashed", size=0.5) + # geom_jitter(data=KD.per.guide, size=0.25, width=0.15, color="red") + # ggdist::stat_halfeye( # data=KD.per.guide, # ## custom bandwidth # adjust = .5, # ## adjust height # width = .3, # ## move geom to the right # justification = -1, # ## remove slab interval # .width = 0, # point_colour = NA, # ## change violin color # fill = "#a0dafa" # ) # ## ## referenced https://www.cedricscherer.com/2021/06/06/visualizing-distributions-with-raincloud-plots-with-ggplot2/ # ## p1 <- KD.per.guide %>% ggplot(aes(x=Gene, y=mean.expression)) + # ## ## add half-violin from {ggdist} package # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = .4, # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## mytheme + # ## geom_hline( # ## yintercept = 100, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) + # ## ylab(paste0(expressed.gene, " RNA Expression\n(% vs control)")) + # ## xlab("Perturbation") # for (t in 1:dim(omega)[2]) { #:dim(omega)[2] # topic <- paste0("topic_", t) # toPlot.all.test <- all.test.guide.w %>% subset(Topic==topic) # toPlot.fdr <- realPvals.df.guide.w %>% subset(Topic == topic) %>% select(Gene,fdr) # # Which topics does the gene regulate? # toPlot.list <- get.average.ptb.gene.expression.based.on.ctrl(topic, ptb.omega.filtered %>% `rownames<-`(ptb.omega.filtered$long.CBC), ctrl.ann.omega.here %>% `rownames<-`(ctrl.ann.omega.here$long.CBC), mode=mode.selection) # toPlot2 <- toPlot.list[["toPlot"]] # # levels(toPlot2$Gene)[grep(topic,levels(toPlot2$Gene))] <- gsub("_", " ", toPlot2$Gene[grep(topic,levels(toPlot2$Gene))]) # change topic_x to "topic x" # toPlot2 <- order.toPlot.Gene(toPlot2) # toCalculate <- toPlot.list[["toCalculate"]] # colnames(toCalculate)[colnames(toCalculate)=="Gene.expression"] <- "mean.expression" # p2.y.step <- (max(toCalculate$mean.expression/5) %/% 20) * 20 # p2.y.step <- ifelse(p2.y.step==0, 20, p2.y.step) # p2.y.step <- ifelse(p2.y.step > 100, 100, p2.y.step) # ## ## old plot # ## p2 <- toPlot2 %>% ggplot(aes(x=Gene,y=mean.expression)) + geom_bar(stat='identity', width=0.5, fill="#38b4f7") + # ## geom_errorbar(aes(ymin=mean.expression-error.bar, ymax=mean.expression+error.bar), width=.15) + # ## # ylim(0, max(toPlot2$mean.expression + toPlot2$error.bar + 50, 150)) + # ## ylab(paste0(gsub("topic_","Topic ", topic), " Expression\n(% vs control)")) + xlab("Guides") + mytheme + # ## # scale_y_continuous(limits = c(0, max(toCalculate$mean.expression + 50, 150)), breaks = seq(0, max(toCalculate$mean.expression + 50, 150), p2.y.step)) + # ## geom_hline(yintercept = 100, color="grey", linetype="dashed", size=0.5) + # ## geom_jitter(data=toCalculate, size=0.25, width=0.15, color="red") # p2 <- toPlot2 %>% ggplot(aes(x=Gene,y=mean.expression)) + geom_bar(stat='identity', width=0.5, fill="#38b4f7") + # ## geom_jitter(data=toCalculate, aes(x=Gene, y=mean.expression), size=0.25, width=0.15, color="red") + # geom_errorbar(aes(ymin=mean.expression-error.bar, ymax=mean.expression+error.bar), width=.15) + # # ylim(0, max(toPlot2$mean.expression + toPlot2$error.bar + 50, 150)) + # ylab(paste0(gsub("topic_","Topic ", topic), " Expression\n(% vs control)")) + xlab("Guides") + mytheme + # # scale_y_continuous(limits = c(0, max(toCalculate$mean.expression + 50, 150)), breaks = seq(0, max(toCalculate$mean.expression + 50, 150), p2.y.step)) + # geom_hline(yintercept = 100, color="grey", linetype="dashed", size=0.5) + # geom_jitter(data=toCalculate, size=0.25, width=0.15, color="red") + # ggdist::stat_halfeye( # data=toCalculate, # ## custom bandwidth # adjust = .5, # ## adjust height # width = .3, # ## move geom to the right # justification = -1, # ## remove slab interval # .width = 0, # point_colour = NA, # ## change violin color # fill = "#a0dafa" # ) # ## ## referenced https://www.cedricscherer.com/2021/06/06/visualizing-distributions-with-raincloud-plots-with-ggplot2/ # ## p2 <- toCalculate %>% ggplot(aes(x=Gene, y = mean.expression)) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = .4, # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## mytheme + # ## geom_hline( # ## yintercept = 100, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) + # ## ylab(paste0(gsub("topic_","Topic ", topic), " Expression\n(% vs control)")) + # ## xlab("Perturbation") # # ## alternative to p2, but basically the same thing # # ## stat_summary removes some values before averaging and causes control to be not at 100% # # p3 <- toPlot.list[["toCalculate"]] %>% ggplot(aes(x=Gene,y=Gene.expression)) + # # geom_bar(stat = "summary", fun = "mean", width=0.5, fill="#38b4f7") + # # stat_summary(fun.data = mean_cl_normal, # # geom = "errorbar", width=0.15) + # # # ylim(0, max(toPlot2$mean.expression + toPlot2$error.bar + 50, 150)) + # # ylab(paste0(gsub("_"," ", topic), " Expression\n(% vs control)")) + xlab("") + mytheme + # # scale_y_continuous(limits = c(0, max(toPlot2$mean.expression + toPlot2$error.bar + 50, 150)), breaks = seq(0, max(toPlot2$mean.expression + toPlot2$error.bar + 50, 150), p2.y.step)) + # # geom_hline(yintercept = 100, color="grey", linetype="dashed") + # # geom_jitter(size=0.25, width=0.15, color="red") # ## #TODO: fix KL specificity score # ## # What cellular program does this topic represent? # ## toPlot <- data.frame(Gene=topFeatures %>% subset(topic == t) %>% pull(genes), # ## Score=topFeatures %>% subset(topic == t) %>% pull(scores)) %>% # ## merge(., gene.def.pathways, by="Gene", all.x=T) # ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # ## p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100, fill=Pathway) ) + geom_col(width=0.5) + theme_minimal() # ## p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("KL score (gene specific to this topic)") + # ## mytheme + theme(legend.position="bottom", legend.direction="vertical") # # raw weight version # ## hand annotated files by Helen # toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), # Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) %>% # merge(., gene.def.pathways, by="Gene", all.x=T) %>% arrange(desc(Score)) %>% slice(1:10) # ## 210804 use Gavin's new summary annotations # ## toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), # ## Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) %>% # ## merge(., summaries %>% select(Symbol, top_class), by.x="Gene", by.y="Symbol", all.x=T) %>% unique %>% `colnames<-`(c("Gene", "Score", "Pathway")) %>% arrange(desc(Score)) %>% slice(1:10) # toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # toPlot$Pathway[toPlot$Pathway == "unclassified"] <- "Other/Unclassified" # p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100, fill=Pathway) ) + geom_col(width=0.5) + theme_minimal() # p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("Specificity Score (z-score)") + # mytheme + theme(legend.position="bottom", legend.direction="vertical") # # plot 4 # # add ABC to gene.set.type.df for this particular plot # ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # # assemble toPlot # toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% # merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # ## merge(.,gene.set.type.df,by="Gene") %>% # mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) %>% # arrange(log2FC) %>% # mutate(x = seq(nrow(.))) # label <- toPlot %>% subset(x <= 3 | x > (nrow(toPlot)-3) | Gene == ptb | adjusted.p.value < fdr.thr) # ## toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # ## merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # ## merge(.,gene.set.type.df,by="Gene") %>% # ## mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) # ## colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # ## toPlot <- toPlot %>% mutate(significant=ifelse((fdr >= fdr.thr | is.na(fdr)), "", "*")) %>% # ## arrange(log2FC) %>% # ## mutate(x = seq(nrow(.))) # ## label <- toPlot %>% subset(fdr < fdr.thr | x <= 3 | x > (nrow(toPlot)-3) | Gene == ptb | adjusted.p.value < fdr.thr) # mytheme <- theme_classic() + theme(axis.text = element_text(size = 13), axis.title = element_text(size = 15), plot.title = element_text(hjust = 0.5, face = "bold")) # p5 <- toPlot %>% ggplot(aes(x=reorder(Gene, log2FC), y=log2FC, color=significant)) + geom_point(size=0.75) + mytheme + # theme(axis.ticks.x=element_blank(), axis.text.x=element_blank()) + scale_color_manual(values = c("#E0E0E0", "#38b4f7")) + # xlab("Perturbed Genes") + ylab(paste0("Topic ", t, " Expression log2 Fold Change")) + # geom_text_repel(data=label, box.padding = 0.5, # aes(label=ann.Gene), size=5, # color="black") + # theme(legend.position = "none") # legend at the bottom? # # plot 5: topic expression vs KD efficacy # # data # ptb.omega.t <- ptb.omega.filtered %>% select(all_of(c("Gene","Gene.full.name","long.CBC","CBC","Guide")), all_of(topic)) # normalized.ptb.omega.t.by.guide <- normalize.by.ctrl.avg(topic, ptb.omega.t, ctrl.ann.omega.here, mode=mode.selection) %>% # select(-Gene, -Gene.count) %>% # remove redundant columns before merging # `colnames<-`(gsub("Gene", "Topic", colnames(.))) # if(mode.selection=="per.cell"){ # merged.ptb.normalized.X.omega.t.by.guide <- merge(normalized.ann.X.ptb.by.guide %>% select(-Guide), normalized.ptb.omega.t.by.guide %>% select(-Guide), by="long.CBC") # } else { # merged.ptb.normalized.X.omega.t.by.guide <- merge(normalized.ann.X.ptb.by.guide, normalized.ptb.omega.t.by.guide, by="Guide") # GOSR2-plus should have 10 guides, but only 8 left # ##debug: rbind(deparse.level, ...): numbers of columns of arguments do not match # } # toPlot <- merged.ptb.normalized.X.omega.t.by.guide %>% order.toPlot(., column="Gene.count") # toPlot.ptb <- toPlot %>% subset(Gene == ptb) # for linear regression # fit <- tryCatch(york(toPlot.ptb %>% select(Gene.expression, Gene.error.bar, Topic.expression, Topic.error.bar)), # error=function(cond){ # return(data.frame(a = 0, b = 0)) # }) # p6a <- toPlot %>% ggplot(aes(x=Gene.expression, y=Topic.expression, color = Gene.count)) + # geom_point(size=1) + # ylab(paste0("Topic ", t, " Expression\n(% vs control)")) + xlab(paste0(expressed.gene, " RNA Expression\n(% vs control)")) + mytheme + # scale_color_manual(values=c("grey", "red"), name="Guide") + theme(legend.position="bottom") + # geom_smooth(data = toPlot.ptb, method="lm", se=F, fullrange=T, size=0.5) + # annotate("text", size = 5, x = min(toPlot$Gene.expression ) + 30, y = min(toPlot$Topic.expression - 15), hjust=0.2, # label = lm_eqn(data.frame(x=toPlot.ptb$Gene.expression, # y=toPlot.ptb$Topic.expression)), parse = TRUE) # , parse = TRUE # p6 <- toPlot %>% ggplot(aes(x=Gene.expression, y=Topic.expression, color = Gene.count)) + # geom_point(size=0.3, color = "gray") + # geom_point(data=toPlot.ptb, size=0.8, color = "red") + ## put perturbation red datapoints on top # ylab(paste0("Topic ", t, " Expression\n(% vs control)")) + xlab(paste0(expressed.gene, " RNA Expression\n(% vs control)")) + mytheme + # scale_color_manual(values=c("grey", "red"), name="Guide") + theme(legend.position="bottom") + # geom_abline(intercept = fit$a[1], slope = fit$b[1], col="red", size=0.5) # if(mode.selection=="per.cell"){ # p6 <- p6 + annotate("text", size = 5, hjust=0.2, x=0, y=0-(max(toPlot$Topic.expression)-min(toPlot$Topic.expression))/20, # label = lm_eqn_manual(fit$a[1], fit$b[1]), parse = TRUE) # } else { # p6 <- p6 + annotate("text", size = 5, x = min(toPlot$Gene.expression ) + 30, y = min(toPlot$Topic.expression - 15), hjust=0.2, # label = lm_eqn_manual(fit$a[1], fit$b[1]), parse = TRUE) # } # # set xlim and ylim to fit annotation! # ## TF motif enrichment volcano plot # toplot <- all.promoter.ttest.df %>% subset(topic==paste0("topic_",t) & top.gene.mean != 0) # volcano.plot <- function(toplot, EP.string, label.type="") { # if( label.type == "pos") { # label <- toplot %>% subset(-log10(p.adjust) > 1 & enrichment.log2fc > 0) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) # } else { # label <- toplot %>% subset(-log10(p.adjust) > 1) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) # } # t <- gsub("topic_", "", toplot$topic[1]) # p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(p.adjust))) + geom_point(size=0.5) + mytheme + # ggtitle(paste0(EP.string, " Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(adjusted p-value)") + # geom_text_repel(data=label, box.padding = 0.5, # aes(label=motif.toshow), size=5, # max.overlaps=25, # color="black") # return(p) # } # p.promoter.motif <- volcano.plot(toplot, "Promoter", label.type="pos") # toplot <- all.enhancer.ttest.df.10en6 %>% subset(topic==paste0("topic_",t) & top.gene.mean != 0) # p.enhancer.motif <- volcano.plot(toplot, "Enhancer", label.type="pos") # ## edgeR p-value vs ptb effect RNA expression for top 100 specific gene in the topic # ##use.edgeR.results # edgeR.expr.gene.names <- rownames(log2fc.edgeR) %>% strsplit(split=":") %>% sapply("[[",1) # ptb.colindex <- which(grepl(ptb, colnames(log2fc.edgeR))) # theta.zscore.t <- theta.zscore[,t] %>% as.data.frame %>% `colnames<-`("topic.zscore.weight") %>% mutate(genes=rownames(.)) %>% arrange(desc(topic.zscore.weight)) %>% mutate(gene.rank = 1:n(), top100=ifelse(gene.rank <= 100, paste0("Top 100 in Topic ", t), paste0("Not in Top 100"))) # ptb.pval.log2fc <- merge(p.value.edgeR[,c(1,ptb.colindex)] %>% `colnames<-`(c("genes", "pval")) %>% mutate(genes = strsplit(genes, split=":") %>% sapply("[[",1)), log2fc.edgeR[,c(1,ptb.colindex)] %>% `colnames<-`(c("genes", "log2fc")) %>% mutate(genes = strsplit(genes, split=":") %>% sapply("[[",1)), by="genes") %>% merge(theta.zscore.t, by="genes") %>% mutate(nlog10pval = -log10(pval)) # ptb.pval.log2fc$nlog10pval[ptb.pval.log2fc$nlog10pval > 15] <- 15 # ptb.pval.log2fc$top100 <- factor(ptb.pval.log2fc$top100) %>% ordered(levels=c(paste0("Top 100 in Topic ", t), "Not in Top 100")) # label <- ptb.pval.log2fc %>% subset(top100!="Not in Top 100") # ptb.10X <- ptb.10X.name.conversion$`Name used by CellRanger`[which(grepl(ptb, ptb.10X.name.conversion$Symbol))] # label.self <- ptb.pval.log2fc %>% subset(genes == ptb.10X) # p.density <- ptb.pval.log2fc %>% ggplot(aes(x=log2fc, fill=top100)) + geom_density(alpha=0.4) + mytheme + xlab("RNA Expression\n(log2FC vs Control") + scale_fill_manual(values=c("red", "gray"), name = "") + theme(legend.position = "none") # + guides(fill=guide_legend(nrow=2, byrow=T)) # p.edgeR <- ggplot(ptb.pval.log2fc, aes(x=log2fc, y=nlog10pval)) + geom_point(size=0.1, color = "gray") + # geom_point(data = label, size=0.1, color = "red") + # mytheme + # geom_text_repel(data=label.self, box.padding = 0.5, # max.overlaps=30, # aes(label=genes), size=4, # color="blue") + # geom_text_repel(data=label, box.padding = 0.5, # max.overlaps=30, # aes(label=genes), size=4, # color="black") + # scale_color_manual(values=c("gray", "red")) + # geom_vline(xintercept=0, col = "#38b4f7", lty=3) + # theme(legend.position="bottom") + # ggtitle(paste0("Topic ", t, " Perturbation ", ptb)) + # xlab("Average RNA Expression (log2FC vs control)") + ylab("p-value (-log10)") # + # ## inset_element(p.density, left = 0.1, bottom = 0.75, right=0.9, top = 0.95) # p.edgeR.combined <- cowplot::plot_grid(p.density, p.edgeR, aign="v", ncol=1, rel_heights=c(0.15, 0.85)) # ## ## old # ## p.top.left <- ggarrange(p1, p2, p6, nrow=1, widths=c(1.5,1.5,2)) # ## ## p.bottom.left <- ggarrange(p6a, p6, nrow=1) # ## p.bottom.left <- ggarrange(p.promoter.motif, p.enhancer.motif, nrow=1) # ## p.left <- ggarrange(p.top.left, p.bottom.left, nrow=2) # ## p <- ggarrange(p.left, p4, p5, ncol=3, nrow=1, widths=c(2,1,1)) # ## # p <- ggarrange(p1, p2, p4, p5, p6, ncol=4, nrow=1, widths=c(1,1,1.5,1.5)) # p.top.left <- ggarrange(p1, p2, p6, nrow=1, widths=c(1.5,1.5,2)) # p.bottom.left <- ggarrange(p.promoter.motif, p.enhancer.motif, nrow=1) # p.left <- ggarrange(p.top.left, p.bottom.left, nrow=2) # p.mid <- ggarrange(p4, p.edgeR, nrow=2, heights=c(1.5,1)) # p <- ggarrange(p.left, p.mid, p5, ncol=3, nrow=1, widths=c(2,1,1)) # print(p) # } # if(!enhancer) dev.off() # } # } ########################################################################## ## slide 26 for all perturbation x topic pairs # function for plot 5 ## # plot 3 data ## toSave.features <- read.delim(paste0(OUTDIRSAMPLE, "/topic.KL.score_K", k,ifelse(SEP, ".sep", ""), ".txt"),header=T, stringsAsFactors=F) ### commented out because it's not necessary in this pipeline ## topFeatures <- toSave.features %>% group_by(topic) %>% arrange(desc(scores)) %>% slice(1:10) # plot 4 data ## ## FGSEA results ## ## load data ## type <- "z.score" ## fgsea.df <- read.delim(file=paste0(FGSEADIR, "/fgsea_", type, "_", SUBSCRIPT, ".txt"), header=T, stringsAsFactors=F) ##HERE ## fgsea.df.GO <- fgsea.df %>% subset(database == "msigdb.c3") ## fgsea.df.all <- fgsea.df %>% subset(database == "msigdb.all") # # ptb.array <- c("GOSR2", "PRKCE", "PHACTR1", "EIF2B2") # ptb.array <- c("RHOA", "PECAM1", "RAP1A", "KLF4", "MEF2C", "EGR1", "CDC42EP2", "CDH5", "KIAA1429","GOSR2","TP53","MAT2A","SKI","EDN1","SMAD3","PHACTR1","EDN1","GGCX","CDKN1A","EGFL7","ELOF1","GPANK1","YLPM1","MESDC1","ITGA5","SKIV2L","LST1","R3HCC1L","UPF2","MEAF6", "CCM2", "KRIT1", "ITGB1BP","HEG1") # ## ptb.array <- c("TP53", "SWAP70", "GOSR2", "PRKCE", "PHACTR1", "EIF2B2", "PPIF", "DMRTA1", "ADAMTS7", "VEZT", "MEAF6", "CDH5") # # ptb.array <- guideCounts$Gene %>% unique() # # ptb.array <- enhancer.set # ## ptb.array <- append(ptb.array, all.test.guide.w$Gene %>% unique()) %>% append(CAD.focus.gene.set) %>% append(gene.set.type.df %>% subset(grepl("CAD_Loci", type)) %>% pull(Gene) %>% sort()) # ptb.array <- append(ptb.array, all.test.guide.w %>%subset(adjusted.p.value < 0.1) %>% pull(Gene) %>% unique()) # # ptb.array <- CAD.focus.gene.set[grepl("E_at_", CAD.focus.gene.set)] # for (mode.selection in c("per.guide", "per.cell")){ # ## new! # for (ptb in ptb.array) { # make a separate one for enhancers, clear up cases like E_at_AAGAB,IQCH,LOC102723493&SMAD3 # print(paste0(ptb, "\n\n")) # if(grepl("E_at_", ptb)) { # target.gene <- strsplit(gsub("E_at_","",ptb %>% strsplit(split="-") %>% unlist() %>% nth(1)), split=",|&") %>% unlist() %>% as.character() # if((target.gene %in% colnames(ann.X.full.filtered)) %>% as.numeric() %>% sum() > 0 & # target gene must express # (ptb %in% ann.X.full.filtered$Gene) %>% as.numeric() %>% sum() > 0) { # enhancer must have enough cells and guides # pdf(file=paste0(FIGDIRSAMPLE,"/gene.by.topic/",ptb, ".", mode.selection, ".pdf"), width=15, height=8) # for(expressed.gene in target.gene) { # if(expressed.gene %in% colnames(ann.X.full.filtered)){ # print(paste0("Enhancer of ", expressed.gene, "\n\n")) # gene.index <- which(grepl(expressed.gene, ptb.10X.name.conversion$Symbol)) # if(length(gene.index) == 1) { # ptb.expression.name <- ptb.10X.name.conversion$`Name used by CellRanger`[gene.index] # } else { # ptb.expression.name <- expressed.gene # } # giant.summary.plot(ptb, ptb.expression.name, expressed.gene, mode.selection, enhancer=T) # } # } # dev.off() # } # } else{ # expressed.gene <- ptb %>% strsplit(split="-") %>% unlist() %>% nth(1) # gene.index <- which(grepl(ptb, ptb.10X.name.conversion$Symbol)) # if(length(gene.index) == 1) { # ptb.expressed.name <- ptb.10X.name.conversion$`Name used by CellRanger`[gene.index] # expressed.gene <- ptb.expressed.name # } else { # ptb.expressed.name <- expressed.gene # } # message(paste0("Expressed gene: ", expressed.gene, ", Perturbation: ", ptb, ", with expressed name: ", ptb.expressed.name)) # if((colnames(X.full) == expressed.gene) %>% as.numeric() %>% sum() > 0) giant.summary.plot(ptb, ptb.expressed.name, expressed.gene, mode.selection, enhancer=F) # } # } # ##HERE # } # if(opt$subsample.type!="ctrl") # ##scratch:210818 # KDefficacyFIGDIR <- paste0(FIGDIRSAMPLE, "/KD.efficacy/") # if(!dir.exists(KDefficacyFIGDIR)) dir.create(KDefficacyFIGDIR) # ptb.array <- c("RHOA", "PECAM1", "RAP1A", "KLF4", "MEF2C", "EGR1", "CDC42EP2", "TSPAN14", "NFATC2") ##210823 list for topic 15 # ptb.array <- ref.table$Symbol %>% unique() # ## per-guide KD efficacy # for (ptb in ptb.array) { # subset.index <- which(X.gene.names == ptb) # gene.index <- which(grepl(ptb, ptb.10X.name.conversion$Symbol)) # if(length(gene.index) == 1) { # expressed.gene <- ptb.10X.name.conversion$`Name used by CellRanger`[gene.index] # } else { # expressed.gene <- ptb # } # ##TODO: add rep1rep2 to ctrl.X %>% subset() # ## for SEP=T, subset ctrl.X to the right sample's control # if(SEP) { # label.here <- strsplit(ptb, split="-") %>% unlist() %>% nth(2) %>% paste0("-",.) # ctrl.X.here <- ctrl.X %>% subset(grepl(label.here,Gene)) # ctrl.ann.omega.here <- ctrl.ann.omega %>% subset(grepl(label.here,Gene)) # } else { # ctrl.X.here <- ctrl.X # ctrl.ann.omega.here <- ctrl.ann.omega # } # ## construct table for plotting # toPlot.list <- get.average.ptb.gene.expression.based.on.ctrl(expressed.gene, ann.X.full.filtered[subset.index,], ctrl.X.here, mode=mode.selection) # toPlot1 <- toPlot.list[["toPlot"]] # order.toPlot.Gene <- function(df) { # df$Gene <- factor(x=df$Gene, levels=df$Gene[c(which(grepl("^Control", df$Gene)) %>% min(), which(!grepl("^Control", df$Gene))%>% min())]) # return(df) # } # order.toPlot <- function(df, column) { # array <- df %>% pull(all_of(column)) # df <- df %>% mutate(!!column:=factor(x=array, levels=array[c(which(grepl("^Control", array)) %>% min(), which(!grepl("^Control", array))%>% min())])) # return(df) # } # toPlot1 <- order.toPlot(toPlot1, column="Gene") # ## data points # KD.per.guide <- toPlot.list[["toCalculate"]] # colnames(KD.per.guide)[colnames(KD.per.guide)=="Gene.expression"] <- "mean.expression" # ## plot # p.KD.efficacy.barplot <- toPlot1 %>% ggplot(aes(x=Gene,y=mean.expression)) + geom_bar(stat='identity', width=0.5, fill="#38b4f7") + # geom_errorbar(data = toPlot1, aes(x=Gene, ymin=mean.expression-error.bar, ymax=mean.expression+error.bar), width=.15) + # # ylim(min(KD.per.guide$mean.expression - 25, 0), max(KD.per.guide$mean.expression + 50, 150)) + # ylab(paste0(expressed.gene, " RNA Expression\n(% vs control)")) + xlab("Guides") + mytheme + # # scale_y_continuous(limits = c(0, max(KD.per.guide$mean.expression + 50, 150)), breaks = round(seq(0, max(KD.per.guide$mean.expression + 50, 150), 20),20) ) + # geom_hline(yintercept = 100, color="grey", linetype="dashed", size=0.5) + # geom_jitter(data=KD.per.guide, size=0.25, width=0.15, color="red") # pdf(file=paste0(KDefficacyFIGDIR, "ptb.", ptb, "_KD.efficacy.pdf"), width=4, height=6) # print(p.KD.efficacy.barplot) # dev.off() # } ## commented out 211013 # ## topic 29 expression log2FC versus CDH5 RNA expression log2FC ##todo:210812 # for (t in 1:k) { # ## t <- 29 ##scratch # topic <- paste0("topic_",t) # ## plot location # perFACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/factor", t, "/") # if(!dir.exists(perFACTORFIGDIR)) dir.create(perFACTORFIGDIR) # perFACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/factor", t, "/", SAMPLE,"_K",k, "_dt_", DENSITY.THRESHOLD,"_factor", t, "_") # ## ptb.name <- "CDH5" ##scratch top perturbations # top.ptb.name <- gene.score %>% as.data.frame %>% mutate(Gene = rownames(.)) %>% select(all_of(topic), Gene) %>% arrange(desc(get(topic))) %>% slice(1:10) %>% pull(Gene) # for( ptb.name in top.ptb.name) { # topic.log2fc.here <- log2fc.omega %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% subset(grepl(ptb.name, Gene)) ## get Topic expression for Gene G # topic.fc.here <- fc.omega %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% subset(grepl(ptb.name, Gene)) # ## inf.index <- topic.log2fc.here %>% pull(all_of(topic)) %>% is.infinite %>% which ## get cell index for zero Topic expression entries # ## neg.inf.index <- (topic.fc.here %>% pull(all_of(topic)) == 0) %>% which # ## pos.inf.index <- which(!(inf.index %in% neg.inf.index)) # ## topic.log2fc.here[[topic]][inf.index] <- 0 # ## min.log2fc.here <- topic.log2fc.here %>% pull(all_of(topic)) %>% min # ## max.log2fc.here <- topic.log2fc.here %>% pull(all_of(topic)) %>% max # ## topic.log2fc.here[[topic]][neg.inf.index] <- floor(min.log2fc.here - 3) # ## topic.log2fc.here[[topic]][pos.inf.index] <- ceil(max.log2fc.here + 3) # remove.inf <- function(topic.log2fc.here, topic.fc.here, topic) { # topic.index <- which(grepl(topic, colnames(topic.log2fc.here))) # inf.index <- topic.log2fc.here[,topic.index] %>% is.infinite %>% which # neg.inf.index <- (topic.fc.here[,topic.index] == 0) %>% which # ## inf.index <- topic.log2fc.here %>% pull(all_of(topic)) %>% is.infinite %>% which ## get cell index for zero Topic expression entries # ## neg.inf.index <- (topic.fc.here %>% pull(all_of(topic)) == 0) %>% which # pos.inf.index <- which(!(inf.index %in% neg.inf.index)) # topic.log2fc.here[inf.index,topic.index] <- 0 # min.log2fc.here <- topic.log2fc.here[,topic.index] %>% min # max.log2fc.here <- topic.log2fc.here[,topic.index] %>% max # ## min.log2fc.here <- topic.log2fc.here %>% pull(all_of(topic)) %>% min # ## max.log2fc.here <- topic.log2fc.here %>% pull(all_of(topic)) %>% max # topic.log2fc.here[neg.inf.index,topic.index] <- floor(min.log2fc.here - 3) # topic.log2fc.here[pos.inf.index,topic.index] <- ceil(max.log2fc.here + 3) # return(topic.log2fc.here) # } # topic.log2fc.here <- remove.inf(topic.log2fc.here, topic.fc.here, topic) # ## select RNA expresion # ## todo: function to extract df # X.full.here <- X.full %>% subset(grepl(ptb.name, rownames(.))) %>% `colnames<-`(gsub(":.*$", "", colnames(.))) # fc.X.full.here <- fc.X.full %>% subset(grepl(ptb.name, rownames(.))) %>% `colnames<-`(gsub(":.*$", "", colnames(.))) # log2fc.X.full.here <- log2fc.X.full %>% subset(grepl(ptb.name, rownames(.))) %>% `colnames<-`(gsub(":.*$", "", colnames(.))) # get.topicFC.vs.RNAexpFC.toPlot <- function(ptb.log2fc.df, expressed.ptb.name, RNA.full.here) { # ## ##### function to combine per cell topic data and gene expression data # ## INPUT # ## ptb.log2fc.df: topic log2FC df subset to one perturbation only # ## expressed.ptb.name: the expressed gene # ## X.full.here: matrix subset to one perturbation only, the values could be RNA expression count, FC, or log2FC. # ## # ## OUTPUT # ## toPlot: a dataframe that has ptb only cells and one topic column and one expressed gene column # ## # expressed.gene.index <- which(grepl(paste0("^", paste0(expressed.ptb.name, collapse="$|^"), "$"), colnames(RNA.full.here))) # expressed.gene.RNA.here <- RNA.full.here[,expressed.gene.index] %>% as.data.frame %>% `colnames<-`(expressed.ptb.name) # toPlot <- merge(ptb.log2fc.df, expressed.gene.RNA.here, by.x="Gene", by.y=0) # return(toPlot) # } # ## ptb.name.index <- which(grepl(ptb.name, colnames(log2fc.X.full.here))) # ## log2fc.RNA.expression.here <- log2fc.X.full.here[,ptb.name.index] %>% as.data.frame %>% `colnames<-`(ptb.name) # ## RNA.expression.here <- X.full.here[,ptb.name.index] %>% as.data.frame %>% `colnames<-`(ptb.name) # ## topic.RNA.expression.here <- merge(log2fc.here, RNA.expression.here, by.x="Gene", by.y=0) # if(ptb.name=="MESDC1") ptb.name <- "TLNRD1" # if(ptb.name %in% colnames(X.full.here)) { # log2fc.X.full.here <- remove.inf(log2fc.X.full.here, fc.X.full.here, ptb.name) # topic.RNA.expression.here <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, ptb.name, X.full.here) # } else { # warning(paste0(ptb.name, " is not in the 10X gene list")) # } # ## plot topic FC vs RNA expression # pdf(file=paste0(perFACTORFIGDIR, "ptb.", ptb.name, "_FC.vs.RNA.expression.pdf"))##here210818 # p <- topic.RNA.expression.here %>% ggplot(aes(x=get(ptb.name), y=get(topic))) + geom_point() + mytheme + # ggtitle(paste0(ptb.name, " Perturbation")) + # xlab(paste0(ptb.name, " RNA Expression")) + ylab(paste0("Topic ", gsub("topic_","",topic), " Expression (log2FC)")) # print(p) # dev.off() # ## EDN1 (or top genes in topic29) expression log2FC in CDH5 KD cells versus topic 29 log2FC in CDH5 KD cells ##todo:210812 # ## get list of top perturbations # top.expressed.gene.in.topic <- theta.zscore[,t] %>% sort(decreasing=T) %>% head(num_ptb) %>% names # ## concatenate all top gene in topic RNA expression # topic.expressed.gene.RNA.df <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, top.expressed.gene.in.topic, X.full.here) # topic.expressed.gene.RNA.fc.df <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, top.expressed.gene.in.topic, fc.X.full.here) # topic.expressed.gene.RNA.log2fc.df <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, top.expressed.gene.in.topic, log2fc.X.full.here) # ## melt the dfs # topic.expressed.gene.RNA.long <- topic.expressed.gene.RNA.df %>% melt(id.vars = c("Gene", topic), value.name = "RNA.expression.count", variable.name = "expressed.gene.name") # topic.expressed.gene.RNA.fc.long <- topic.expressed.gene.RNA.fc.df %>% melt(id.vars = c("Gene", topic), value.name = "RNA.expression.fc", variable.name = "expressed.gene.name") # topic.expressed.gene.RNA.log2fc.long <- topic.expressed.gene.RNA.log2fc.df %>% melt(id.vars = c("Gene", topic), value.name = "RNA.expression.log2fc", variable.name = "expressed.gene.name") # ## ## violin plot # ## p.violin <- topic.expressed.gene.RNA.long %>% ggplot(aes(x=expressed.gene.name, y=RNA.expression.count)) + mytheme + # ## xlab(paste0("Top Specific Gene in Topic ", t)) + ylab("RNA expression count") + # ## ggtitle(paste0(ptb.name, " Perturbation, Topic ", t, " Top Specific Gene Expression")) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = .4, # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") # ## p.fc.violin <- topic.expressed.gene.RNA.fc.long %>% ggplot(aes(x=expressed.gene.name, y=RNA.expression.fc)) + mytheme + # ## xlab(paste0("Top Specific Gene in Topic ", t)) + ylab("RNA expression (FC)") + # ## ggtitle(paste0(ptb.name, " Perturbation, Topic ", t, " Top Specific Gene Expression")) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = .4, # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## geom_hline( # ## yintercept = 1, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) # ## p.log2fc.violin <- topic.expressed.gene.RNA.log2fc.long %>% ggplot(aes(x=expressed.gene.name, y=RNA.expression.log2fc)) + mytheme + # ## xlab(paste0("Top Specific Gene in Topic ", t)) + ylab("RNA expression (log2FC)") + # ## ggtitle(paste0(ptb.name, " Perturbation, Topic ", t, " Top Specific Gene Expression")) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = .4, # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## geom_hline( # ## yintercept = 0, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) # ## pdf(file=paste0(perFACTORFIGDIR, "ptb.", ptb.name, "_top.gene.RNA.expression.violin.pdf")) # ## print(p.violin) # ## print(p.fc.violin) # ## print(p.log2fc.violin) # ## dev.off() # ## get RNA expression FC and topic FC df for KD efficacy plot and topic expressio change plot # ## negative control df # ctrl.topic.fc <- fc.omega %>% select(all_of(topic)) %>% mutate(Gene = rownames(.)) %>% subset(grepl("negative|safe", Gene)) # ctrl.topic.RNA.expression.here <- get.topicFC.vs.RNAexpFC.toPlot(ctrl.topic.fc, top.expressed.gene.in.topic, ctrl.X) %>% mutate(CBC=Gene, Gene="Negative Control") # ## perturbation df # ptb.topic.RNA.expression.here <- get.topicFC.vs.RNAexpFC.toPlot(topic.fc.here, top.expressed.gene.in.topic, X.full.here) %>% mutate(CBC=Gene, Gene=ptb.name) # ## combine perturbation and control dfs # ptb.with.ctrl <- rbind(ctrl.topic.RNA.expression.here, ptb.topic.RNA.expression.here) # ## plot gene expression (% vs control) distribution for perturbation and for control side-by-side (violin plot) # p.list <- vector("list",length(top.expressed.gene.in.topic)) # for ( expr.gene.index in 1:length(top.expressed.gene.in.topic) ) { # expr.gene <- top.expressed.gene.in.topic[expr.gene.index] # toPlot <- ptb.with.ctrl %>% select(Gene, all_of(topic), all_of(expr.gene)) # ptb.ary <- toPlot %>% subset(Gene == ptb.name) %>% pull(all_of(expr.gene)) # ctr.ary <- toPlot %>% subset(Gene == "Negative Control") %>% pull(all_of(expr.gene)) # toPlot <- toPlot %>% melt(id.vars=c("Gene", topic), variable.name = "expr.gene", value.name = "log2fc.RNA.expression") # p.value <- wilcox.test(ptb.ary, ctr.ary)$p.value # p <- toPlot %>% ggplot(aes(x=expr.gene, y=log2fc.RNA.expression, fill=Gene)) + geom_split_violin() + mytheme + # xlab(paste0("(p-value: ", format.pval(p.value, digits=4), ")")) + ylab("RNA Expression (count)") # p.list[[expr.gene.index]] <- p # } # num_plot_row <- ceil(length(p.list)/5) # p <- ggarrange(plotlist = p.list, ncol = 5, nrow = num_plot_row, common.legend=T, legend="bottom") # annotate_figure(p, left = "RNA Expression (log2FC vs Control)") # pdf(paste0(perFACTORFIGDIR, "ptb.", ptb.name, "_top.gene.RNA.expression.violin.pdf"), width=10, height=3*num_plot_row) # print(p) # dev.off() # ## ## plot KD efficacy # ## p.KD.efficacy <- ptb.with.ctrl %>% ggplot(aes(x=Gene, y=get(ptb.name)*100)) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = 0.3, # ## ## control verticle range of jitter # ## transformation = position_jitter(height = 10), # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## mytheme + # ## geom_hline( # ## yintercept = 100, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) + # ## ylab(paste0(ptb.name, " RNA Expression\n(% vs control)")) + # ## xlab("Perturbation") # ## ## copied from summary plots # ## if(SEP) { # ## label.here <- strsplit(ptb, split="-") %>% unlist() %>% nth(2) %>% paste0("-",.) # ## ctrl.X.here <- ctrl.X %>% subset(grepl(label.here,Gene)) # ## ctrl.ann.omega.here <- ctrl.ann.omega %>% subset(grepl(label.here,Gene)) # ## } else { # ## ctrl.X.here <- ctrl.X # ## ctrl.ann.omega.here <- ctrl.ann.omega # ## } # ## p.topic.expression.with.ctrl <- ptb.with.ctrl %>% ggplot(aes(x=Gene, y=get(topic)*100)) + # ## ggdist::stat_halfeye( # ## ## custom bandwidth # ## adjust = .5, # ## ## adjust height # ## width = .3, # ## ## move geom to the right # ## justification = -.4, # ## ## remove slab interval # ## .width = 0, # ## point_colour = NA # ## ) + # ## geom_boxplot( # ## width = .1, # ## ## remove outliers # ## outlier.color = NA ## `outlier.shape = NA` works as well # ## ) + # ## ## add justified jitter from the {gghalves} package # ## gghalves::geom_half_point( # ## ## control point size # ## size = 0.5, # ## ## draw jitter on the left # ## side = "l", # ## ## control range of jitter # ## range_scale = 0.3, # ## ## control verticle range of jitter # ## transformation = position_jitter(height = 10), # ## ## add some transparency # ## alpha = .3 # ## ) + # ## coord_cartesian(xlim = c(1.2, NA), clip = "off") + # ## mytheme + # ## geom_hline( # ## yintercept = 100, # ## linetype = "dashed", # ## color = "#38b4f7", # ## size = 0.5 # ## ) + # ## ylab(paste0("Topic ", t, " Expression\n(% vs control)")) + # ## xlab("Perturbation") # ## pdf(file=paste0(perFACTORFIGDIR, "ptb.", ptb.name, "_topic", t, ".expression.vs.ctrl.pdf")) # ## print(p.topic.expression.with.ctrl) # ## dev.off() # pdf(file=paste0(perFACTORFIGDIR, "ptb.", ptb.name, "_log2FC.vs.top.gene.RNA.expression.pdf")) # for (expressed.gene.here in top.expressed.gene.in.topic) { # print(paste0("plotting ", expressed.gene.here)) # ## expressed.gene.here <- "EDN1" # loop over top genes in topic # ## old (converted to get.topicFC.vs.RNAexpFC.toPlot()) # ## expressed.ptb.name.index <- which(grepl(expressed.gene.here, colnames(log2fc.X.full.here))) # ## expressed.gene.RNA.here <- X.full.here[,expressed.ptb.name.index] %>% as.data.frame %>% `colnames<-`(expressed.gene.here) # ## topic.expressed.gene.RNA.here <- merge(topic.log2fc.here, expressed.gene.RNA.here, by.x="Gene", by.y=0) # topic.expressed.gene.RNA.here <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, expressed.gene.here, X.full.here) # topic.expressed.gene.RNA.fc.here <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, expressed.gene.here, fc.X.full.here) # topic.expressed.gene.RNA.log2fc.here <- get.topicFC.vs.RNAexpFC.toPlot(topic.log2fc.here, expressed.gene.here, log2fc.X.full.here) # p.raw.x <- topic.expressed.gene.RNA.here %>% ggplot(aes(y=get(expressed.gene.here), x=get(topic))) + geom_point() + mytheme + # ggtitle(paste0(ptb.name, " Perturbation")) + # ylab(paste0(expressed.gene.here, " RNA Expression")) + xlab(paste0("Topic ", gsub("topic_","",topic), " Expression (log2FC)")) # p.fc.x <- topic.expressed.gene.RNA.fc.here %>% ggplot(aes(y=get(expressed.gene.here), x=get(topic))) + geom_point() + mytheme + # ggtitle(paste0(ptb.name, " Perturbation")) + # ylab(paste0(expressed.gene.here, " RNA Expression (FC)")) + xlab(paste0("Topic ", gsub("topic_","",topic), " Expression (log2FC)")) # p.log2fc.x <- topic.expressed.gene.RNA.log2fc.here %>% ggplot(aes(y=get(expressed.gene.here), x=get(topic))) + geom_point() + mytheme + # ggtitle(paste0(ptb.name, " Perturbation")) + # ylab(paste0(expressed.gene.here, " RNA Expression (log2FC)")) + xlab(paste0("Topic ", gsub("topic_","",topic), " Expression (log2FC)")) # ## output plots # print(p.raw.x) ## fit a line # print(p.fc.x) # print(p.log2fc.x) # } # dev.off() # } # } # ##end of scratch:210818 # ##scratch:210825 # t <- 15 # topic <- paste0("topic_",t) # perFACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/factor", t, "/") # if(!dir.exists(perFACTORFIGDIR)) dir.create(perFACTORFIGDIR) # perFACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/factor", t, "/", SAMPLE,"_K",k, "_dt_", DENSITY.THRESHOLD,"_factor", t, "_") # top.ptb.name <- gene.score %>% as.data.frame %>% mutate(Gene = rownames(.)) %>% select(all_of(topic), Gene) %>% arrange(desc(get(topic))) %>% slice(1:10) %>% pull(Gene) # top.ptb.name <- gsub("WTAP2", "WTAP", top.ptb.name) # ptb <- "MESDC1" # pdf(paste0(perFACTORFIGDIR, "top.ptb_pval.log2fc.expr.gene.volcano.pdf"), width=5, height=5) # for ( ptb in top.ptb.name ) { # ##use.edgeR.results # edgeR.expr.gene.names <- rownames(log2fc.edgeR) %>% strsplit(split=":") %>% sapply("[[",1) # ptb.colindex <- which(grepl(ptb, colnames(log2fc.edgeR))) # theta.zscore.t <- theta.zscore[,t] %>% as.data.frame %>% `colnames<-`("topic.zscore.weight") %>% mutate(genes=rownames(.)) %>% arrange(desc(topic.zscore.weight)) %>% mutate(gene.rank = 1:n(), top100=ifelse(gene.rank <= 100, paste0("Top 100 in Topic ", t), paste0("Not in Top 100"))) # ptb.pval.log2fc <- merge(p.value.edgeR[,c(1,ptb.colindex)] %>% `colnames<-`(c("genes", "pval")) %>% mutate(genes = strsplit(genes, split=":") %>% sapply("[[",1)), log2fc.edgeR[,c(1,ptb.colindex)] %>% `colnames<-`(c("genes", "log2fc")) %>% mutate(genes = strsplit(genes, split=":") %>% sapply("[[",1)), by="genes") %>% merge(theta.zscore.t, by="genes") %>% mutate(nlog10pval = -log10(pval)) # ptb.pval.log2fc$nlog10pval[ptb.pval.log2fc$nlog10pval > 15] <- 15 # ptb.pval.log2fc$top100 <- factor(ptb.pval.log2fc$top100) %>% ordered(levels=c(paste0("Top 100 in Topic ", t), "Not in Top 100")) # label <- ptb.pval.log2fc %>% subset(top100!="Not in Top 100") # ptb.10X <- ptb.10X.name.conversion$`Name used by CellRanger`[which(grepl(ptb, ptb.10X.name.conversion$Symbol))] # label.self <- ptb.pval.log2fc %>% subset(genes == ptb.10X) # p.density <- ptb.pval.log2fc %>% ggplot(aes(x=log2fc, fill=top100)) + geom_density(alpha=0.4) + mytheme + xlab("RNA Expression\n(log2FC vs Control") + coord_flip() + scale_fill_manual(values=c("red", "gray"), name = "") + theme(legend.position = "none") # + guides(fill=guide_legend(nrow=2, byrow=T)) # p <- ggplot(ptb.pval.log2fc, aes(x=log2fc, y=nlog10pval)) + geom_point(size=0.1, color = "gray") + # geom_point(data = label, size=0.1, color = "red") + # mytheme + # geom_text_repel(data=label.self, box.padding = 0.5, # max.overlaps=30, # aes(label=genes), size=4, # color="blue") + # geom_text_repel(data=label, box.padding = 0.5, # max.overlaps=30, # aes(label=genes), size=4, # color="black") + # scale_color_manual(values=c("gray", "red")) + # geom_vline(xintercept=0, col = "#38b4f7", lty=3) + # theme(legend.position="bottom") + # ggtitle(paste0("Topic ", t, " Perturbation ", ptb)) + # xlab("Average RNA Expression (log2FC vs control)") + ylab("p-value (-log10)") + # inset_element(p.density, left = 0.7, bottom = 0.05, right=0.95, top = 0.55) # print(p) # } # dev.off() # ##end of use.edgeR.results # ## ##redo.wilcoxon.test # ## ptb.ctrl.rowindex <- which(grepl(ptb,X.gene.names) | grepl("negative|safe", X.gene.names)) # ## ptb.ctrl.X <- X.full[ptb.ctrl.rowindex,] # ## ptb.rowindex.new <- which(grepl(ptb, rownames(ptb.ctrl.X))) ## cells with ptb # ## ctrl.rowindex.new <- which(!grepl(ptb, rownames(ptb.ctrl.X))) ## ctrl cells # ## log2fc.ptb.ctrl.X <- log2fc.X.full[ptb.ctrl.rowindex,] # ## df <- do.call(rbind, lapply(1:dim(ptb.ctrl.X)[2], function(expr.gene.index) { # ## expr.gene <- colnames(ptb.ctrl.X)[expr.gene.index] # ## p.value <- wilcox.test(ptb.ctrl.X[ptb.rowindex.new,expr.gene.index], ptb.ctrl.X[ctrl.rowindex.new,expr.gene.index])$p.value # ## return(data.frame(ptb = ptb, # ## expr.gene = expr.gene, # ## avg.log2fc.RNA.expr = log2fc.ptb.ctrl.X[ptb.rowindex.new, expr.gene.index] %>% mean, # ## p.value = p.value)) # ## })) # ## ##end of redo.wilcoxon.test # ##end of scratch:210825 # ####################################################### # ## scatter plot of ( average log2FC of gene KD compared to control ) vs ( weight in the topic ) # ##scratch:210823 # t <- 15 ## loop over topics # topic <- paste0("topic_",t) # FACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/factor", t, "/") # if(!dir.exists(FACTORFIGDIR)) dir.create(FACTORFIGDIR) # FACTORFIGDIR <- paste0(FACTORFIGDIR, SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_factor", t, "_") # pdf(paste0(FACTORFIGDIR, "top10.ptb.in.topic_RNA.expr.avg.log2fc_zscore.weight.pdf")) # top.ptb <- gene.score %>% as.data.frame %>% select(all_of(topic)) %>% arrange(desc(get(topic))) # top.gene.zscore <- theta.zscore[,t] %>% sort(decreasing=T) # top.gene.names <- names(top.gene.zscore)[1:100] # top.gene.col.index <- which(grepl(paste0("^", paste0(top.gene.names, collapse="$|^"), "$"), colnames(fc.X.full))) # num_ptb <- 10 # toPlot.list <- vector("list", num_ptb) # for (ptb.index in 1:num_ptb) { # ptb <- rownames(top.ptb)[ptb.index] # ## ptb <- "MESDC1" ## loop based on top perturbation # ## subset RNA expression log2fc matrix # ptb.row.index <- X.gene.names %>% grepl(paste0("^",ptb,"$"), .) %>% which # ptb.fc.top100.topic.genes <- fc.X.full[ptb.row.index, top.gene.col.index] # # adjust for -Inf # ptb.mean.log2fc.top100.topic.genes <- ptb.fc.top100.topic.genes %>% apply(2, mean) %>% log2 ## apply log2 after the normalization and average # ## merge y-axis average expression with x-axis gene weight in topic # toPlot <- merge(data.frame(theta.zscore=top.gene.zscore[1:100], # Gene=names(top.gene.zscore[1:100])), # data.frame(log2fc.RNA.expr=ptb.mean.log2fc.top100.topic.genes, # Gene=names(ptb.mean.log2fc.top100.topic.genes)), # by="Gene") # toPlot.list[[ptb.index]] <- toPlot %>% mutate(Perturbation = ptb, log2fc.topic.expr=top.ptb %>% subset(rownames(.) == ptb) %>% pull(all_of(topic))) # ## scatter plot # p <- toPlot %>% ggplot(aes(x=theta.zscore, y=log2fc.RNA.expr)) + geom_point() + mytheme + # xlab(paste0("Topic ", t, " z-score (Specificity) Weight")) + ylab(paste0("Average RNA Expression (log2FC)")) + # ggtitle(paste0(ptb, " Perturbation Top 100 (by z-score) Gene Expression in Topic ", t)) + # geom_text_repel(data=toPlot, box.padding = 0.5, # aes(label=Gene), size=4, # color="black") # ## pdf(paste0(FACTORFIGDIR, "ptb.", ptb, "_RNA.expr.avg.log2fc_zscore.weight.pdf")) # print(p) # } # dev.off() # toPlot <- do.call(rbind, toPlot.list) # pdf(paste0(FACTORFIGDIR, "top10.ptb.in.topic_RNA.expr.avg.log2fc_zscore.weight.scratch.pdf")) # p <- toPlot %>% ggplot(aes(x = log2fc.topic.expr, y = log2fc.RNA.expr, color = theta.zscore)) + geom_point(size=0.5) + mytheme + # xlab(paste0("Topic ", t, " Expression (log2FC)")) + ylab(paste0("Average RNA Expression (log2FC)")) # print(p) # p <- toPlot %>% ggplot(aes(x=theta.zscore, y=log2fc.RNA.expr, color=Perturbation)) + geom_point(size=0.1) + mytheme + # xlab(paste0("Topic ", t, " z-score (Specificity) Weight")) + ylab(paste0("Average RNA Expression (log2FC)")) + # ggtitle(paste0("Top 100 (by z-score) Gene Expression in Topic ", t)) # print(p) # dev.off() # ##end of scratch:210823 # ##scratch:210824 # ## violin plot of top 30 genes # ##end of scratch:210824 # ## per factor summary plot by test # FACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/") # if(!dir.exists(FACTORFIGDIR)) dir.create(FACTORFIGDIR) # FACTORFIGDIR <- paste0(FIGDIRSAMPLE, "factor.summary/", SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") # ptb.zscore.long <- ptb.zscore %>% as.data.frame %>% mutate(Gene=rownames(.)) %>% melt(id.vars = "Gene", value.name = "perturbation.zscore", variable.name = "Topic") ## Perturbation z-score plot # for(test.type.here in c("per.cell.wilcoxon", "per.guide.wilcoxon")) { # all.test.w <- all.test %>% subset(test.type==test.type.here) # realPvals.df.w <- realPvals.df %>% subset(test.type==test.type.here) # for (t in 1:dim(omega)[2]) { #:dim(omega)[2] # figure.path <- paste0(FACTORFIGDIR, "factor", t, "_with.sig.", test.type.here, ".0.1.perturbation") # ## raw weight version # topic <- paste0("topic_",t) # toPlot.all.test <- all.test.w %>% subset(Topic==topic) # toPlot.fdr <- realPvals.df.w %>% subset(Topic == topic) %>% select(Gene,fdr) # ## toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), # ## Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) %>% # ## merge(., gene.def.pathways, by="Gene", all.x=T) %>% arrange(desc(Score)) %>% slice(1:10) # ## toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # toPlot <- data.frame(Gene=topFeatures.raw.weight %>% subset(topic == t) %>% pull(Gene), # Score=topFeatures.raw.weight %>% subset(topic == t) %>% pull(scores)) %>% # merge(., summaries %>% select(Symbol, top_class), by.x="Gene", by.y="Symbol", all.x=T) %>% `colnames<-`(c("Gene", "Score", "Pathway")) %>% arrange(desc(Score)) %>% slice(1:10) # toPlot$Pathway[is.na(toPlot$Pathway)] <- "Other/Unclassified" # toPlot$Pathway[toPlot$Pathway == "unclassified"] <- "Other/Unclassified" # p4 <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score*100, fill=Pathway) ) + geom_col(width=0.5) + theme_minimal() # p4 <- p4 + coord_flip() + xlab("Top 10 Genes") + ylab("Raw Weights (z-score)") + # mytheme + theme(legend.position="bottom", legend.direction="vertical") # # plot 4 # # add ABC to gene.set.type.df for this particular plot # ## gene.set.type.df$type[which(gene.set.type.df$Gene %in% gene.set)] <- "ABC" # # assemble toPlot # toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% # merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # ## merge(.,gene.set.type.df,by="Gene") %>% # mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) %>% # arrange(log2FC) %>% # mutate(x = seq(nrow(.))) # label <- toPlot %>% subset(x <= 3 | x > (nrow(toPlot)-3) | adjusted.p.value < fdr.thr) # mytheme <- theme_classic() + theme(axis.text = element_text(size = 13), axis.title = element_text(size = 15), plot.title = element_text(hjust = 0.5, face = "bold")) # p5 <- toPlot %>% ggplot(aes(x=reorder(Gene, log2FC), y=log2FC, color=significant)) + geom_point(size=0.75) + mytheme + # theme(axis.ticks.x=element_blank(), axis.text.x=element_blank()) + scale_color_manual(values = c("#E0E0E0", "#38b4f7")) + # xlab("Perturbed Genes") + ylab(paste0("Factor ", t, " Expression log2 Fold Change")) + # geom_text_repel(data=label, box.padding = 0.5, # aes(label=ann.Gene), size=4, # color="black") + # theme(legend.position = "none") # legend at the bottom? # ## top perturbation list # ## toPlot.all.test <- all.test.w # ## toPlot.fdr <- realPvals.df.w %>% subset(Topic == topic) %>% select(Gene,adjusted.p.value) # # assemble toPlot # toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene") %>% # mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$type == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", type), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", type), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # ## toPlot <- gene.score %>% select(all_of(topic)) %>% mutate(Gene=rownames(.)) %>% merge(.,toPlot.all.test,by="Gene", all.x=T) %>% # ## merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # ## merge(.,gene.set.type.df,by="Gene") %>% # ## mutate(Gene = gsub("Enhancer-at-CAD-SNP-","",Gene)) # ## colnames(toPlot)[which(colnames(toPlot)==topic)] <- "log2FC" # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(log2FC)) %>% slice(1:25) # toPlot.bottom <- toPlot %>% arrange(log2FC) %>% slice(1:25) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(type %in% c("ABC","CAD focus"), "red", # ifelse(type=="non-expressed", "grey", # ifelse(type=="other", "blue", "black")))) %>% # mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) # # mutate(Gene = ifelse(type=="ABC", paste0("[ ", Gene, " ]"), Gene)) # p.ptb.list <- toPlot.extreme %>% arrange(desc(log2FC)) %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=reorder(Gene, log2FC), y=log2FC, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("log2 Fold Change") + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none")#, axis.text.y = element_text(colour = toPlot.extreme$color)) # ## ## enriched gene sets by fgsea # ## fgsea.here <- fgsea.df.all %>% subset(topic == t) %>% arrange(padj) # ## Perturbation z-score # toPlot.all.test <- all.test %>% subset(test.type=="per.cell.wilcoxon" & Topic==topic) # toPlot.fdr <- realPvals.df %>% subset(test.type=="per.cell.wilcoxon" & Topic == topic) %>% select(Gene,fdr) # # assemble toPlot # toPlot <- ptb.zscore.long %>% subset(Topic == topic) %>% merge(.,toPlot.all.test,by=c("Gene","Topic"), all.x=T) %>% # merge(.,toPlot.fdr,by="Gene", all.x=T) %>% # merge(.,gene.set.type.df,by="Gene", all.x=T) %>% ##here210809 # ## merge(.,gene.def.pathways, by="Gene", all.x=T) %>% # merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Gene", by.y="Symbol", all.x=T) %>% # mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% # mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), # ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% # mutate(ann.Gene = paste0(Gene, GWAS.class.text, EC_ctrl_text)) # toPlot <- toPlot %>% mutate(significant=ifelse((adjusted.p.value >= fdr.thr | is.na(adjusted.p.value)), "", "*")) # toPlot.top <- toPlot %>% arrange(desc(perturbation.zscore)) %>% slice(1:25) # toPlot.bottom <- toPlot %>% arrange(perturbation.zscore) %>% slice(1:25) # toPlot.extreme <- rbind(toPlot.top, toPlot.bottom) %>% # mutate(color=ifelse(grepl("CAD", type), "red", # ifelse(type=="non-expressed", "gray", # ifelse(type=="EC_ctrls", "blue", "black")))) %>% # mutate(color=ifelse(is.na(type), "black", color)) # ## colors <- toPlot.extreme$color[order(toPlot.extreme %>% arrange(desc(perturbation.zscore)) %>% pull(color))] # toPlot.extreme <- toPlot.extreme %>% arrange(perturbation.zscore) # ## add gene distance to CAD # toPlot.extreme$ann.Gene <- factor(toPlot.extreme$ann.Gene, levels = toPlot.extreme$ann.Gene) # ## color y.axis.label # p.ptb.zscore <- toPlot.extreme %>% #mutate(Gene = paste0("<span style = 'color: ", color, ";'>", Gene, "</span>")) %>% # ggplot(aes(x=ann.Gene, y=perturbation.zscore, fill=significant)) + geom_col() + theme_minimal() + # coord_flip() + xlab("Most Extreme Gene (Perturbation)") + ylab("Perturbation z-score") + # scale_fill_manual(values=c("grey", "#38b4f7")) + # geom_text(aes(label = significant)) + # theme(legend.position = "none", axis.text.y = element_text(colour = toPlot.extreme$color)) # ## enriched TF motifs # toplot <- all.promoter.ttest.df %>% subset(topic==paste0("topic_",t) & top.gene.mean != 0 & !grepl("X.NA.",motif)) ##here:210816 # p.promoter.motif <- volcano.plot(toplot, ep.type="promoter", ranking.type="z-score") # toplot <- all.enhancer.ttest.df.10en6 %>% subset(topic==paste0("topic_",t) & top.gene.mean != 0 & !grepl("X.NA.",motif)) # p.enhancer.motif <- volcano.plot(toplot, ep.type="enhancer", ranking.type="z-score") # ## old as of 210816 # ## volcano.plot <- function(toplot) { # ## label <- toplot %>% subset(-log10(p.adjust) > 1) %>% mutate(motif.toshow = gsub("HUMAN.H11MO.", "", motif)) # ## t <- gsub("topic_", "", toplot$topic[1]) # ## p <- toplot %>% ggplot(aes(x=enrichment.log2fc, y=-log10(p.adjust))) + geom_point(size=0.5) + mytheme + # ## ggtitle(paste0("Top 100 z-score Specific Promoter Motif Enrichment")) + xlab("Motif Enrichment (log2FC)") + ylab("-log10(adjusted p-value)") + # ## geom_text_repel(data=label, box.padding = 0.5, # ## aes(label=motif.toshow), size=5, # ## color="black") # ## return(p) # ## } # ## p.motif <- volcano.plot(toplot) # ## Factor expression on UMAP # plot.features <- paste0("K",k,"_",colnames(omega)) # feature.name <- plot.features[grepl(paste0("_",t, "$"), plot.features)] ## make sure the seruat object has this feature # if ( grepl(feature.name, colnames(s@meta.data)) %>% as.numeric() %>% sum() > 0 ) { # p.umap <- FeaturePlot(s, reduction = "umap", features=feature.name) # } else { # p.umap <- DimPlot(s, reduction = "umap") # } # p <- ggarrange(ggarrange(p4, p5, p.ptb.list, p.ptb.zscore, nrow=1), ggarrange(p.promoter.motif, p.enhancer.motif, p.umap, nrow=1), nrow=2) # ## p.left <- ggarrange(ggarrange(p4, p5, nrow=1), ggarrange(p.motif, p.umap, nrow=1), nrow=2) # ## p <- ggarrange(p.left, p.ptb.list, nrow=1, width=c(2.5,1)) # p <- annotate_figure(p, top = text_grob(paste0("K = ", k, ", Factor ", t), face = "bold", size = 16)) # ## pdf(paste0(figure.path, ".pdf"), width=10, height=12) # ## print(p) # ## dev.off() # png(paste0(figure.path, ".png"), width=1600, height=1200) # print(p) # dev.off() # } # ## ## convert the original pdf to png # ## im.convert(pdf.path, output = paste0(pdf.path, ".png"), extra.opts="-density 100") ## takes > 10 minutes # } # ## java options for allocating more memory to write Excel sheet # options(java.parameters = "-Xmx16000m") ## 16 GB # ## make an Excel file that has (topic x top gene x GeneCard summaries) and (topic x top perturbations x GeneCard summaries) # ## load gene summary files # theta.zscore.long <- theta.zscore %>% as.data.frame %>% mutate(Gene = rownames(.)) %>% melt(value.name = "score", id.vars="Gene", variable.name = "Factor") # ## theta.zscore.annotation <- merge(theta.zscore.long %>% group_by(Factor) %>% arrange(desc(score)) %>% slice(1:100), gene.summary) # ann_omega_long <- gene.score %>% as.data.frame %>% mutate(Gene=rownames(.)) %>% melt(value.name = "log2FC", id.vars = "Gene", variable.name = "Factor") # ## ann.omega.long <- sqldf("select ann_omega_long.*, summaries.* from ann_omega_long left join summaries on instr(ann_omega_long.Gene, summaries.Symbol)") ## takes a while # ann.omega.long <- merge(ann_omega_long, summaries, by.x="Gene", by.y="Symbol", all.x=T) %>% merge(., all.test %>% subset(test.type=="per.cell.wilcoxon"), by.x=c("Gene","Factor"), by.y=c("Gene","Topic"), all.x=T) # ann.omega.long.top <- ann.omega.long %>% group_by(Factor) %>% arrange(desc(log2FC)) %>% slice(1:50) %>% as.data.frame # ann.omega.long.bottom <- ann.omega.long %>% group_by(Factor) %>% arrange(log2FC) %>% slice(1:50) %>% as.data.frame # ann.omega.long.sig <- ann.omega.long %>% subset(adjusted.p.value < 0.1) %>% as.data.frame # ann.omega.long.output <- rbind(ann.omega.long.top, ann.omega.long.bottom, ann.omega.long.sig) %>% unique %>% arrange(Factor, desc(log2FC)) %>% relocate(c(adjusted.p.value,p.value,top_class,classes), .after="log2FC") # ## load Gavin's top genes in factors with GeneCards information # theta.zscore.long.output <- read_xlsx(paste0(opt$datadir, "210730_cNMF_topic_model_anal.xlsx"), sheet="top100_annotated") # ## combine topic definition and perturbation information # omega.theta.zscore.topic.analysis <- rbind(ann.omega.long.output %>% select(Factor, classes, top_class) %>% mutate(datasource = "omega"), # theta.zscore.long.output %>% select(Topic, Classes, Top_Class) %>% `colnames<-`(c("Factor", "classes", "top_class")) %>% mutate(datasource = "theta.zscore")) %>% # group_by(Factor, top_class) %>% # mutate(top_class_count = n()) %>% ungroup() %>% # filter(!is.na(top_class)) %>% ## remove rows with NA in column `top_class` # separate_rows(classes, sep=";", convert=T) %>% ## separate classes by ";" and expand the rows # filter(!is.na(classes)) %>% subset(classes != "") %>% # group_by(Factor, classes) %>% # mutate(class_count = n()) %>% ungroup() %>% unique() %>% # arrange(Factor, desc(top_class_count), desc(class_count)) # ## separate top_class and class into two dfs # classes.output <- omega.theta.zscore.topic.analysis %>% select(Factor, classes, class_count) %>% as.data.frame %>% unique %>% arrange(Factor,desc(class_count)) # top_class.output <- omega.theta.zscore.topic.analysis %>% select(Factor, top_class, top_class_count) %>% as.data.frame %>% unique %>% arrange(Factor,desc(top_class_count)) # ## write to xlsx # topic.ptb.summary.xlsx.path <- paste0(OUTDIRSAMPLE, "Significant.or.Top50Ptb_per.cell.wilcoxon.adj.0.1_Summary_", SUBSCRIPT, ".xlsx") # write_xlsx(list(Annotations = ann.omega.long.output, # Factor_pathway_summary = omega.theta.zscore.topic.analysis, # Classes_summary = classes.output, # Top_class_summary = top_class.output), # path=topic.ptb.summary.xlsx.path, col_names=T) # ## heatmap for all classes # classes.heatmap <- classes.output # classes.heatmap$class_count <- as.numeric(classes.heatmap$class_count) # classes.heatmap <- classes.heatmap %>% spread(key = classes, value = class_count, fill = as.numeric(0)) %>% `rownames<-`(gsub("topic", "factor",.$Factor)) %>% select(-Factor) %>% as.matrix # ## plot heatmap # plotHeatmap <- function(mtx){ # heatmap.2( # mtx, # Rowv=T, # Colv=T, # trace='none', # key=T, # col=palette, # labCol=colnames(mtx), # margins=c(15,5), # cex.main=0.1, # cexCol=2.5/(nrow(mtx)^(1/3)), cexRow=1.7/(ncol(mtx)^(1/3)), # main=paste0(SAMPLE, ", K=", k, ", Factor Pathway Enrichment") # ) # } # pdf(paste0(FIGDIRTOP, "factor.pathway.enrichment.pdf"), width= 12, height = 9) # mtx <- classes.heatmap # plotHeatmap(mtx) # mtx <- classes.heatmap %>% apply(2, function(x) x/sum(x)) # plotHeatmap(mtx) # dev.off() # ptb.array <- c("GOSR2", "TP53", "CDKN1A", "EDN1", "NOS3", "FGD6", "ELN") # ptb.array <- c("TP53", "ELN", "PHB", "LRPPRC", "MESDC1") # ## manual QC # pdf(paste0(FIGDIRSAMPLE, "/manual.QC.pdf")) # toPlot <- cell.per.ptb <- ann.omega %>% subset(!grepl("^neg|^safe",Gene)) %>% group_by(Gene) %>% summarize(cell.count = n()) # number of cell per perturbation # p <- toPlot %>% ggplot(aes(x=cell.count)) + stat_ecdf() + mytheme + ggtitle("Number of Cells per Perturbation") + xlab("Number of Cells") + ylab("Fraction of Perturbed Genes") # print(p) # p <- toPlot %>% ggplot(aes(x=cell.count)) + geom_histogram() + mytheme + ggtitle("Number of Cells per Perturbation") + xlab("Number of Cells") + ylab("Number of Perturbed Genes") # print(p) # toPlot <- guide.per.ptb <- ann.omega %>% subset(!grepl("^neg|^safe",Gene)) %>% select(Gene,Guide) %>% unique() %>% group_by(Gene) %>% summarize(guide.count = n()) # p <- toPlot %>% ggplot(aes(x=guide.count)) + stat_ecdf() + mytheme + ggtitle("Number of Guides per Perturbation") + xlab("Number of Guides") + ylab("Fraction of Perturbed Genes") # print(p) # p <- toPlot %>% ggplot(aes(x=guide.count)) + geom_histogram() + mytheme + ggtitle("Number of Guides per Perturbation") + xlab("Number of Guides") + ylab("Number of Perturbed Genes") # print(p) # toPlot <- ann.omega %>% group_by(Guide) %>% summarize(count = n()) # p <- toPlot %>% ggplot(aes(x=count)) + stat_ecdf() + mytheme + ggtitle("Number of Cells per Guide") + xlab("Number of Cells") + ylab("Fraction of Perturbed Guides") # print(p) # p <- toPlot %>% ggplot(aes(x=count)) + stat_ecdf() + mytheme + ggtitle("Number of Cells per Guide") + xlim(0,20) + xlab("Number of Cells") + ylab("Fraction of Perturbed Guides") # print(p) # dev.off() ## cell.QC.data <- merge(cell.per.ptb, guide.per.ptb, by="Gene") # ########################################################################## # ## dotplot of log2FC for perturbation and control # ## CDF of log2FC for perturbation and control # ## per gene and per cell # # create a directory for this type of files (one perturbation per file) # FIGDIR.HERE=paste0(FIGDIRTOP,"log2FC_dotplot_CDF_each.perturbation/") # if(!dir.exists(FIGDIR.HERE)) dir.create(FIGDIR.HERE, recursive=T) # |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 | import numpy as np import pandas as pd # import scipy.sparse as sp # import scanpy as sc import anndata as ad from cnmf import cNMF import argparse import re ## argparse parser = argparse.ArgumentParser() parser.add_argument('--path_to_topics', type=str, help='path to the topic (cNMF directory) to project data on', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes_acrossK') parser.add_argument('--topic_sampleName', type=str, help='sample name for topics to project on, use the same sample name as used for the cNMF directory', default='WeissmanK562gwps') # parser.add_argument('--tpm_counts_path', type=str, help='path to tpm input cell x gene matrix', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes_acrossK/2kG.library_overdispersedGenes/cnmf_tmp/2kG.library_overdispersedGenes.tpm.h5ad') #/scratch/groups/engreitz/Users/kangh/cNMF_pipeline/220505_snakemake_moreK_findK/all_genes/K60/worker0/2kG.library/cnmf_tmp/2kG.library.norm_counts.h5ad') parser.add_argument('--outdir', dest = 'outdir', type=str, help = 'path to output directory', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K10/threshold_0_2/IGVF_format/') parser.add_argument('--k', dest = 'k', type=int, help = 'number of components', default='10') parser.add_argument('--density_threshold', dest = 'density_threshold', type=float, help = 'component spectra clustering threshold, 2 for no filtering, recommend 0_2 (means 0.2)', default="0.2") parser.add_argument('--barcode_dir', dest = 'barcode_dir', type=str, default='/oak/stanford/groups/engreitz/Users/kangh/collab_data/IGVF/mouse_ENCODE_heart/auxiliary_data/snrna/heart_Parse_10x_integrated_metadata.csv', help='Directory to barcodes, require columns CBC and Gene') args = parser.parse_args() # ## sdev for IGVF_b01_LeftCortex, all_genes, K=60 # args.outdir = '/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/IGVF_b01_LeftCortex/K60/threshold_0_2/IGVF_format/' # args.path_to_topics = '/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes_acrossK' # args.topic_sampleName = 'IGVF_b01_LeftCortex' # args.k = 60 # args.density_threshold = 0.2 # args.barcode_dir = '/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_igvf_b01_LeftCortex_data/IGVF_b01_LeftCortex.barcodes.txt' sample = args.topic_sampleName # output_sample = args.output_sampleName # tpm_counts_path = args.tpm_counts_path OUTDIR = args.outdir selected_K = args.k density_threshold = args.density_threshold output_directory = args.path_to_topics run_name = args.topic_sampleName barcode_dir = args.barcode_dir cnmf_obj = cNMF(output_dir=output_directory, name=run_name) usage_norm, gep_scores, gep_tpm, topgenes = cnmf_obj.load_results(K=selected_K, density_threshold=density_threshold) ## load cell barcode inforamtion # barcode_dir = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" if barcode_dir.endswith('csv'): barcodes_df = pd.read_csv(barcode_dir, index_col="CBC") else: barcodes_df = pd.read_csv(barcode_dir, sep="\t", index_col="CBC") ## organize program name programNames = [run_name + "_K" + str(selected_K) + "_" + str(i) for i in usage_norm.columns] programNames_df = pd.DataFrame({"ProgramNames": programNames}, index=programNames) usage_norm.columns = programNames barcodes_df = barcodes_df.loc[usage_norm.index,:] ## sort the barcodes df to match with usage_norm ## create AnnData adata = ad.AnnData( X = usage_norm, obs = barcodes_df, var = programNames_df ) ## save results fileName = OUTDIR + sample + ".k_" + str(selected_K) + ".dt_" + re.sub("[.]", "_", str(density_threshold)) + ".cellxgene.h5ad" adata.write_h5ad(fileName) |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("rename", "dplyr") suppressPackageStartupMessages({ library(optparse) library(dplyr) library(tidyr) library(reshape2) library(ggplot2) library(ggpubr) ## ggarrange library(gplots) ## heatmap.2 library(ggrepel) library(readxl) library(xlsx) ## might not need this package library(writexl) library(org.Hs.eg.db) }) option.list <- list( make_option("--sampleName", type="character", default="2kG.library", help="Name of the sample"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/K25/threshold_0_2/", help="Output directory"), make_option("--scratch.outdir", type="character", default="", help="Scratch space for temporary files"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), ## make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), ## make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--perturbSeq", type="logical", default=TRUE, help="Whether this is a Perturb-seq experiment") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## K562 gwps 2k overdispersed genes ## ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K90/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/" ## opt$K.val <- 90 ## opt$sampleName <- "WeissmanK562gwps" ## opt$perturbSeq <- TRUE ## opt$scratch.outdir <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/K90/analysis/comprehensive_program_summary/" ## opt$barcodeDir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## OUTDIR <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220316_regulator_topic_definition_table/outputs/" ## FIGDIR <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220316_regulator_topic_definition_table/figures/" ## SCRATCHOUTDIR <- "/scratch/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220316_regulator_topic_definition_table/outputs/" OUTDIR <- opt$outdir SCRATCHOUTDIR <- opt$scratch.outidr check.dir <- c(OUTDIR, SCRATCHOUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) mytheme <- theme_classic() + theme(axis.text = element_text(size = 12), axis.title = element_text(size = 16), plot.title = element_text(hjust = 0.5, face = "bold")) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## parameters ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/" OUTDIRSAMPLE <- opt$outdir k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) ## SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## ## load ref table (for Perturbation distance to GWAS loci annotation in Perturb_plus column) ## ref.table <- read.delim(file="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/ref.table.txt", header=T, check.names=F, stringsAsFactors=F) ## ## load test results ## all.test.combined.df <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220312_compare_statistical_test/outputs/all.test.combined.df.txt"), stringsAsFactors=F) ## all.test.MAST.df <- all.test.combined.df %>% subset(test.type == "batch.correction") ## MAST.df.4n.input <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/2kG.library4n3.99x_MAST.txt"), stringsAsFactors=F, check.names=F) ## MAST.df.4n <- MAST.df.4n.input %>% ## remove multiTarget entries ## subset(!grepl("multiTarget", perturbation)) %>% ## group_by(zlm.model.name) %>% ## mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method="fdr")) %>% ## ungroup() %>% ## subset(zlm.model.name == "batch.correction") %>% ## select(-zlm.model.name) %>% ## as.data.frame ## colnames(MAST.df.4n) <- c("Topic", "p.value", "log2FC", "log2FC.ci.hi", "log2fc.ci.lo", "fdr", "Perturbation", "fdr.across.ptb") ## Load Regulator Data if(opt$perturbSeq) { MAST.file.name <- paste0(OUTDIR, "/", SAMPLE, "_MAST_DEtopics.txt") message(paste0("loading ", MAST.file.name)) MAST.df <- read.delim(MAST.file.name, stringsAsFactors=F, check.names=F) %>% rename("Perturbation" = "perturbation") %>% rename("log2FC" = "coef", "log2FC.ci.hi" = ci.hi, "log2FC.ci.lo" = ci.lo, "p.value" = "Pr(>Chisq)") if(grepl("topic", MAST.df$primerid) %>% sum > 0) MAST.df <- MAST.df %>% mutate(ProgramID = paste0("K", k, "_", gsub("topic_", "", primerid))) %>% as.data.frame } ## ## 2n MAST ## file.name <- paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes//Perturb_2kG_dup4/acrossK/aggregated.outputs.findK.perturb-seq.RData") ## load(file.name) ## MAST.df.2n <- MAST.df %>% ## filter(K == 60) %>% ## select(-K) %>% ## select(-zlm.model.name) ## colnames(MAST.df.2n) <- c("Topic", "p.value", "log2FC", "log2FC.ci.hi", "log2fc.ci.lo", "fdr", "Perturbation", "fdr.across.ptb") ## load gene annotations (Refseq + Uniprot) gene.summaries.path <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/combined.gene.summaries.txt" gene.summary <- read.delim(gene.summaries.path, stringsAsFactors=F) ## load topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } ## modify theta.zscore if Gene is in ENSGID db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") gene.ary <- theta.zscore %>% rownames if(grepl("^ENSG", gene.ary) %>% as.numeric %>% sum == nrow(theta.zscore)) { GeneSymbol.ary <- mapIds(get(db), keys=gene.ary, keytype = "ENSEMBL", column = "SYMBOL") GeneSymbol.ary[is.na(GeneSymbol.ary)] <- row.names(theta.zscore)[is.na(GeneSymbol.ary)] rownames(theta.zscore) <- GeneSymbol.ary } ## omega.4n <- omega ## theta.zscore.4n <- theta.zscore ## theta.raw.4n <- theta.raw meta_data <- read.delim(opt$barcodeDir, stringsAsFactors=F) ## ## batch topics ## batch.topics.4n <- read.delim(file=paste0(OUTDIRSAMPLE, "/batch.topics.txt"), stringsAsFactors=F) %>% as.matrix %>% as.character ann.omega <- merge(meta_data, omega, by.x="CBC", by.y=0, all.T) ########################################################################################## ## create table create_topic_definition_table <- function(theta.zscore, t) { out <- theta.zscore[,t] %>% as.data.frame %>% `colnames<-`(c("zscore")) %>% mutate(Perturbation = rownames(theta.zscore), .before="zscore") %>% merge(gene.summary, by.x="Perturbation", by.y="Gene", all.x=T) %>% arrange(desc(zscore)) %>% mutate(Rank = 1:n(), .before="Perturbation") %>% mutate(ProgramID = paste0("K", k, "_", t), .before="zscore") %>% arrange(Rank) %>% mutate(My_summary = "", .after = "zscore") %>% select(Rank, ProgramID, Perturbation, zscore, My_summary, FullName, Summary) } create_topic_regulator_table <- function(all.test, program.here, fdr.thr = 0.1) { out <- MAST.df %>% subset(ProgramID == program.here & fdr.across.ptb < fdr.thr) %>% select(Perturbation, fdr.across.ptb, log2FC, log2FC.ci.hi, log2FC.ci.lo, fdr, p.value) %>% merge(gene.summary, by.x="Perturbation", by.y="Gene", all.x=T) %>% arrange(fdr.across.ptb, desc(log2FC)) %>% mutate(Rank = 1:n(), .before="Perturbation") %>% mutate(My_summary = "", .after="Perturbation") %>% mutate(ProgramID = program.here, .after="Rank") %>% ## merge(., ref.table %>% select("Symbol", "TSS.dist.to.SNP", "GWAS.classification"), by.x="Perturbation", by.y="Symbol", all.x=T) %>% ## mutate(EC_ctrl_text = ifelse(.$GWAS.classification == "EC_ctrls", "(+)", "")) %>% ## mutate(GWAS.class.text = ifelse(grepl("CAD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb"), ## ifelse(grepl("IBD", GWAS.classification), paste0("_", floor(TSS.dist.to.SNP/1000),"kb_IBD"), ""))) %>% ## mutate(Perturb_plus = paste0(Perturbation, GWAS.class.text, EC_ctrl_text)) %>% select(Rank, ProgramID, Perturbation, fdr.across.ptb, log2FC, My_summary, FullName, Summary, log2FC.ci.hi, log2FC.ci.lo, fdr, p.value) %>% ## removed Perturb_plus arrange(Rank) } create_summary_table <- function(ann.omega, theta.zscore, all.test, meta_data) { df.list <- vector("list", k) for (t in 1:k) { program.here <- paste0("K", k, "_", t) ## topic defining genes ann.theta.zscore <- theta.zscore %>% create_topic_definition_table(t) ann.top.theta.zscore <- ann.theta.zscore %>% subset(Rank <= 100) ## select the top 100 topic defining genes to output ## regulators if(opt$perturbSeq) regulator.MAST.df <- all.test %>% create_topic_regulator_table(program.here, 0.3) ## write table to scratch dir file.name <- paste0(SCRATCHOUTDIR, program.here, "_table.csv") sink(file=file.name) ## open the document ## cat("Author,PERTURBATIONS SUMMARIES\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\n\n\nAuthor,TOPIC SUMMARIES\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\nAuthor,TESTABLE HYPOTHESIS IDEAS:\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\nAuthor,OTHER THOUGHTS:\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\nTOPIC DEFINING GENES (TOP 100),\n") ## headers cat("Author,PERTURBATIONS SUMMARIES,,,,,,,,,,,\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\"\nAuthor,TOPIC SUMMARIES\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\"\nAuthor,TESTABLE HYPOTHESIS IDEAS:\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\"\nAuthor,OTHER THOUGHTS:\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\"\n\nTOPIC DEFINING GENES (TOP 100),\n") ## headers write.csv(ann.top.theta.zscore, row.names=F) ## topic defining genes cat("\n\"\n\"\nPERTURBATIONS REGULATING TOPIC AT FDR < 0.3 (most significant on top),,,,,,,,,,,\n") ## headers if(opt$perturbSeq) write.csv(regulator.MAST.df, row.names=F) ## regulators sink() ## close the document ## read the assembled table to save to list df.list[[t]] <- read.delim(file.name, stringsAsFactors=F, check.names=F, sep=",") } ## output to xlsx names(df.list) <- paste0("Program ", 1:k) write_xlsx(df.list, paste0(OUTDIR, "/", SAMPLE, "_k_", k, ".dt_", DENSITY.THRESHOLD, "_ComprehensiveProgramSummary.xlsx")) return(df.list) } if(opt$perturbSeq == "F") MAST.df <- data.frame() df <- create_summary_table(ann.omega, theta.zscore, MAST.df, meta_data) |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 | suppressPackageStartupMessages(library(conflicted)) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("first", "dplyr") suppressPackageStartupMessages({ library(optparse) library(dplyr) library(tidyr) library(reshape2) ## library(ggplot2) ## library(cowplot) ## library(ggpubr) ## ggarrange ## library(gplots) ## heatmap.2 ## library(scales) ## geom_tile gradient rescale ## library(ggrepel) library(stringr) library(stringi) library(svglite) library(Seurat) library(SeuratObject) library(xlsx) }) ########################################################################################## ## Constants and Directories option.list <- list( make_option("--sampleName", type="character", default="2kG.library", help="Name of the sample"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/K25/threshold_0_2/", help="Output directory"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), make_option("--perturbSeq", type="logical", default=TRUE, help="Whether this is a Perturb-seq experiment") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## K562 gwps 2k overdispersed genes ## ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K90/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/" ## opt$K.val <- 90 ## opt$sampleName <- "WeissmanK562gwps" ## opt$perturbSeq <- TRUE ## ## ENCODE mouse heart ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes/mouse_ENCODE_heart/K10/threshold_0_2/" ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$K.val <- 10 ## opt$perturbSeq <- FALSE OUTDIR <- opt$outdir SAMPLE <- opt$sampleName k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) check.dir <- c(OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) fdr.thr <- 0.05 db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") suppressPackageStartupMessages(library(!!db)) ## load the appropriate database ## Load Data ## Batch Correlation table load(paste0(OUTDIR, "/batch.correlation.RDS")) if("topic" %in% colnames(max.batch.correlation.df)){ max.batch.correlation.df <- max.batch.correlation.df %>% mutate(ProgramID = paste0("K", k, "_", gsub("topic_", "", topic))) %>% as.data.frame } if ("ProgramID" %in% colnames(max.batch.correlation.df)) { max.batch.correlation.df <- max.batch.correlation.df %>% mutate(ProgramID = paste0("K", k, "_", gsub("topic_", "", ProgramID))) %>% as.data.frame } ################################################## ## load cNMF results cNMF.result.file <- paste0(OUTDIR,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") if(file.exists(cNMF.result.file)) { message(paste0("loading cNMF result file: \n", cNMF.result.file)) load(cNMF.result.file) } else { print(paste0(cNMF.result.file, " does not exist")) } ## helper function to map between ENSGID and SYMBOL map.ENSGID.SYMBOL <- function(df) { ## need column `Gene` to be present in df ## detect gene data type (e.g. ENSGID, Entrez Symbol) if(!("Gene" %in% colnames(df))) df$Gene = df$ENSGID gene.type <- ifelse(nrow(df) == sum(as.numeric(grepl("^ENS", df$Gene))), "ENSGID", "Gene") if(gene.type == "ENSGID") { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "ENSEMBL", column = "SYMBOL") df <- df %>% mutate(ENSGID = Gene, Gene = mapped.genes) } else { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "SYMBOL", column = "ENSEMBL") df <- df %>% mutate(ENSGID = mapped.genes) } df <- df %>% mutate(Gene = ifelse(is.na(Gene), ENSGID, Gene)) return(df) } if(sum(c("ENSGID", "Gene") %in% colnames(median.spectra.zscore.df)) < 2) { median.spectra.zscore.df <- median.spectra.zscore.df %>% map.ENSGID.SYMBOL } ## get list of topic defining genes by z-score coefficients if(!("theta.zscore.rank.df" %in% ls())) { theta.zscore.rank.list <- vector("list", ncol(theta.zscore))## initialize storage list for(i in 1:ncol(theta.zscore)) { topic <- paste0("topic_", colnames(theta.zscore)[i]) theta.zscore.rank.list[[i]] <- theta.zscore %>% as.data.frame %>% select(all_of(i)) %>%##here `colnames<-`("topic.zscore") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.zscore), .before="topic.zscore") %>% mutate(zscore.specificity.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.zscore.rank.df <- do.call(rbind, theta.zscore.rank.list) %>% ## combine list to df `colnames<-`(c("topic.zscore", "Gene", "zscore.specificity.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL } if(!("theta.tpm.rank.df" %in% ls())) { theta.tpm.rank.list <- vector("list", ncol(theta.raw))## initialize storage list for(i in 1:ncol(theta.zscore)) { topic <- paste0("topic_", colnames(theta.raw)[i]) theta.tpm.rank.list[[i]] <- theta %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("topic.zscore") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.zscore), .before="topic.zscore") %>% mutate(zscore.specificity.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.tpm.rank.df <- do.call(rbind, theta.tpm.rank.list) %>% ## combine list to df `colnames<-`(c("program.tpm.coef", "Gene", "tpm.coef.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL %>% mutate(Gene = ifelse(is.na(Gene), ENSGID, Gene)) } ## get list of topic genes by median spectra weight if(!("median.spectra.rank.df" %in% ls())) { median.spectra.rank.list <- vector("list", ncol(median.spectra))## initialize storage list for(i in 1:ncol(median.spectra)) { topic <- paste0("topic_", colnames(median.spectra)[i]) median.spectra.rank.list[[i]] <- median.spectra %>% as.data.frame %>% select(all_of(i)) %>% `colnames<-`("median.spectra") %>% mutate(Gene = rownames(.)) %>% arrange(desc(median.spectra), .before="median.spectra") %>% mutate(median.spectra.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } median.spectra.rank.df <- do.call(rbind, median.spectra.rank.list) %>% ## combine list to df `colnames<-`(c("median.spectra", "Gene", "median.spectra.rank", "ProgramID")) %>% mutate(ProgramID = gsub("topic_", paste0("K", k, "_"), ProgramID)) %>% as.data.frame %>% map.ENSGID.SYMBOL %>% mutate(Gene = ifelse(is.na(Gene), ENSGID, Gene)) ## median.spectra.zscore.df <- median.spectra.zscore.df %>% mutate(Gene = ENSGID) ## quick fix, need to add "Gene" column to this dataframe in analysis script } ## Load Regulator Data if(opt$perturbSeq) { MAST.file.name <- paste0(OUTDIR, "/", SAMPLE, "_MAST_DEtopics.txt") message(paste0("loading ", MAST.file.name)) MAST.df <- read.delim(MAST.file.name, stringsAsFactors=F, check.names=F) if(grepl("topic", MAST.df$primerid) %>% sum > 0) MAST.df <- MAST.df %>% mutate(ProgramID = paste0("K", k, "_", gsub("topic_", "", primerid))) %>% as.data.frame } ## Load Promoter and Enhancer TF Motif Enrichment Data add.ProgramID <- function(df) { if("topic" %in% c(df %>% colnames)) { return(df %>% mutate(ProgramID = paste0('K', k, '_', gsub('topic_', '', topic))) %>% as.data.frame) } } num.top.genes <- 300 ## all.ttest.df.path <- paste0(OUTDIRSAMPLE,"/", ep.type, ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.", motif.match.thr.str, "_", SUBSCRIPT.SHORT,".txt") promoter.ttest.df.path <- paste0(OUTDIR,"/", "promoter", ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.", "pval1e-4", "_", SUBSCRIPT.SHORT,".txt") enhancer.ttest.df.path <- paste0(OUTDIR,"/", "promoter", ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.", "pval1e-6", "_", SUBSCRIPT.SHORT,".txt") promoter.ttest.df <- read.delim(promoter.ttest.df.path, stringsAsFactors=F) %>% add.ProgramID enhancer.ttest.df <- read.delim(enhancer.ttest.df.path, stringsAsFactors=F) %>% add.ProgramID ## GO terms ## file.name <- paste0(OUTDIRSAMPLE, "/clusterProfiler_GeneRankingType", ranking.type.here, "_EnrichmentType", GSEA.type,".txt") file.name <- paste0(OUTDIR, "/clusterProfiler_GeneRankingType", "zscore", "_EnrichmentType", "GOEnrichment",".txt") theta.zscore.GO.df <- read.delim(file.name, stringsAsFactors=F) file.name <- paste0(OUTDIR, "/clusterProfiler_GeneRankingType", "median_spectra_zscore", "_EnrichmentType", "GOEnrichment",".txt") median_spectra_zscore.GO.df <- read.delim(file.name, stringsAsFactors=F) #################################################################################################### ## Gather all columns ## MaxBatchCorrelation MaxBatchCorrelation <- max.batch.correlation.df %>% select(ProgramID, maxPearsonCorrelation) %>% `colnames<-`(c("ProgramID", "MaxBatchCorrelation")) %>% as.data.frame ## nSigPerturbationsProgramUp SigPerturbations.df <- MAST.df %>% subset(fdr.across.ptb < fdr.thr) %>% as.data.frame nSigPerturbationsProgramUp <- SigPerturbations.df %>% subset(coef > log(1.1)) %>% group_by(ProgramID) %>% summarize(nSigPerturbationsProgramUp = n()) %>% as.data.frame nSigPerturbationsProgramDown <- SigPerturbations.df %>% subset(coef < log(0.9)) %>% group_by(ProgramID) %>% summarize(nSigPerturbationsProgramDown = n()) %>% as.data.frame ## nSigMotifsPromoter nSigMotifsPromoter <- promoter.ttest.df %>% subset(one.sided.p.adjust < fdr.thr) %>% group_by(ProgramID) %>% summarize(nSigMotifsPromoter = n()) %>% as.data.frame ## nSigMotifsEnhancer nSigMotifsEnhancer <- enhancer.ttest.df %>% subset(one.sided.p.adjust < fdr.thr) %>% group_by(ProgramID) %>% summarize(nSigMotifsEnhancer = n()) %>% as.data.frame ## ProgramGenesZScoreCoefficientTop10 ProgramGenesZScoreCoefficientTop10 <- theta.zscore.rank.df %>% mutate(Gene = ifelse(is.na(Gene), ENSGID, Gene)) %>% subset(zscore.specificity.rank < 10) %>% group_by(ProgramID) %>% arrange(desc(topic.zscore)) %>% summarize(ProgramGenesZScoreCoefficientTop10 = paste0(Gene, collapse=",")) %>% as.data.frame ## ProgramGenesTPMCoefficientTop10 ProgramGenesTPMCoefficientTop10 <- theta.tpm.rank.df %>% subset(tpm.coef.rank < 10) %>% group_by(ProgramID) %>% arrange(desc(program.tpm.coef)) %>% summarize(ProgramGenesTPMCoefficientTop10 = paste0(Gene, collapse=",")) %>% as.data.frame ## ProgramGenesMedianSpectraTop10 ProgramGenesMedianSpectraTop10 <- median.spectra.rank.df %>% subset(median.spectra.rank < 10) %>% mutate(Gene = ifelse(is.na(Gene), ENSGID, Gene)) %>% group_by(ProgramID) %>% arrange(desc(median.spectra)) %>% summarize(ProgramGenesMedianSpectraTop10 = paste0(Gene, collapse = ",")) %>% as.data.frame ## ProgramGenesMedianSpectraZScoreTop10 ProgramGenesMedianSpectraZScoreTop10 <- median.spectra.zscore.df %>% subset(median.spectra.zscore.rank < 10) %>% group_by(ProgramID) %>% arrange(desc(median.spectra.zscore)) %>% summarize(ProgramGenesMedianSpectraZScoreTop10 = paste0(Gene, collapse = ",")) %>% as.data.frame ## ProgramGenesMotifsPromoter ProgramGenesMotifsPromoter <- promoter.ttest.df %>% subset(one.sided.p.adjust < fdr.thr) %>% group_by(ProgramID) %>% arrange(one.sided.p.adjust) %>% summarize(ProgramGenesMotifsPromoter = paste0(motif, collapse = ",")) %>% as.data.frame ## ProgramGenesMotifsEnhancer ProgramGenesMotifsEnhancer <- enhancer.ttest.df %>% subset(one.sided.p.adjust < fdr.thr) %>% group_by(ProgramID) %>% arrange(one.sided.p.adjust) %>% summarize(ProgramGenesMotifsEnhancer = paste0(motif, collapse = ",")) %>% as.data.frame ## ProgramGenesZScoreCoefficientGOTermsTop10 ProgramGenesZScoreCoefficientGOTermsTop10 <- theta.zscore.GO.df %>% group_by(ProgramID) %>% arrange(fdr.across.ont) %>% slice(1:10) %>% mutate(GOTerm = paste0(ONTOLOGY, ":", ID, ":", Description)) %>% summarize(ProgramGenesZScoreCoefficientGOTermsTop10 = paste0(GOTerm, collapse = ",")) %>% as.data.frame ## ProgramGenesMedianSpectraZScoreGOTermsTop10 ProgramGenesMedianSpectraZScoreGOTermsTop10 <- median_spectra_zscore.GO.df %>% group_by(ProgramID) %>% arrange(fdr.across.ont) %>% slice(1:10) %>% mutate(GOTerm = paste0(ONTOLOGY, ":", ID, ":", Description)) %>% summarize(ProgramGenesMedianSpectraZScoreGOTermsTop10 = paste0(GOTerm, collapse = ",")) %>% as.data.frame ## ProgramGenesTF* ## Combine all columns ProgramSummary.list <- list( MaxBatchCorrelation = MaxBatchCorrelation, nSigPerturbationsProgramUp = nSigPerturbationsProgramUp, nSigPerturbationsProgramDown = nSigPerturbationsProgramDown, nSigMotifsPromoter = nSigMotifsPromoter, nSigMotifsEnhancer = nSigMotifsEnhancer, ProgramGenesZScoreCoefficientTop10 = ProgramGenesZScoreCoefficientTop10, ProgramGenesTPMCoefficientTop10 = ProgramGenesTPMCoefficientTop10, ProgramGenesMedianSpectraTop10 = ProgramGenesMedianSpectraTop10, ProgramGenesMedianSpectraZScoreTop10 = ProgramGenesMedianSpectraZScoreTop10, ProgramGenesMotifsPromoter = ProgramGenesMotifsPromoter, ProgramGenesMotifsEnhancer = ProgramGenesMotifsEnhancer, ProgramGenesZScoreCoefficientGOTermsTop10 = ProgramGenesZScoreCoefficientGOTermsTop10, ProgramGenesMedianSpectraZScoreGOTermsTop10 = ProgramGenesMedianSpectraZScoreGOTermsTop10 ) ProgramSummary.df <- Reduce(function(x, y, ...) full_join(x, y, by = "ProgramID", ...), ProgramSummary.list) ## ## Write Table to Text File fileName <- paste0(SAMPLE, "_ProgramSummary_", SUBSCRIPT.SHORT) write.table(ProgramSummary.df, file=paste0(OUTDIR, "/", fileName, ".xlsx"), row.names=F, quote=F, sep="\t") ## Write Excel File write.xlsx(ProgramSummary.df, file=paste0(OUTDIR, "/", fileName, ".xlsx"), sheetName="Gene Selection Table (For experimental design)", row.names=F, showNA=F) |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | import os import pandas as pd import numpy as np from scipy.io import mmread import scipy.sparse as sp # import matplotlib.pyplot as plt #from IPython.display import Image import scanpy as sc import argparse parser = argparse.ArgumentParser() parser.add_argument('--inputPath', dest = 'inputPath', type=str, default='/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.txt', help = 'path to count matrix file') # parser.add_argument('--outPath', dest = 'outPath', type=str, help = 'path to output folder') # parser.add_argument('--run_name',dest ='run_name', type=str, help = 'sample name') parser.add_argument('--output_h5ad_mtx', dest = 'output_h5ad_mtx', type=str, help = 'path to output folder', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/211014_test_txt_to_h5ad/outputs/test.h5ad') parser.add_argument('--output_gene_name_txt', dest = 'output_gene_name_txt', type=str, help = 'path to gene name output txt file', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/211014_test_txt_to_h5ad/outputs/test.h5ad.all.genes.txt') args = parser.parse_args() # ## 200 gene library 230610, no IL1B # args.inputPath = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/no_IL1B.raw.h5ad' # args.output_h5ad_mtx = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/no_IL1B.h5ad' # args.output_gene_name_txt = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/no_IL1B.h5ad.all.genes.txt' # ## 200 gene library 230610, plus IL1B # args.inputPath = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/plus_IL1B.raw.h5ad' # args.output_h5ad_mtx = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/plus_IL1B.h5ad' # args.output_gene_name_txt = '/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/plus_IL1B.h5ad.all.genes.txt' ## load scanpy from txt adata = sc.read(args.inputPath) # remove non protein coding genes df_columns = pd.Series(adata.var_names.values.flatten()).astype(str) AL = df_columns[df_columns.str.contains('^AL[0-9][0-9][0-9][0-9][0-9][0-9]\\.').tolist()] AC = df_columns[df_columns.str.contains('^AC[0-9][0-9][0-9][0-9][0-9][0-9]\\.').tolist()] AP = df_columns[df_columns.str.contains('^AP[0-9][0-9][0-9][0-9][0-9][0-9]\\.').tolist()] LINC = df_columns[df_columns.str.contains('LINC').tolist()] allPattern = df_columns[df_columns.str.contains('^[A-Za-z][A-Za-z][0-9][0-9][0-9][0-9][0-9][0-9]\\.').tolist()] toremove = allPattern.append(LINC) tokeep = ~df_columns.isin(toremove) adata = adata[:,tokeep] # filter cells sc.pp.filter_cells(adata, min_genes=200) # filter cells with fewer than 200 genes sc.pp.filter_cells(adata, min_counts=200) # This is a weaker threshold than above. It is just to population the n_counts column in adata sc.pp.filter_genes(adata, min_cells=10) # filter genes detected in fewer than 3 cells # save to h5ad file sc.write(args.output_h5ad_mtx, adata) # get gene names after filtering filtered_genes = pd.DataFrame(adata.var_names.values.flatten()).astype(str) # save gene names filtered_genes.to_csv(args.output_gene_name_txt, sep="\t", header=False, index=False) |
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 | COORD=$1 FASTA=$2 OUTFASTA=$3 # PROJECT=$OAK/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/ # DATADIR=$PROJECT/data/ # FILEDIR=$OAK/Data/hg38 # OUTDIR=${PROJECT}/outputs/ # ABCDIR=$OAK/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/ # TOPDATADIR=$OAK/Users/kangh/2009_endothelial_perturbseq_analysis/data/ # TOPDATADIRABC=$OAK/Users/kangh/2009_endothelial_perturbseq_analysis/data/ABC/ # SCRATCHDIR=${SCRATCH}/210509_topic_motif_enrichment/ # mkdir -p $OUTDIR # mkdir -p $DATADIR # mkdir -p $TOPDATADIRABC # mkdir -p $SCRATCHDIR # LOG=${PROJECT}/logs/ # mkdir -p $LOG # QSUB=/home/groups/engreitz/bin/quick-sub # hg38FASTA=/oak/stanford/groups/engreitz/Data/hg38/Sequence/hg38.fa # hg19FASTA=/oak/stanford/groups/engreitz/Data/hg19/Sequence/hg19.fa ############################################################ ## get fasta for the enhancer regions # chr, start, end, name, class, activity_base, TargetGene, TargetGeneTSS, TargetGeneExpression, TargetGenePromoterActivityQuantile, TargetGeneIsExpressed, distance, isSelfPromoter, powerlaw_contact, powerlaw_contact_reference, hic_contact, hic_contact_pl_scaled, hic_pseudocount, hic_contact_pl_scaled_adj, ABC.Score.Numerator, ABC.Score, powerlaw.Score.Numerator, powerlaw.Score, CellType bedtools getfasta -name -fi ${FASTA} -bed <(awk 'OFS="\t" {print $1,$2,$3,$1":"$2"-"$3"|"$4"|"$7}' ${COORD}) -fo ${OUTFASTA} # bedtools getfasta -name -fi ${hg19FASTA} -bed <(awk 'OFS="\t" {print $1,$2,$3,$1":"$2"-"$3"|"$4"|"$7}' ${COORD}) -fo ${TOPDATADIRABC}/${sample}_Predictions.AvgHiC.ABC0.015.minus150.fa |
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("first", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("filter", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr") #, "grid", "gtable", "gridExtra","ggrepel",#"ramify", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ########################################################################################## ## Constants and Directories option.list <- list( make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--sampleName", type="character", default="2kG.library", help="sample name"), make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/2kG.library.barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), ## make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), ## make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--outdirsample", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/", help="path to cNMF analysis results"), ## or for 2n1.99x: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/K60/threshold_0_2/" make_option("--scatteroutput", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/K35/threshold_0_2/", help="path to gene breakdown table output"), make_option("--total.scatter.gene.group", type="numeric", default=50, help="Total number of groups for running MAST"), ## script dir make_option("--scriptdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts/", help="location for this script and functions script") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev debug K562 gwps ## opt$sampleName <- "WeissmanK562gwps" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K35/threshold_0_2/" ## opt$scatteroutput <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/WeissmanK562gwps/MAST/K35/threshold_0_2/" ## opt$total.scatter.gene.group <- 494 ## opt$scriptdir <- "/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts" ## opt$K.val <- 35 k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) ## SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) OUTDIRSAMPLE <- opt$outdirsample SCATTEROUTDIR <- opt$scatteroutput NUMGROUPS <- opt$total.scatter.gene.group OUTDIR <- OUTDIRSAMPLE check.dir <- c(OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## load data MAST.df.list <- vector("list", NUMGROUPS) for (i in 1:NUMGROUPS) { MAST.df.list[[i]] <- read.delim(paste0(SCATTEROUTDIR, "/", SAMPLE, "_MAST_DEtopics_Group", i, ".txt"), stringsAsFactors=F, check.names=F) } MAST.df <- do.call(rbind, MAST.df.list) %>% group_by(zlm.model.name) %>% mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method='fdr')) %>% as.data.frame write.table(MAST.df, file=paste0(OUTDIRSAMPLE, "/", SAMPLE, "_MAST_DEtopics.txt"), quote=F, row.names=F, sep="\t") |
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("filter", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel",#"ramify", "ggpubr","gridExtra", "parallel", "future", "org.Hs.eg.db","limma","conflicted", #"fgsea", "cluster","textshape","readxl", "ggdist", "gghalves", "Seurat", "writexl", "SingleCellExperiment", "MAST") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ########################################################################################## ## Constants and Directories option.list <- list( make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--sampleName", type="character", default="2kG.library", help="sample name"), make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/2kG.library.barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), ## make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), ## make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--outdirsample", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/", help="path to cNMF analysis results"), ## or for 2n1.99x: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/K60/threshold_0_2/" make_option("--scatteroutput", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/", help="path to gene breakdown table output"), make_option("--numCtrl", type="numeric", default=5000, help="number of control cells to use for MAST") ## ## script dir ## make_option("--scriptdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts/", help="location for this script and functions script") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev debug K562 gwps ## opt$sampleName <- "WeissmanK562gwps" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K35/threshold_0_2/" ## opt$scatteroutput <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/K35/threshold_0_2/" ## opt$scatter.gene.group <- 496 ## opt$scriptdir <- "/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts" ## opt$K.val <- 35 k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) ## SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K31/threshold_0_2/" OUTDIRSAMPLE <- opt$outdirsample SCATTEROUTDIR <- opt$scatteroutput SCATTERINDEX <- opt$scatter.gene.group INPUTDIR <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/inputs/" OUTDIR <- OUTDIRSAMPLE check.dir <- c(INPUTDIR, OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ########################################################################################## ## load data ## load topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } barcode.names <- read.delim(opt$barcode.names, stringsAsFactors=F) ## separate out control cells omega.ctrl.index <- barcode.names %>% mutate(rowindex = 1:n()) %>% filter(Gene == "negative-control") %>% pull(rowindex) omega.ptb.index <- barcode.names %>% mutate(rowindex = 1:n()) %>% filter(Gene != "negative-control") %>% pull(rowindex) omega.ctrl <- omega[omega.ctrl.index,] omega.ptb <- omega[omega.ptb.index,] barcode.names.ctrl <- barcode.names[omega.ctrl.index,] barcode.names.ptb <- barcode.names[omega.ptb.index,] ## randomly subset to 5000 cells ctrl.subset.index <- sample(1:length(omega.ctrl.index), min(length(omega.ctrl.index), opt$numCtrl), replace=FALSE) omega.ctrl.subset <- omega.ctrl[ctrl.subset.index,] barcode.names.ctrl.subset <- barcode.names.ctrl[ctrl.subset.index,] omega.new <- rbind(omega.ptb, omega.ctrl.subset) barcode.names.subset <- rbind(barcode.names.ptb, barcode.names.ctrl.subset) omega.tpm <- omega.new %>% as.matrix %>% apply(2, function(x) x / sum(x) * 1000000) ## convert to TPM log2.omega <- (omega.tpm + 1) %>% log2 ## log2(TPM + 1) ## output log2(TPM + 1) save(log2.omega, barcode.names.subset, file=paste0(opt$scatteroutput, "/", SAMPLE, "_MAST_log2TPM_barcodes.RDS")) |
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("filter", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel",#"ramify", "ggpubr","gridExtra", "parallel", "future", "org.Hs.eg.db","limma","conflicted", #"fgsea", "cluster","textshape","readxl", "ggdist", "gghalves", "Seurat", "writexl", "SingleCellExperiment", "MAST") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ########################################################################################## ## Constants and Directories option.list <- list( make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--sampleName", type="character", default="2kG.library", help="sample name"), make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/2kG.library.barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--outdirsample", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/", help="path to cNMF analysis results"), ## or for 2n1.99x: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/K60/threshold_0_2/" make_option("--scatteroutput", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/K80/threshold_0_2/", help="path to gene breakdown table output"), make_option("--gene.group.list", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/GeneNames_Group43.txt"), make_option("--scatter.gene.group", type="numeric", default=50, help="Gene group index"), ## script dir make_option("--scriptdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts/", help="location for this script and functions script") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev debug K562 gwps ## opt$sampleName <- "WeissmanK562gwps" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K35/threshold_0_2/" ## opt$scatteroutput <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/WeissmanK562gwps/MAST/K80/threshold_0_2/" ## opt$gene.group.list <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/GeneNames_Group43.txt" ## opt$scatter.gene.group <- 43 ## opt$scriptdir <- "/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts" ## opt$K.val <- 80 ## ## no IL1B ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/data/no_IL1B.barcodes.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/analysis/top2000VariableGenes/no_IL1B/K15/threshold_0_2/" ## opt$scatteroutput <- "/scratch/groups/engreitz/Users/kangh/cNMF_pipeline/tutorials/2306_V2G2P_prep/top2000VariableGenes/no_IL1B/MAST/K15/threshold_0_2/" ## opt$gene.group.list <- "/scratch/groups/engreitz/Users/kangh/cNMF_pipeline/tutorials/2306_V2G2P_prep/top2000VariableGenes/no_IL1B/MAST/GeneNames_Group12.txt" ## opt$scatter.gene.group <- 12 ## opt$sampleName <- "no_IL1B" ## opt$K.val <- 15 ## opt$density.thr <- 0.2 ## opt$scriptdir <- "/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts" k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K31/threshold_0_2/" OUTDIRSAMPLE <- opt$outdirsample SCATTEROUTDIR <- opt$scatteroutput SCATTERINDEX <- opt$scatter.gene.group INPUTDIR <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/inputs/" OUTDIR <- OUTDIRSAMPLE check.dir <- c(INPUTDIR, OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## mytheme <- theme_classic() + theme(axis.text = element_text(size = 12), axis.title = element_text(size = 16), plot.title = element_text(hjust = 0.5, face = "bold")) ## palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## p.adjust.thr <- 0.1 ########################################################################################## ## load data ## ## load known CAD genes ## CAD.genes <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/known_CAD_gene_set.txt", header=F, stringsAsFactors=F) %>% as.matrix %>% as.character ## ## load topic model results ## cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") ## print(cNMF.result.file) ## if(file.exists(cNMF.result.file)) { ## print("loading cNMF result file") ## load(cNMF.result.file) ## } ## file.name <- paste0(OUTDIRSAMPLE,"/cNMFAnalysis.",SUBSCRIPT,".RData") ## print(file.name) ## if(file.exists((file.name))) { ## print(paste0("loading ",file.name)) ## load(file.name) ## } ## ## add UMI / cell information ## umiPerCell.df <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/outputs/UMIPerCell.txt", stringsAsFactors=F, row.names=1) %>% mutate(long.CBC = rownames(.)) %>% ## separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% ## separate(col="CBC", into=c("CBC", "sample"), sep="-scRNAseq_2kG_", remove=F) ## sample.to.10X.lane <- data.frame(sample_num = 1:20, ## sample = umiPerCell.df$sample %>% unique %>% sort) ## umiPerCell.df <- merge(umiPerCell.df, sample.to.10X.lane, by="sample") %>% ## mutate(CBC_10x = paste0(CBC, "-", sample_num)) ## ## add guide / cell information ## guidePerCell.df <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/inputs/_UPDATED_ALL_SAMPLES_dup4_NON_NA_CALLS_FOR_EA_CBC_FULL_INFO.txt", stringsAsFactors=F) ## ## add genes detected / cell information ## geneDetectedPerCell.df <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/outputs/geneDetectedPerCell.txt", stringsAsFactors=F) %>% ## mutate(long.CBC = rownames(.)) %>% ## separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% ## separate(col="CBC", into=c("CBC", "sample"), sep="-scRNAseq_2kG_", remove=F) %>% ## merge(sample.to.10X.lane, by="sample") %>% ## mutate(CBC_10x = paste0(CBC, "-", sample_num)) ## load gene list gene.ary <- read.delim(opt$gene.group.list, header=F, stringsAsFactors=F) %>% unlist %>% as.character ## load log2(TPM + 1) load(paste0(opt$scatteroutput, "/", SAMPLE, "_MAST_log2TPM_barcodes.RDS")) print("organize meta data") ## get metadata from log2.X.full rownames # if(SAMPLE %in% c("2kG.library", "Perturb_2kG_dup4")) { if( grepl("2kG.library|Perturb_2kG_dup4", SAMPLE) ) { barcode.names <- read.table(opt$barcode.names, header=F, stringsAsFactors=F) %>% `colnames<-`("long.CBC") ## rownames(omega) <- barcode.names %>% pull(long.CBC) %>% gsub("CSNK2B-and-CSNK2B", "CSNK2B",.) meta_data <- barcode.names %>% rownames %>% as.data.frame %>% `colnames<-`("long.CBC") %>% separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% ## separate(col="CBC", into=c("CBC", "sample"), sep="-", remove=F) %>% separate(col="CBC", into=c("CBC", "sample"), sep="-scRNAseq_2kG_", remove=F) %>% mutate(Gene = gsub("-TSS2", "", Gene.full.name), CBC = gsub("RHOA-and-", "", CBC)) %>% filter(Gene %in% c(gene.ary, "negative-control", "safe-targeting")) %>% as.data.frame sample.to.10X.lane <- data.frame(sample_num = 1:20, sample = meta_data$sample %>% unique %>% sort) ## meta_data <- merge(meta_data, sample.to.10X.lane, by="sample") %>% ## mutate(CBC_10x = paste0(CBC, "-", sample_num)) %>% ## merge(guidePerCell.df %>% select(CBC_10x, guides_per_cbc, max_umi_ct), by="CBC_10x") %>% ## merge(umiPerCell.df, by="long.CBC") %>% ## merge(geneDetectedPerCell.df, by.x="long.CBC", by.y=0) } else { ## barcode.names <- read.table(opt$barcode.names, header=T, stringsAsFactors=F) ## %>% `colnames<-`("long.CBC") ## print("finished loading barcode names") ## print(paste0("omega dimensions: ", dim(omega))) ## print(paste0("barcode names dimensions: ", dim(barcode.names))) meta_data <- barcode.names.subset %>% filter(Gene %in% c(gene.ary, "negative-control")) meta_data <- meta_data %>% mutate(sample = factor(sample)) } ## load model fitting fomulas test.cmd.df <- read.delim(paste0(INPUTDIR, "MAST_model_formulas.txt"), stringsAsFactors=F) ## ## 220222 scratch ## if ( !( "ann.omega" %in% ls()) ) { ## if( grepl("2kG.library|Perturb_2kG_dup4", SAMPLE) ) { ## ann.omega <- omega %>% ## as.data.frame %>% ## mutate(long.CBC = rownames(.)) %>% ## separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% ## mutate(Gene = gsub("-TSS2$", "", Gene.full.name)) ## } else { ## ann.omega <- omega %>% as.data.frame %>% ## mutate(CBC = rownames(.)) %>% ## merge(barcode.names, by="CBC", all.x=T) ## ## meta_data <- barcode.names ## } ## } else { ## ann.omega <- ann.omega %>% ## mutate(Gene = gsub("-TSS2$", "", Gene.full.name)) ## remove TSS2 annotation ## } ## test ## omega.tpm <- omega %>% as.matrix %>% apply(2, function(x) x / sum(x) * 1000000) ## convert to TPM ## ## end of test ## ## omega.tpm <- ann.omega[,1:k] %>% as.matrix %>% apply(2, function(x) x / sum(x) * 1000000) ## convert to TPM ## log2.omega <- (omega.tpm + 1) %>% log2 ## log2(TPM + 1) ## ## ann.omega.original <- ann.omega ## ## load log2(TPM + 1) ## log2.omega <- readRDS(paste0(opt$scatteroutput, "/", SAMPLE, "_MAST_log2TPM.RDS")) if( grepl("2kG.library|Perturb_2kG_dup4", SAMPLE) ) { df <- merge(log2.omega %>% as.data.frame %>% mutate(long.CBC = rownames(.)), meta_data, by="long.CBC") } else { df <- merge(log2.omega %>% as.data.frame %>% mutate(CBC = rownames(.)), meta_data, by="CBC", all.y=T) } ## gene.here <- "MESDC1" gene.list <- gene.ary num.ptb <- length(gene.list) cat(paste0("number of genes: ", num.ptb, "\n")) ## MAST.list <- vector("list", num.ptb) MAST.list <- mclapply(1:num.ptb, function(i) { gene.here <- gene.list[i] out <- tryCatch( { suppressWarnings({ totest.df <- df %>% mutate(Gene = gsub("safe-targeting","negative-control", Gene)) %>% ## combine safe targeting and negative control guide to call them "negative-control" subset(Gene %in% c("negative-control", gene.here)) ## take a small slice of data (one perturbation one control) scaRaw <- FromMatrix(totest.df %>% select(-any_of(colnames(meta_data))) %>% t) colData(scaRaw)$perturb_status <- totest.df %>% pull(Gene) colData(scaRaw)$lane <- totest.df %>% pull(sample) cond <- factor(colData(scaRaw)$perturb_status) cond <- relevel(cond, "negative-control") colData(scaRaw)$condition <- cond format.zlm.result <- function(zlmCond) { condition.str <- paste0('condition', gene.here) summaryCond <- summary(zlmCond, doLRT=condition.str, parallel = T) summaryDt <- summaryCond$datatable fcHurdle <- merge(summaryDt[contrast==condition.str & component=='H',.(primerid, `Pr(>Chisq)`)], #hurdle P values summaryDt[contrast==condition.str & component=='logFC', .(primerid, coef, ci.hi, ci.lo)], by='primerid') %>% #logFC coefficients mutate(fdr:=p.adjust(`Pr(>Chisq)`, 'fdr'), perturbation = gene.here) return(fcHurdle) } fcHurdle <- do.call(rbind, lapply(1:nrow(test.cmd.df), function(i) { zlmCond <- eval(parse(text = paste0("zlmCond <- zlm(", test.cmd.df$zlm.model.command[i], ", scaRaw)"))) fcHurdle.here <- format.zlm.result(zlmCond) %>% mutate(zlm.model.name = test.cmd.df$zlm.model.name[i]) return(fcHurdle.here) })) cat(gene.here) return(fcHurdle) })}, warning = function(cond) { return(fcHurdle) }, error = function(cond) { return(data.frame(primerid = NA, `Pr(>Chisq)`= NA, coef = NA, ci.hi = NA, ci.lo = NA, fdr = NA, perturbation = gene.here, check.names=F, zlm.model.name=NA) ) } ) ## MAST.list[[i]] <- fcHurdle %>% mutate(perturbation = gene.here) }, mc.cores = max(1, floor(availableCores() - 1))) ## 64G, K=35, 3 perturbations took 30 minutes MAST.df <- do.call(rbind, MAST.list) %>% ## group_by(zlm.model.name) %>% ## mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method='fdr')) %>% as.data.frame write.table(MAST.df, file=paste0(SCATTEROUTDIR, "/", SAMPLE, "_MAST_DEtopics_Group", opt$scatter.gene.group, ".txt"), quote=F, row.names=F, sep="\t") |
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 | library(conflicted) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("filter", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr") #, "grid", "gtable", "gridExtra","ggrepel",#"ramify", ## "ggpubr","gridExtra", "parallel", "future", ## "org.Hs.eg.db","limma","conflicted", #"fgsea", ## "cluster","textshape","readxl", ## "ggdist", "gghalves", "Seurat", "writexl", "SingleCellExperiment", "MAST") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", xfun::pkg_attach(packages) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") ########################################################################################## ## Constants and Directories option.list <- list( make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--sampleName", type="character", default="2kG.library", help="sample name"), make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/2kG.library.barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--outdirsample", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/", help="path to cNMF analysis results"), ## or for 2n1.99x: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/K60/threshold_0_2/" make_option("--num.genes.per.MAST.runGroup", type="numeric", default=494, help="Number of MAST parallel processes to create"), make_option("--scatteroutput", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/", help="path to gene breakdown table output"), ## script dir make_option("--scriptdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts/", help="location for this script and functions script") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev debug K562 gwps ## opt$sampleName <- "WeissmanK562gwps" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/data/K562_gwps_raw_singlecell_01_metadata.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K35/threshold_0_2/" ## opt$scatteroutput <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/230104_snakemake_WeissmanLabData/top2000VariableGenes/MAST/" ## opt$scriptdir <- "/oak/stanford/groups/engreitz/Users/kangh/cNMF_pipeline/Perturb-seq/workflow/scripts" ## opt$K.val <- 35 k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K31/threshold_0_2/" ## OUTDIRSAMPLE <- opt$outdirsample SCATTEROUTDIR <- opt$scatteroutput ## INPUTDIR <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/inputs/" ## OUTDIR <- OUTDIRSAMPLE ## check.dir <- c(INPUTDIR, OUTDIR, SCATTEROUTDIR) check.dir <- c(SCATTEROUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## mytheme <- theme_classic() + theme(axis.text = element_text(size = 12), axis.title = element_text(size = 16), plot.title = element_text(hjust = 0.5, face = "bold")) ## palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## p.adjust.thr <- 0.1 ## ########################################################################################## ## ## load data ## ## load known CAD genes ## CAD.genes <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/known_CAD_gene_set.txt", header=F, stringsAsFactors=F) %>% as.matrix %>% as.character ## ## load topic model results ## cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") ## print(cNMF.result.file) ## if(file.exists(cNMF.result.file)) { ## print("loading cNMF result file") ## load(cNMF.result.file) ## } ## Organize metadata print("organize meta data") ## get metadata from log2.X.full rownames # if(SAMPLE %in% c("2kG.library", "Perturb_2kG_dup4")) { if( grepl("2kG.library|Perturb_2kG_dup4", SAMPLE) ) { barcode.names <- read.table(opt$barcode.names, header=F, stringsAsFactors=F) %>% `colnames<-`("long.CBC") ## rownames(omega) <- barcode.names %>% pull(long.CBC) %>% gsub("CSNK2B-and-CSNK2B", "CSNK2B",.) barcode.names <- barcode.names %>% rownames %>% as.data.frame %>% `colnames<-`("long.CBC") %>% separate(col="long.CBC", into=c("Gene.full.name", "Guide", "CBC"), sep=":", remove=F) %>% ## separate(col="CBC", into=c("CBC", "sample"), sep="-", remove=F) %>% separate(col="CBC", into=c("CBC", "sample"), sep="-scRNAseq_2kG_", remove=F) %>% mutate(Gene = gsub("-TSS2", "", Gene.full.name), CBC = gsub("RHOA-and-", "", CBC)) %>% as.data.frame sample.to.10X.lane <- data.frame(sample_num = 1:20, sample = meta_data$sample %>% unique %>% sort) ## meta_data <- merge(meta_data, sample.to.10X.lane, by="sample") %>% ## mutate(CBC_10x = paste0(CBC, "-", sample_num)) %>% ## merge(guidePerCell.df %>% select(CBC_10x, guides_per_cbc, max_umi_ct), by="CBC_10x") %>% ## merge(umiPerCell.df, by="long.CBC") %>% ## merge(geneDetectedPerCell.df, by.x="long.CBC", by.y=0) } else { barcode.names <- read.table(opt$barcode.names, header=T, stringsAsFactors=F) ## %>% `colnames<-`("long.CBC") } cat("finished loading barcode names\n") ## print(paste0("omega dimensions: ", dim(omega))) cat(paste0("barcode names dimensions: ", dim(barcode.names)[1], " x ", dim(barcode.names)[2], "\n")) num_genes <- barcode.names %>% pull(Gene) %>% unique %>% length cat(paste0("total number of perturbations: ", num_genes, "\n")) numGenesPerRun <- opt$num.genes.per.MAST.runGroup num_MAST_runs <- floor(num_genes / numGenesPerRun) + 1 cat(paste0("total number of MAST runs: ", num_MAST_runs, "\n")) gene.ary <- barcode.names %>% pull(Gene) %>% unique %>% sort ## separate into numGenesPerRun perturbations per group for MAST for (i in 1:(num_MAST_runs-1)) { gene.ary.i <- gene.ary[((i-1)*numGenesPerRun+1) : (numGenesPerRun*i)] write.table(gene.ary.i, paste0(SCATTEROUTDIR, "/GeneNames_Group", i, ".txt"), sep="\n", quote=F, row.names=F, col.names=F) } i <- i + 1 # the last one gene.ary.i <- gene.ary[((i-1)*numGenesPerRun+1) : length(gene.ary)] write.table(gene.ary.i, paste0(SCATTEROUTDIR, "/GeneNames_Group", i, ".txt"), sep="\n", quote=F, row.names=F, col.names=F) |
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 | library(conflicted) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel","ramify", "ggpubr","gridExtra", "org.Hs.eg.db","limma","fgsea", "cluster","textshape","readxl", "ggdist", "gghalves", "writexl") # "GGally","RNOmni","usedist","GSEA","clusterProfiler","IsoplotR","wesanderson", ## packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", ## "CountClust", "Hmisc", "tidyr", "grid", "gtable", "gridExtra","ggrepel","ramify", ## "GGally","RNOmni","usedist","ggpubr","gridExtra","GSEA", ## "org.Hs.eg.db","limma","clusterProfiler","fgsea", "conflicted", ## "cluster","textshape","readxl", "IsoplotR", "wesanderson", ## "ggdist", "gghalves", "Seurat", "writexl") # library(Seurat) xfun::pkg_attach(packages) conflict_prefer("slice", "dplyr") ## source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") source("workflow/scripts/motif_enrichment_functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220307_prioritized_topic_motif_enrichment/figures/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220307_prioritized_topic_motif_enrichment/outputs/", help="Output directory"), # make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/", help="Input 10x data directory"), # make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210707_snakemake_maxParallel/all_genes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="FT010_fresh_4min", help="Name of Samples to be processed, separated by commas"), # make_option("--sep", type="logical", default=F, help="Whether to separate replicates or samples"), make_option("--K.list", type="character", default="2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,19,21,23,25", help="K values available for analysis"), make_option("--K.val", type="numeric", default=20, help="K value to analyze"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--ABCdir",type="character", default="/oak/stanford/groups/engreitz/Projects/ABC/200220_CAD/ABC_out/TeloHAEC_Ctrl/Neighborhoods/", help="Path to ABC enhancer directory"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), # make_option("--raw.mtx.dir",type="character",default="stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/data/no_IL1B_filtered.normalized.ptb.by.gene.mtx.filtered.txt", help="input matrix to cNMF pipeline"), # make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), # the first lane: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.expandedMultiTargetGuide.RDS" # make_option("--subsample.type", type="character", default="", help="Type of cells to keep. Currently only support ctrl"), # make_option("--barcode.names", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/barcodes.tsv", help="barcodes.tsv for all cells"), make_option("--reference.table", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/210702_2kglib_adding_more_brief_ca0713.xlsx"), ## fisher motif enrichment ## make_option("--outputTable", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputTableBinary", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.table.binary.k_14.df_0_2.txt", help="Output directory"), ## make_option("--outputEnrichment", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210607_snakemake_output/outputs/no_IL1B/topic.top.100.zscore.gene.motif.fisher.enrichment.k_14.df_0_2.txt", help="Output directory"), make_option("--motif.promoter.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv", help="All promoter's motif matches"), make_option("--motif.enhancer.background", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv", help="All enhancer's motif matches specific to {no,plus}_IL1B"), make_option("--enhancer.fimo.threshold", type="character", default="1.0E-4", help="Enhancer fimo motif match threshold"), make_option("--ep.type", type="character", default="enhancer", help="motif enrichment for enhancer or promoter, specify 'enhancer' or 'promoter'"), #summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute"), make_option("--motif.match.thr.str", type="character", default="pval0.0001", help="threshold for subsetting motif matches"), ## Organism flag make_option("--organism", type="character", default="human", help="Organism type, accept org.Hs.eg.db. Only support human and mouse.") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" ## opt$K.val <- 60 ## opt$sampleName <- "2kG.library" # ## ## all genes directories (for sdev) # opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/" # opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" # opt$K.val <- 60 # opt$sampleName <- "2kG.library" ## ## debug ctrl ## opt$topic.model.result.dir <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210810_snakemake_ctrls/all_genes_acrossK/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation" ## opt$sampleName <- "2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210810_snakemake_ctrls/figures/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210810_snakemake_ctrls/analysis/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation/all_genes/" ## opt$barcode.names <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210806_curate_ctrl_mtx/outputs/2kG.library.no.DE.gene.with.FDR.less.than.0.1.perturbation.barcodes.tsv" ## opt$K.val <- 60 ## ## K562 gwps sdev ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$K.val <- 100 ## opt$ep.type <- "enhancer" ## opt$motif.match.thr.str <- "qval0.1" ## opt$motif.enhancer.background <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/fimo/fimo_out/fimo.formatted.tsv" ## opt$motif.promoter.background <- "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv" ## ## ENCODE mouse heart ## opt$sampleName <- "mouse_ENCODE_heart" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes" ## opt$K.val <- 5 ## opt$ep.type <- "enhancer" ## opt$motif.match.thr.str <- "pval1e-4" ## opt$motif.enhancer.background <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230116_snakemake_mouse_ENCODE_heart/analysis/top2000VariableGenes/mouse_ENCODE_heart/fimo/fimo_out/fimo.txt" ## opt$motif.promoter.background <- "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv" ## ## no_IL1B 200 gene library ## opt$sampleName <- "no_IL1B" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/tutorials/2306_V2G2P_prep/analysis/all_genes" ## opt$K.val <- 14 ## opt$ep.type <- "promoter" ## opt$motif.match.thr.str <- "qval0.1" ## opt$motif.enhancer.background <- "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.tsv" ## opt$motif.promoter.background <- "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv" ## ## IGVF b01_LeftCortex sdev ## opt$sampleName <- "IGVF_b01_LeftCortex" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/figures/all_genes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes" ## opt$K.val <- 20 ## opt$ep.type <- "enhancer" ## opt$organism <- "mouse" ## opt$motif.match.thr.str <- "pval1e-6" ## opt$motif.enhancer.background <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/IGVF_b01_LeftCortex/fimo/fimo_out/fimo.txt" ## opt$motif.promoter.background <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/IGVF_b01_LeftCortex/fimo/fimo_out/fimo.txt" mytheme <- theme_classic() + theme(axis.text = element_text(size = 9), axis.title = element_text(size = 11), plot.title = element_text(hjust = 0.5, face = "bold")) SAMPLE=strsplit(opt$sampleName,",") %>% unlist() DATADIR=opt$olddatadir # "/seq/lincRNA/Gavin/200829_200g_anal/scRNAseq/" OUTDIR=opt$outdir # TMDIR=opt$topic.model.result.dir # SEP=opt$sep # K.list <- strsplit(opt$K.list,",") %>% unlist() %>% as.numeric() k <- opt$K.val num.top.genes <- 300 ## number of topic defining genes ep.type <- opt$ep.type DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIR=opt$figdir FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") FGSEADIR=paste0(OUTDIRSAMPLE,"/fgsea/") FGSEAFIG=paste0(FIGDIRSAMPLE,"/fgsea/") ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr motif.match.thr.str <- opt$motif.match.thr.str ## create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE, FGSEADIR, FGSEAFIG) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ###################################################################### ## Load topic model results cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } ## not used # ## load hg38 promoter region file # promoter.region.hg38.original <- read.table(file=paste0("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/RefSeqCurated.170308.bed.CollapsedGeneBounds.TSS300bp.hg38.bed"), header = F, stringsAsFactors = F) %>% # `colnames<-`(c("chr","start","end", "gene","cell.type","strand")) %>% # mutate(sequence_name = paste0(chr, ":", start, "-", end, "(", strand, ")")) # ## keep only the expressed gene's motifs for background # expressed.genes <- theta %>% rownames() # promoter.region.hg38 <- promoter.region.hg38.original %>% # mutate(expressed = (gene %in% expressed.genes) ) %>% # filter(expressed) ########### ## load FIMO matched motifs ## ifelse on promoter vs enhancer if (ep.type == "promoter") { ## load promoter motif matches motif.background <- read.delim(file=paste0(ifelse(opt$motif.promoter.background!="", opt$motif.promoter.background, "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv")), header=F, stringsAsFactors=F) ## %>% filter(!grepl("#", motif_id)) # 30 seconds if(ncol(motif.background) > 9) { motif.background <- motif.background %>% `colnames<-`(c("motif_id", "motif_alt_id", "enhancer_region", "enhancer_type", "gene_region","sequence_name","start","stop","motif.matched.strand","score","p.value","q.value","matched_sequence")) %>% filter(!grepl("#|motif_id", motif_id)) # more than 30 seconds, minutes? motif.background <- motif.background %>% filter(grepl("promoter", enhancer_type)) } else { motif.background <- motif.background %>% `colnames<-`(c("motif_id", "sequence_name", "start", "stop", "motif.matched.strand", "score", "p.value", "q.value", "matched_sequence")) } ## old ## motif.background <- read.delim(file=paste0(ifelse(opt$motif.promoter.background!="", opt$motif.promoter.background, "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModel/2104_remove_lincRNA/data/fimo_out_all_promoters_thresh1.0E-4/fimo.tsv")), header=T, stringsAsFactors=F) ## %>% filter(!grepl("#", motif_id)) # 30 seconds ## ## colnames(motif.background) <- c("motif_id", "motif_alt_id", "enhancer_region", "enhancer_type", "gene_region","sequence_name","start","stop","motif.matched.strand","score","p.value","q.value","matched_sequence") ## colnames(motif.background)[colnames(motif.background) == "strand"] <- "motif.matched.strand" ## motif.background <- motif.background %>% ## mutate(motif.short = strsplit(motif_id, split="_") %>% sapply("[[", 1) %>% as.character) ## end of old } else { ## load enhancer motif matches print(opt$motif.enhancer.background) motif.background <- read.delim(file=paste0(ifelse(opt$motif.enhancer.background!="", opt$motif.enhancer.background, "/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/cNMF/2104_all_genes/data/fimo_out_ABC_TeloHAEC_Ctrl_thresh1.0E-4/fimo.formatted.tsv")), header=F, stringsAsFactors=F) if(ncol(motif.background) > 9) { motif.background <- motif.background %>% `colnames<-`(c("motif_id", "motif_alt_id", "enhancer_region", "enhancer_type", "gene_region","sequence_name","start","stop","motif.matched.strand","score","p.value","q.value","matched_sequence")) %>% filter(!grepl("#|motif_id", motif_id)) # more than 30 seconds, minutes? motif.background <- motif.background %>% filter(!grepl("promoter", enhancer_type)) } else { motif.background <- motif.background %>% `colnames<-`(c("motif_id", "sequence_name", "start", "stop", "motif.matched.strand", "score", "p.value", "q.value", "matched_sequence")) } } motif.background <- motif.background %>% mutate(motif.short = strsplit(motif_id, split="_") %>% sapply("[[", 1) %>% as.character) message("finished loading motif input") ## subset to q.value < 0.1 if (grepl("qval", motif.match.thr.str)) { motif.background <- motif.background %>% subset(q.value < 0.1) } else { print(paste0("subset to ", motif.match.thr.str)) threshold <- gsub("pval", "", motif.match.thr.str) %>% as.numeric motif.background <- motif.background %>% subset(p.value < threshold) } if(ep.type == "enhancer") { if(ncol(motif.background) == 9 | sum(as.numeric(grepl("|", motif.background$sequence_name))) == nrow(motif.background)) { motif.background <- motif.background %>% filter(!grepl("promoter", sequence_name) & !grepl("start", start)) %>% separate(col="sequence_name", into=c("enhancer_region", "enhancer_type", "gene_region", "gene_name_sequence_region"), sep="[|]") %>% separate(col="gene_name_sequence_region", into=c("sequence_name", "to_remove"), sep="::") %>% select(-to_remove) } colnames(motif.background)[colnames(motif.background) == "strand"] <- "motif.matched.strand" motif.background <- motif.background %>% mutate(motif.short = strsplit(motif_id, split="_") %>% sapply("[[", 1) %>% as.character) } expressed.genes <- rownames(theta.zscore) ## todo: convert expressed genes to symbol if they are not in symbol db <- ifelse(grepl("mouse|org.Mm.eg.db", opt$organism), "org.Mm.eg.db", "org.Hs.eg.db") gene.type <- ifelse(length(expressed.genes) == sum(as.numeric(grepl("^ENS", expressed.genes))), "ENSGID", "Gene") if(gene.type == "ENSGID") expressed.genes = mapIds(get(db), keys=expressed.genes, keytype="ENSEMBL", column="SYMBOL") motif.background <- motif.background %>% subset(sequence_name %in% expressed.genes) #################################################################################################### ## get list of topic defining genes theta.rank.list <- vector("list", ncol(theta.zscore))## initialize storage list for(i in 1:ncol(theta.zscore)) { topic <- paste0("topic_", colnames(theta.zscore)[i]) theta.rank.list[[i]] <- theta.zscore[,i] %>% as.data.frame %>% `colnames<-`("topic.zscore") %>% mutate(Gene = rownames(.)) %>% arrange(desc(topic.zscore), .before="topic.zscore") %>% mutate(zscore.specificity.rank = 1:n()) %>% ## add rank column mutate(Topic = topic) ## add topic column } theta.rank.df <- do.call(rbind, theta.rank.list) ## combine list to df topic.defining.gene.df <- theta.rank.df %>% subset(zscore.specificity.rank <= num.top.genes) ## select top 300 genes for each topic gene.type <- ifelse(nrow(topic.defining.gene.df) == sum(as.numeric(grepl("^ENS", topic.defining.gene.df$Gene))), "ENSGID", "Gene") if(gene.type=="ENSGID") topic.defining.gene.df <- topic.defining.gene.df %>% mutate(ENSGID = Gene) %>% mutate(Gene = mapIds(get(db), keys=.$ENSGID, keytype="ENSEMBL", column="SYMBOL")) ## topic.defining.gene.df <- topic.defining.gene.df %>% mutate(Gene = toupper(Gene)) topic.motif.match.df <- merge(topic.defining.gene.df, motif.background %>% select(motif_id, motif.short, sequence_name, score, p.value, q.value, motif.matched.strand), by.x="Gene", by.y="sequence_name", all.y=T) ## filtered motif.background to genes expressed in this data set, so keep all motif.id.type <- "motif.short" ## or "motif_id" topic.motif.match.df.long <- topic.motif.match.df %>% group_by(Gene, get(motif.id.type), Topic) %>% summarize(count = n()) %>% `colnames<-`(c("gene", motif.id.type, "Topic", "count")) %>% as.data.frame wide <- topic.motif.match.df.long %>% spread(key=motif.id.type, value="count", fill=0) %>% ## get each motif's count in each top promoter mutate(topic.value=1) %>% spread(key=Topic, value=topic.value, fill=0) ## %>% select(-topic_NA) ## get presence/absense of promoter in topic ## 5 seconds print(paste0(ep.type, " motif (", motif.match.thr.str, ") count t-test")) ttest.df <- ttest.on.motifs(wide) ## significant.motifs.df <- ttest.df %>% ## group_by(topic) %>% ## subset(p.adjust < 0.1 & ## enrichment.log2fc > 0) %>% ## arrange(p.adjust) %>% ## summarize(motifs = paste0(motif %>% sort, collapse=",")) %>% ## as.data.frame #################################################################################################### ## save results all.ttest.df.path <- paste0(OUTDIRSAMPLE,"/", ep.type, ".topic.top.", num.top.genes, ".zscore.gene_motif.count.ttest.enrichment_motif.thr.", motif.match.thr.str, "_", SUBSCRIPT.SHORT,".txt") print(paste0("saving to ", all.ttest.df.path)) write.table(ttest.df, all.ttest.df.path, sep="\t", quote=F, row.names=F) |
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 | suppressPackageStartupMessages(library(conflicted)) conflict_prefer("combine", "dplyr") conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("first", "dplyr") conflict_prefer("rename", "dplyr") suppressPackageStartupMessages({ library(optparse) library(dplyr) library(tidyr) library(reshape2) ## library(ggplot2) ## library(cowplot) ## library(ggpubr) ## ggarrange ## library(gplots) ## heatmap.2 ## library(scales) ## geom_tile gradient rescale ## library(ggrepel) library(stringr) library(stringi) library(svglite) ## library(Seurat) ## library(SeuratObject) library(xlsx) library(yaml) }) ########################################################################################## ## Constants and Directories option.list <- list( make_option("--sampleName", type="character", default="WeissmanK562gwps", help="Name of the sample"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/", help="Output directory"), make_option("--K.val", type="numeric", default=90, help="K value to analyze"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--perturbSeq", type="logical", default=TRUE, help="Whether this is a Perturb-seq experiment"), make_option("--level", type="character", default="cell line", help="Sample type (e.g. tissue, cell line, primary cells"), make_option("--cell.type", type="character", default="teloHAEC", help="Cell type description (e.g. brain, teloHAEC, K562)") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev IGVF b01_LeftCortex ## opt$sampleName <- "IGVF_b01_LeftCortex" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes/" ## opt$K.val <- 15 ## opt$perturbSeq <- FALSE ## opt$level <- "tissue" ## opt$cell.type <- "brain" SAMPLE=strsplit(opt$sampleName,",") %>% unlist() OUTDIR=opt$outdir k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") OUTDIRSAMPLEIGVF = paste0(OUTDIRSAMPLE, "IGVF_format/") check.dir <- c(OUTDIRSAMPLEIGVF) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## subscript for files SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) ## SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## cNMF direct output file (GEP) cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") ## cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.k_", k, ".dt_", density.threshold, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } else { print(paste0("file ", cNMF.result.file, " not found")) } db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) ## load the appropriate database topic.gene.names <- rownames(theta.zscore) topic.gene.name.type <- ifelse(grepl("^ENSG", topic.gene.names) %>% as.numeric %>% sum == length(topic.gene.names), "ENSGID", "SYMBOL") if(topic.gene.name.type == "ENSGID") { ENSGID.gene.names <- topic.gene.names SYMBOL.gene.names <- mapIds(get(db), keys=topic.gene.names, keytype = "ENSEMBL", column = "SYMBOL") SYMBOL.gene.names[is.na(SYMBOL.gene.names)] <- ENSGID.gene.names[is.na(SYMBOL.gene.names)] } else { SYMBOL.gene.names <- topic.gene.names ENSGID.gene.names <- mapIds(get(db), keys=topic.gene.names, keytype = "SYMBOL", column = "ENSEMBL") ENSGID.gene.names[is.na(ENSGID.gene.names)] <- SYMBOL.gene.names[is.na(ENSGID.gene.names)] } ## load variance explained variance.explained.df <- read.delim(paste0(OUTDIRSAMPLE, "metrics.varianceExplained.df.txt"), stringsAsFactors=F) ## 1. Model YAML file: Capture all the information about the dataset that you used and which method and how you run it along with Topic_ID which points to Topic YAML files and Cell-Topic participation ID for pointing out the Cell-Topic participation h5ad file. out <- list("Assay" = NULL, "Cell-Topic participation ID" = NULL, "Experiment ID" = SAMPLE, "Name of method" = "cNMF", "Number of topics" = k, "Technology" = "10x", "cNMF spectra threshold" = opt$density.thr, "Topic IDs" = paste0(SAMPLE, "_K", k, "_", 1:k), "level" = opt$level, "cell type" = opt$cell.type) write_yaml(out, paste0(OUTDIRSAMPLEIGVF, SAMPLE, ".", SUBSCRIPT.SHORT, ".modelYAML.yaml")) ## 2. Topics YAML files: Capture all the information about Topics including Topic_ID, gene_weight, gene_id and gene_name and any other information that suits your data theta.zscore.long <- theta.zscore %>% as.data.frame %>% `colnames<-`(paste0(SAMPLE, "_K", k, "_", colnames(.))) %>% mutate(Gene = SYMBOL.gene.names, ENSGID = ENSGID.gene.names) %>% melt(id.vars=c("Gene", "ENSGID"), value.name="Gene weights", variable.name="Topic ID") %>% rename("gene_id" = "ENSGID") for( t in 1:k ) { theta.zscore.long.here <- theta.zscore.long %>% subset(grepl(paste0("_", t, "$"), `Topic ID`)) duplicated.index <- duplicated(theta.zscore.long.here$Gene) theta.zscore.long.here$Gene[duplicated.index] <- paste0(theta.zscore.long.here$Gene[duplicated.index], "_", theta.zscore.long.here$ENSGID[duplicated.index]) variance.here <- variance.explained.df %>% subset(ProgramID == paste0("K", k, "_", t)) %>% pull(VarianceExplained) ## create output list ## out <- list("gene_id" = theta.zscore.long.here %>% ## `rownames<-`(.$Gene) %>% ## select(gene_id) %>% t %>% as.data.frame, ## "Gene weights" = theta.zscore.long.here %>% ## `rownames<-`(.$Gene) %>% ## select(`Gene weights`) %>% t %>% as.data.frame, ## "Topic ID" = paste0(SAMPLE, "_K", k, "_", t)) out <- list("Gene information" = list("gene_id" = theta.zscore.long.here %>% `rownames<-`(.$Gene) %>% select(gene_id) %>% t %>% as.data.frame), "Gene weights" = theta.zscore.long.here %>% `rownames<-`(.$Gene) %>% select(`Gene weights`) %>% t %>% as.data.frame, "Topic ID" = paste0(SAMPLE, "_K", k, "_", t), "Topic Information" = list("variance" = variance.here)) write_yaml(out, paste0(OUTDIRSAMPLEIGVF, SAMPLE, "_", SUBSCRIPT.SHORT, "_program", t, "_topicYAML.yaml")) } |
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 | suppressPackageStartupMessages({ library(optparse) library(dplyr) library(tidyr) library(reshape2) library(ggplot2) library(cowplot) library(ggpubr) ## ggarrange library(gplots) ## heatmap.2 library(scales) ## geom_tile gradient rescale library(ggrepel) library(stringr) library(svglite) library(ggseqlogo) library(universalmotif) }) ########################################################################################## ## Constants and Directories ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211205_sig_topic_TPM/figures/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/" option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/", help="Output directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of Samples to be processed, separated by commas"), make_option("--K.val", type="numeric", default=60, help="K value to analyze"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--raw.mtx.RDS.dir",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210623_aggregate_samples/outputs/aggregated.2kG.library.mtx.cell_x_gene.RDS", help="input matrix to cNMF pipeline"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), ## GSEA parameters make_option("--ranking.type", type="character", default="zscore", help="{zscore, raw} ranking for the top program genes"), make_option("--GSEA.type", type="character", default="GOEnrichment", help="{GOEnrichment, ByWeightGSEA, GSEA}") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## overdispersed genes (for sdev) ## opt$sampleName <- "2kG.library_overdispersedGenes" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes" ## opt$K.val <- 100 ## ## K562 gwps sdev ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/" ## opt$K.val <- 55 ## opt$ranking.type <- "zscore" ## opt$GSEA.type <- "GSEA" ## IGVF b01_LeftCortex sdev opt$sampleName <- "IGVF_b01_LeftCortex" opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/figures/all_genes/" opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/all_genes" opt$K.val <- 10 opt$ranking.type <- "median_spectra" opt$GSEA.type <- "GSEA" OUTDIR <- opt$outdir FIGDIR <- opt$figdir SAMPLE <- opt$sampleName k <- opt$K.val DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) FIGDIRSAMPLE=paste0(FIGDIR, "/", SAMPLE, "/K",k,"/") FIGDIRTOP=paste0(FIGDIRSAMPLE,"/",SAMPLE,"_K",k,"_dt_", DENSITY.THRESHOLD,"_") OUTDIRSAMPLE=paste0(OUTDIR, "/", SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) # SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) ## GSEA specific parameters ranking.type.here <- opt$ranking.type GSEA.type <- opt$GSEA.type message(FIGDIRTOP) ## adjusted p-value threshold fdr.thr <- opt$adj.p.value.thr p.value.thr <- opt$adj.p.value.thr # create dir if not already check.dir <- c(OUTDIR, FIGDIR, paste0(FIGDIR,SAMPLE,"/"), paste0(FIGDIR,SAMPLE,"/K",k,"/"), paste0(OUTDIR,SAMPLE,"/"), OUTDIRSAMPLE, FIGDIRSAMPLE) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) mytheme <- theme_classic() + theme(axis.text = element_text(size = 7), axis.title = element_text(size = 8), plot.title = element_text(hjust = 0.5, face = "bold", size=10), axis.line = element_line(color = "black", size = 0.25), axis.ticks = element_line(color = "black", size = 0.25), legend.key.size = unit(10, units="pt"), legend.text = element_text(size=7), legend.title = element_text(size=8) ) file.name <- paste0(OUTDIRSAMPLE, "/clusterProfiler_GeneRankingType", ranking.type.here, "_EnrichmentType", GSEA.type,".txt") gsea.df <- read.delim(file.name, stringsAsFactors=F) if(nrow(gsea.df) == 0) { toplot <- data.frame() } else { toplot <- gsea.df %>% subset(p.adjust < fdr.thr) %>% group_by(ProgramID) %>% arrange(p.adjust) %>% unique %>% slice(1:10) %>% mutate(TruncatedDescription = str_trunc(paste0(ID, "; ", Description), width=50, side="right"), t = gsub("K60_", "", ProgramID) %>% as.numeric) %>% arrange(t, p.adjust) %>% as.data.frame } plot.title <- paste0(ifelse(grepl("GO", GSEA.type), "GO Term Enrichment", "MSigDB Pathway Enrichment"), "\non ", ifelse(grepl("zscore", ranking.type.here), "Program Gene Specificity", "Raw Weight"), "\nby ", ifelse(grepl("ByWeight", GSEA.type), "All Gene Weight", "Top 300 Gene Set")) pdf(file=paste0(FIGDIRTOP,"top10EnrichedPathways_GeneRankingType", ranking.type.here, "_EnrichmentType", GSEA.type, ".pdf"), width=4, height=4) for(program in (paste0("K", k, "_", c(1:k)))) { t <- strsplit(program, split="_") %>% sapply(`[[`,2) toplot.here <- toplot %>% subset(ProgramID %in% program) %>% arrange(p.adjust) labels <- toplot.here$TruncatedDescription %>% unique %>% rev toplot.here <- toplot.here %>% mutate(TruncatedDescription = factor(TruncatedDescription, levels = labels)) p <- toplot.here %>% ggplot(aes(x=TruncatedDescription, y=-log10(p.adjust))) + geom_col(fill="gray") + coord_flip() + mytheme + xlab(ifelse(grepl("GO", GSEA.type), "GO Terms", "Pathways")) + ylab("FDR (-log10)") + ggtitle(paste0(plot.title, "\nProgram ", t)) print(p) } dev.off() |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 | library(conflicted) conflict_prefer("first", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("collapse", "dplyr") conflict_prefer("combine", "dplyr") packages <- c("optparse","dplyr", "data.table", "reshape2", "ggplot2", "tidyr", "textshape","readxl", "AnnotationDbi") xfun::pkg_attach(packages) conflict_prefer("select", "dplyr") option.list <- list( make_option("--feature.dir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/data/features/pops_features_raw/"), make_option("--output", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211101_normalized_features/outputs/", help="output directory") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## for sdev ## opt$feature.dir <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/211101_20sample_snakemake/pops/features/pops_features_raw" ## opt$output <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/pops/" ## opt$prefix <- "CAD_aug6_cNMF30" OUTDIR=opt$output # PREFIX=opt$prefix ## ## load cNMF features ## features <- read.delim(opt$, stringsAsFactors=F) ## get files in feature directory feature.file.paths <- dir(opt$feature.dir) ## load all features num_feature_files = length(feature.file.paths) all.features.list <- vector("list", num_feature_files) for (feature.raw.index in 1:num_feature_files) { all.features.list[[feature.raw.index]] <- read.delim(paste0(opt$feature.dir, "/", feature.file.paths[[feature.raw.index]]), stringsAsFactors=F) } all.features <- Reduce(function(x,y) merge(x,y,by="ENSGID"), all.features.list) ## features without cNMF # all.features <- data.frame(1) ## all.features.cNMF <- merge(all.features, features, by="ENSGID") ## all features with cNMF print(paste0("Saving all.features to ", OUTDIR, " as an RDS file")) saveRDS(all.features, file=paste0(OUTDIR, "/full_external_features.RDS")) print(paste0("Saving all.features to ", OUTDIR, " as a text file")) write.table(all.features, file=paste0(OUTDIR, "/full_external_features.txt"), row.names=F, quote=F, sep="\t") print("done!") |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | library(conflicted) conflict_prefer("first", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("collapse", "dplyr") conflict_prefer("combine", "dplyr") packages <- c("optparse","dplyr", "data.table", "reshape2", "ggplot2", "tidyr", "textshape","readxl", "AnnotationDbi") xfun::pkg_attach(packages) conflict_prefer("select", "dplyr") option.list <- list( make_option("--feature.RDS", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/data/features/pops_features_raw/"), make_option("--cNMF.features", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211101_normalized_features/outputs/", help="cNMF features in ENSGID to add to all features"), make_option("--output", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211101_normalized_features/outputs/", help="output directory"), make_option("--prefix", type="character", default="CAD_aug6_cNMF60", help="use a format of MAGMA_{with, without}cNMF to specify which features are included") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## for sdev ## opt$feature.dir <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/211101_20sample_snakemake/pops/features/pops_features_raw" ## opt$output <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/pops/" ## opt$prefix <- "CAD_aug6_cNMF30" OUTDIR=opt$output PREFIX=opt$prefix ## load cNMF features features <- read.delim(opt$cNMF.features, stringsAsFactors=F) ## load external features RDS all.features <- readRDS(opt$feature.RDS) ## features without cNMF ## combine cNMF features and external features all.features.cNMF <- merge(all.features, features, by="ENSGID") ## all features with cNMF saveRDS(all.features.cNMF, file=paste0(OUTDIR, "/full_features_", PREFIX, ".RDS")) write.table(all.features.cNMF, file=paste0(OUTDIR, "/full_features_", PREFIX, ".txt", row.names=F, quote=F, sep="\t")) |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 | library(conflicted) conflict_prefer("first", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("collapse", "dplyr") conflict_prefer("combine", "dplyr") packages <- c("optparse","dplyr", "data.table", "reshape2", "ggplot2", "tidyr", "textshape","readxl", "gplots", "AnnotationDbi", "org.Hs.eg.db", "ggrepel", "gplots") xfun::pkg_attach(packages) conflict_prefer("select", "dplyr") option.list <- list( make_option("--project", type="character", default = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/", help="project directory"), make_option("--output", type="character", default = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/", help="output directory"), make_option("--scratch.output", type="character", default="", help="output directory for large files"), make_option("--coefs_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.coefs", help=""), make_option("--marginals_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.marginals", help=""), make_option("--preds_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.preds", help=""), make_option("--coefs_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.coefs", help=""), make_option("--marginals_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.marginals", help=""), make_option("--preds_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.preds", help=""), make_option("--prefix", type="character", default="CAD_aug6_cNMF60", help="magma file name (before genes.raw)"), make_option("--external.features.metadata", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/metadata/metadata_jul17.txt", help="annotations for each external features"), make_option("--cNMF.features", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/data/features/pops_features_raw/topic.zscore.ensembl.scaled_k_60.dt_0_2.txt", help="normalized cNMF weights, unit variance and zero mean"), make_option("--all.features", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211101_normalized_features/outputs/full_features_with_cNMF.RDS", help=".RDS file with all features input into PoPS"), make_option("--recompute", type="logical", default=F, help="T for rerunning the entire script, F for only outputting the missing data") ) opt <- parse_args(OptionParser(option_list=option.list)) SAMPLE=opt$prefix OUTDIR=opt$output SCRATCH.OUTDIR=opt$scratch.output PREFIX=opt$prefix check.dir <- c(OUTDIR, SCRATCH.OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## graphing constants mytheme <- theme_classic() + theme(axis.text = element_text(size = 12), axis.title = element_text(size = 14), plot.title = element_text(hjust = 0.5, face = "bold", size=14)) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## load metadata print("loading metadata") meta.data.path <- opt$external.features.metadata print(meta.data.path) metadata <- read.delim(meta.data.path, stringsAsFactors=F) ## load all features print(paste0("loading all features from ", opt$all.features)) all.features.cNMF <- readRDS(opt$all.features) ## load data print("loading PoPS results") preds <- read.table(file=opt$preds_with_cNMF,header=T, stringsAsFactors=F, sep="\t") colnames(preds) <- paste0(colnames(preds), "_with.cNMF") colnames(preds)[1] <- "ENSGID" preds.before <- read.table(file=paste0(opt$preds_without_cNMF), header=T, stringsAsFactors=F, sep="\t") colnames(preds.before) <- paste0(colnames(preds.before), "_without.cNMF") colnames(preds.before)[1] <- "ENSGID" preds.combined <- merge(preds, preds.before, by="ENSGID") marginals <- read.table(file=opt$marginals_with_cNMF,header=T, stringsAsFactors=F, sep="\t") coefs <- read.table(file=opt$coefs_with_cNMF,header=T, stringsAsFactors=F, sep="\t") coefs.df <- coefs[4:nrow(coefs),] %>% arrange(desc(beta)) coefs.df$beta <- coefs.df$beta %>% as.numeric ## map ids x <- org.Hs.egENSEMBL mapped_genes <- mappedkeys(x) xx.entrez.to.ensembl <- as.list(x[mapped_genes]) # EntrezID to Ensembl xx.ensembl.to.entrez <- as.list(org.Hs.egENSEMBL2EG) # Ensembl to EntrezID y <- org.Hs.egGENENAME y_mapped_genes <- mappedkeys(y) entrez.to.genename <- as.list(y[y_mapped_genes]) genename.to.entrez <- as.list(org.Hs.egGENENAME) z <- org.Hs.egSYMBOL z_mapped_genes <- mappedkeys(z) entrez.to.symbol <- as.list(z[z_mapped_genes]) symbol.to.entrez <- as.list(org.Hs.egSYMBOL) ## function for adding name to df add_gene_name_to_df <- function(df) { out <- df %>% mutate(EntrezID = xx.ensembl.to.entrez[df$ENSGID %>% as.character] %>% sapply("[[",1)) %>% mutate(Gene.name = entrez.to.genename[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character) return(out) } ## add Gene name to preds.df preds.df <- preds %>% mutate(EntrezID = xx.ensembl.to.entrez[preds$ENSGID %>% as.character] %>% sapply("[[",1)) %>% mutate(Gene.name = entrez.to.genename[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character) file.name <- paste0(OUTDIR, "/", PREFIX, ".combined.preds") if( !file.exists(file.name) | opt$recompute ) { preds.combined.df <- preds.combined %>% mutate(EntrezID = xx.ensembl.to.entrez[preds.combined$ENSGID %>% as.character] %>% sapply("[[",1)) %>% mutate(Gene.name = entrez.to.genename[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character) write.table(preds.combined.df %>% apply(2, as.character), file=file.name, row.names=F, quote=F, sep="\t") } else { preds.combined.df <- read.delim(file.name, stringsAsFactors = F) } file.name <- paste0(SCRATCH.OUTDIR, "/", PREFIX, "_coefs.marginals.feature.outer.prod.RDS") ## if( !file.exists(file.name) | opt$recompute ) { ## load cNMF features ## features <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/data/features/pops_features_raw/topic.zscore.ensembl.scaled_k_60.dt_0_2.txt", stringsAsFactors=F) features <- read.delim(opt$cNMF.features, stringsAsFactors=F) ## sort features and subset coefs, take a product ((gene x features) x (features x beta scalar)) coefs.cnmf <- coefs %>% subset(grepl("zscore", parameter)) coefs.cnmf.mtx <- coefs.cnmf %>% `rownames<-`(.$parameter) %>% select(-parameter) coefs.cnmf.mtx$beta <- coefs.cnmf.mtx$beta %>% as.numeric coefs.cnmf.mtx <- coefs.cnmf.mtx %>% as.matrix ## subset features features.names.tokeep <- coefs.cnmf.mtx %>% rownames features.tokeep <- features %>% `rownames<-`(.$ENSGID) %>% select(all_of(features.names.tokeep)) %>% as.matrix ## also test marginals features.names.tokeep.df <- data.frame(X=features.names.tokeep) marginals.tokeep <- inner_join(features.names.tokeep.df, marginals, by="X") %>% `rownames<-`(.$X) %>% select(beta) %>% as.matrix coefs.mtx <- coefs.df %>% `rownames<-`(.$parameter) %>% select(-parameter) coefs.mtx$beta <- coefs.mtx$beta %>% as.numeric coefs.mtx <- coefs.mtx %>% as.matrix all.features.cNMF.names.tokeep <- coefs.mtx %>% rownames all.features.cNMF.tokeep <- all.features.cNMF %>% `rownames<-`(.$ENSGID) %>% select(all_of(all.features.cNMF.names.tokeep)) %>% as.matrix all.features.cNMF.tokeep[all.features.cNMF.tokeep=="True"] <- 1 all.features.cNMF.tokeep[all.features.cNMF.tokeep=="False"] <- 0 storage.mode(all.features.cNMF.tokeep) <- "numeric" all.features.cNMF.names.tokeep.df <- data.frame(X=all.features.cNMF.names.tokeep) ## to subset marginals all.marginals.cNMF.tokeep <- inner_join(all.features.cNMF.names.tokeep.df, marginals, by="X") %>% `rownames<-`(.$X) %>% select(beta) %>% as.matrix saveRDS(all.features.cNMF.names.tokeep, file=paste0(OUTDIR, "/all.features.cNMF.keep.prioritized.mtx.RDS")) ## ## check if PoPS_Score = coefs * features ## ## multiply features with beta ## PoPS_Score.coefs.manual <- features.tokeep %*% coefs.cnmf.mtx %>% `colnames<-`("PoPS_Score.coefs.manual") ## PoPS_Score.marginals.manual <- features.tokeep %*% marginals.tokeep %>% `colnames<-`("PoPS_Score.marginals.manual") ## PoPS_Score.coefs.all <- all.features.cNMF.tokeep %*% coefs.mtx %>% `colnames<-`("PoPS_Score_all.coefs") ## PoPS_Score.marginals.all <- all.features.cNMF.tokeep %*% marginals.tokeep %>% `colnames<-`("PoPS_Score.all.marginals") ## ## PoPS_Score.marginals.manual.all <- features %>% `rownames<-`(.$ENSGID) %>% select(-ENSGID) %*% marginals ## get outer products of features and marginals PoPS_Score.coefs.manual.outer.ENSG <- sweep(features.tokeep, 2, (coefs.cnmf.mtx %>% t), `*`) ### store this matrix PoPS_Score.marginals.manual.outer.ENSG <- sweep(features.tokeep, 2, (marginals.tokeep %>% t), `*`) ### save this matrix PoPS_Score.coefs.all.outer.ENSG <- sweep(all.features.cNMF.tokeep, 2, (coefs.mtx %>% t), `*`) PoPS_Score.marginals.all.outer.ENSG <- sweep(all.features.cNMF.tokeep, 2, (all.marginals.cNMF.tokeep %>% t), `*`) ## function to convert the outer product ENSG name to Gene name add_gene_name <- function(df) { out <- df %>% as.data.frame %>% mutate(EntrezID = xx.ensembl.to.entrez[df %>% rownames %>% as.character] %>% sapply("[[",1)) %>% mutate(Gene.name = entrez.to.genename[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character) return(out) } PoPS_Score.coefs.manual.outer <- add_gene_name(PoPS_Score.coefs.manual.outer.ENSG) PoPS_Score.marginals.manual.outer <- add_gene_name(PoPS_Score.marginals.manual.outer.ENSG) PoPS_Score.coefs.all.outer <- add_gene_name(PoPS_Score.coefs.all.outer.ENSG) PoPS_Score.marginals.all.outer <- add_gene_name(PoPS_Score.marginals.all.outer.ENSG) ## save the results write.table(PoPS_Score.coefs.manual.outer %>% apply(2, as.character), file=paste0(OUTDIR, "/", PREFIX, "_coefs.feature.outer.prod.txt"), quote=F, sep="\t") write.table(PoPS_Score.marginals.manual.outer %>% apply(2, as.character), file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_marginals.feature.outer.prod.txt"), quote=F, sep="\t") write.table(PoPS_Score.coefs.all.outer %>% apply(2, as.character), file=paste0(OUTDIR, "/", PREFIX, "_coefs.all.feature.outer.prod.txt"), quote=F, sep="\t") write.table(PoPS_Score.marginals.all.outer %>% apply(2, as.character), file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_marginals.all.feature.outer.prod.txt"), quote=F, sep="\t") ## find top topic that define PoPS for each gene ## function for sorting the feature x gene importance value sort_feature_x_gene_importance <- function(df) { out <- df %>% mutate(ENSGID=rownames(.)) %>% melt(id.vars = c("Gene.name", "EntrezID", "Gene", "ENSGID"), variable.name="topic", value.name="gene.feature_x_beta") %>% group_by(Gene) %>% arrange(desc(gene.feature_x_beta)) %>% as.data.frame return(out) } coefs.defining.top.topic.df <- PoPS_Score.coefs.manual.outer %>% sort_feature_x_gene_importance marginals.defining.top.topic.df <- PoPS_Score.marginals.manual.outer %>% sort_feature_x_gene_importance all.coefs.defining.top.topic.df <- PoPS_Score.coefs.all.outer %>% sort_feature_x_gene_importance all.marginals.defining.top.topic.df <- PoPS_Score.marginals.all.outer %>% sort_feature_x_gene_importance ## these are large files, so store them in $GROUP_SCRATCH saveRDS(marginals.defining.top.topic.df, file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_marginals.defining.top.topic.RDS")) saveRDS(coefs.defining.top.topic.df, file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_coefs.defining.top.topic.RDS")) saveRDS(all.marginals.defining.top.topic.df, file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_all.marginals.defining.top.topic.RDS")) saveRDS(all.coefs.defining.top.topic.df, file=paste0(SCRATCH.OUTDIR, "/", PREFIX, "_all.coefs.defining.top.topic.RDS")) save(PoPS_Score.coefs.manual.outer, PoPS_Score.marginals.manual.outer, PoPS_Score.coefs.all.outer, PoPS_Score.marginals.all.outer, coefs.defining.top.topic.df, marginals.defining.top.topic.df, all.coefs.defining.top.topic.df, all.marginals.defining.top.topic.df, file=file.name) ## } else { ## load(file.name) ## } ## output a table, one row per gene, with columns for gene symbol, PoPS score without cNMF, PoPS score with cNMF, top features from any source important for that gene, top topic features important for that gene coefs.defining.top.topic.df.subset <- coefs.defining.top.topic.df %>% subset(grepl("^ENSG",ENSGID)) %>% group_by(ENSGID) %>% arrange(desc(gene.feature_x_beta)) %>% slice(1:10) %>% as.data.frame all.coefs.defining.top.topic.df.subset <- all.coefs.defining.top.topic.df %>% subset(grepl("^ENSG",ENSGID)) %>% group_by(ENSGID) %>% arrange(desc(gene.feature_x_beta)) %>% slice(1:10) %>% as.data.frame PoPS_preds.importance.score <- merge(preds.combined.df, all.coefs.defining.top.topic.df.subset %>% select(-Gene.name, -Gene, -EntrezID), by="ENSGID") %>% merge(., metadata, by.x="topic", by.y="X", all.x=T) colnames(PoPS_preds.importance.score)[which(colnames(PoPS_preds.importance.score)=="topic")] <- "pathway" write.table(PoPS_preds.importance.score %>% apply(2, as.character), file=paste0(OUTDIR, "/", PREFIX, "_PoPS_preds.importance.score.all.columns.txt"), sep="\t", quote=F, row.names=F) PoPS_preds.importance.score.key <- PoPS_preds.importance.score %>% select(Gene, Gene.name, PoPS_Score_with.cNMF, PoPS_Score_without.cNMF, pathway, Long_Name, gene.feature_x_beta) write.table(PoPS_preds.importance.score.key %>% apply(2, as.character), file=paste0(OUTDIR, "/", PREFIX, "_PoPS_preds.importance.score.key.columns.txt"), sep="\t", quote=F, row.names=F) ## cNMF topics only gene.feature_x_beta score PoPS_preds.importance.score.cNMF <- merge(preds.combined.df, coefs.defining.top.topic.df.subset %>% select(-Gene.name, -Gene, -EntrezID), by="ENSGID") %>% merge(., metadata, by.x="topic", by.y="X", all.x=T) colnames(PoPS_preds.importance.score.cNMF)[which(colnames(PoPS_preds.importance.score.cNMF)=="topic")] <- "pathway" write.table(PoPS_preds.importance.score %>% apply(2, as.character), file=paste0(OUTDIR, "/", PREFIX, "_PoPS_preds.importance.score.all.columns.cNMF.Topics.only.txt"), sep="\t", quote=F, row.names=F) PoPS_preds.importance.score.cNMF.key <- PoPS_preds.importance.score.cNMF %>% select(Gene, Gene.name, PoPS_Score_with.cNMF, PoPS_Score_without.cNMF, pathway, Long_Name, gene.feature_x_beta) write.table(PoPS_preds.importance.score.cNMF.key %>% apply(2, as.character), file=paste0(OUTDIR, "/PoPS_preds.importance.score.key.columns.cNMF.Topics.only.txt"), sep="\t", quote=F, row.names=F) ## get top features' top genes top.features.names <- coefs.df %>% arrange(desc(beta)) %>% slice(1:10) %>% pull(parameter) top.features.definition <- all.features.cNMF %>% select(all_of(c("ENSGID",top.features.names))) %>% add_gene_name_to_df slice_top_genes_from_features <- function(df) { out <- df %>% melt(value.name="relative.importance", variable.name="features", id.vars=c("ENSGID", "Gene", "Gene.name", "EntrezID")) %>% group_by(features) %>% arrange(desc(relative.importance)) %>% slice(1:10) } top.genes.in.top.features <- top.features.definition %>% slice_top_genes_from_features %>% as.data.frame %>% merge(coefs.df, by.x="features", by.y="parameter") %>% arrange(desc(beta)) write.table(top.genes.in.top.features, file=paste0(OUTDIR, "/top.genes.in.top.features.coefs.txt"), row.names=F, quote=F, sep="\t") |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 | import numpy as np import pandas as pd import glob import argparse if __name__ == '__main__': parser = argparse.ArgumentParser(description='Converts a directory of feature files into efficient NumPy format, written out to multiple chunks, amenable for use downstream with PoPS.') parser.add_argument("--gene_annot_path", help="Path to gene annotation table. For the purposes of this script, only require that there is an ENSGID column.") parser.add_argument("--feature_dir", help="Directory where raw feature files live. Each feature file must be a tab-separated file with a header for column names and the first column must be the ENSGID. Will process every file in the directory so make sure every file is a feature file and there are no hidden files. Please also make sure the column names are unique across all feature files. The easiest way to ensure this is to prefix every column with the filename.") parser.add_argument("--nan_policy", default="raise", help="What to do if a feature file is missing ENSGIDs that are in gene_annot_path. Takes the values \"raise\" (raise an error), \"ignore\" (ignore and write out with nans), \"mean\" (impute the mean of the feature), and \"zero\" (impute 0). Default is \"raise\".") parser.add_argument("--save_prefix", help="Prefix to the output path. For each chunk i, 2 files will be written: {save_prefix}_mat.{i}.npy, {save_prefix}_cols.{i}.txt. Furthermore, row data will be written to {save_prefix}_rows.txt") parser.add_argument("--max_cols", default=5000, type=int, help="Maximum number of columns per output chunk. Default is 5000.") args = parser.parse_args() gene_annot_path = args.gene_annot_path feature_dir = args.feature_dir nan_policy = args.nan_policy save_prefix = args.save_prefix MAX_COLS = args.max_cols assert nan_policy in ["raise", "ignore", "mean", "zero"], "Invalid argument for flag --nan_policy. Accepts \"raise\", \"ignore\", \"mean\", and \"zero\"." gene_annot_df = pd.read_csv(gene_annot_path, sep="\t", index_col="ENSGID").iloc[:,0:0] row_data = gene_annot_df.index.values np.savetxt(save_prefix + ".rows.txt", row_data, fmt="%s") #### Sort for canonical ordering all_feature_files = sorted([f for f in glob.glob(feature_dir + "/*")]) all_mat_data = [] all_col_data = [] curr_block_index = 0 for f in all_feature_files: print(f) ## added by Helen for debugging f_df = pd.read_csv(f, sep="\t", index_col=0).astype(np.float64) # import pdb; pdb.set_trace(); ## added by Helen for debugging f_df = gene_annot_df.merge(f_df, how="left", left_index=True, right_index=True) if len(f_df.index[f_df.index.duplicated(keep='first')]) > 0: print("duplicated ENSGID: " + f_df.index[f_df.index.duplicated()]) f_df = f_df[~f_df.index.duplicated(keep='first')] ## added by Helen to remove duplicated ENSGID ## need to make sure if nan_policy == "raise": assert not f_df.isnull().values.any(), "Missing genes in feature matrix." elif nan_policy == "ignore": pass elif nan_policy == "mean": f_df = f_df.fillna(f_df.mean()) elif nan_policy == "zero": f_df = f_df.fillna(0) mat = f_df.values cols = f_df.columns.values all_mat_data.append(mat) all_col_data += list(cols) while len(all_col_data) >= MAX_COLS: ### Flush MAX_COLS columns to disk at a time # import pdb; pdb.set_trace(); ## added by Helen for debugging mat = np.hstack(all_mat_data) save_mat = mat[:,:MAX_COLS] keep_mat = mat[:,MAX_COLS:] save_cols = all_col_data[:MAX_COLS] keep_cols = all_col_data[MAX_COLS:] ### Save np.save(save_prefix + ".mat.{}.npy".format(curr_block_index), save_mat) np.savetxt(save_prefix + ".cols.{}.txt".format(curr_block_index), save_cols, fmt="%s") ### Update variables all_mat_data = [keep_mat] all_col_data = keep_cols curr_block_index += 1 ### Flush last block if len(all_col_data) > 0: # import pdb; pdb.set_trace(); ## added by Helen for debugging mat = np.hstack(all_mat_data) np.save(save_prefix + ".mat.{}.npy".format(curr_block_index), mat) np.savetxt(save_prefix + ".cols.{}.txt".format(curr_block_index), all_col_data, fmt="%s") |
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 | library(conflicted) conflict_prefer("first", "dplyr") conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("collapse", "dplyr") conflict_prefer("combine", "dplyr") packages <- c("optparse","dplyr", "data.table", "reshape2", "ggplot2", "tidyr", "textshape","readxl", "gplots", "AnnotationDbi", "org.Hs.eg.db", "ggrepel", "gplots") xfun::pkg_attach(packages) conflict_prefer("select", "dplyr") option.list <- list( make_option("--project", type="character", default = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211130_test_PoPS.plots/", help="project directory"), make_option("--sampleName", type="character", default="2kG.library", help="project name"), make_option("--output", type="character", default = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211130_test_PoPS.plots/outputs/", help="output directory"), make_option("--figure", type="character", default = "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211130_test_PoPS.plots/figures/", help="figure directory"), make_option("--scratch.output", type="character", default="", help="output directory for large files"), make_option("--k.val", type="numeric", default=60, help="the value of K in this run"), make_option("--PoPS_outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/", help="PoPS output directory"), make_option("--coefs_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.coefs", help=""), make_option("--marginals_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.marginals", help=""), make_option("--preds_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.preds", help=""), make_option("--coefs_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.coefs", help=""), make_option("--marginals_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.marginals", help=""), make_option("--preds_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.preds", help=""), make_option("--prefix", type="character", default="CAD_aug6_cNMF60", help="magma file name (before genes.raw)"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--cNMF.features", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/data/features/pops_features_raw/topic.zscore.ensembl.scaled_k_60.dt_0_2.txt", help="normalized cNMF weights, unit variance and zero mean"), make_option("--all.features", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211101_normalized_features/outputs/full_features_with_cNMF.RDS", help=".RDS file with all features input into PoPS"), make_option("--external.features.metadata", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/metadata/metadata_jul17.txt", help="annotations for each external features"), make_option("--combined.preds", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.combined.preds", help="preds file with results from no_cNMF run and with_cNMF run"), make_option("--coefs.defining.top.topic.RDS", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60_coefs.defining.top.topic.RDS", help=""), make_option("--preds.importance.score.key.columns", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60_preds.importance.score.key.columns.txt", help="") ) opt <- parse_args(OptionParser(option_list=option.list)) ## debug PoPS.plots.R using scRNAseq_11AMDox_1 sample opt$sampleName <- "scRNAseq_2kG_11AMDox_1" opt$output <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/" opt$figure <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/figures/all_genes/scRNAseq_2kG_11AMDox_1/K5/" opt$scratch.output <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/211101_20sample_snakemake/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/" opt$prefix <- "CAD_aug6_cNMF5" opt$k.val <- 5 opt$coefs_with_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5.coefs" opt$preds_with_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5.preds" opt$marginals_with_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5.marginals" opt$coefs_without_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/pops/CAD_aug6.coefs" opt$preds_without_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/pops/CAD_aug6.preds" opt$marginals_without_cNMF <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5.marginals" opt$cNMF.features <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/topic.zscore.ensembl.scaled_k_5.dt_0_2.txt" opt$all.features <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/full_features_CAD_aug6_cNMF5.RDS" opt$external.features.metadata <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/metadata/metadata_jul17.txt" opt$combined.preds <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5.combined.preds" opt$coefs.defining.top.topic.RDS <- "/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/211101_20sample_snakemake/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5_coefs.defining.top.topic.RDS" opt$preds.importance.score.key.columns <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211101_20sample_snakemake/analysis/all_genes/scRNAseq_2kG_11AMDox_1/K5/threshold_0_2/pops/CAD_aug6_cNMF5_PoPS_preds.importance.score.key.columns.txt" SAMPLE=opt$sampleName OUTDIR=opt$output SCRATCH.OUTDIR=opt$scratch.output FIGDIR=opt$figure PREFIX=opt$prefix DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) DENSITY.THRESHOLD.FILENAME =paste0("dt_", DENSITY.THRESHOLD) k <- opt$k.val check.dir <- c(OUTDIR, FIGDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ## graphing constants mytheme <- theme_classic() + theme(axis.text = element_text(size = 12), axis.title = element_text(size = 14), plot.title = element_text(hjust = 0.5, face = "bold", size=14)) palette = colorRampPalette(c("#38b4f7", "white", "red"))(n = 100) ## load metadata meta.data.path <- opt$external.features.metadata metadata <- read.delim(meta.data.path, stringsAsFactors=F) ## load data preds <- read.table(file=opt$preds_with_cNMF,header=T, stringsAsFactors=F, sep="\t") colnames(preds) <- paste0(colnames(preds), "_with.cNMF") colnames(preds)[1] <- "ENSGID" preds.before <- read.table(file=paste0(opt$preds_without_cNMF), header=T, stringsAsFactors=F, sep="\t") colnames(preds.before) <- paste0(colnames(preds.before), "_without.cNMF") colnames(preds.before)[1] <- "ENSGID" preds.combined <- merge(preds, preds.before, by="ENSGID") marginals <- read.table(file=opt$marginals_with_cNMF,header=T, stringsAsFactors=F, sep="\t") %>% merge(metadata, by="X") coefs <- read.table(file=opt$coefs_with_cNMF,header=T, stringsAsFactors=F, sep="\t") coefs.df <- coefs[4:nrow(coefs),] %>% merge(metadata, by.x="parameter", by.y="X") %>% arrange(desc(beta)) coefs.df$beta <- coefs.df$beta %>% as.numeric ## map ids x <- org.Hs.egENSEMBL mapped_genes <- mappedkeys(x) xx.entrez.to.ensembl <- as.list(x[mapped_genes]) # EntrezID to Ensembl xx.ensembl.to.entrez <- as.list(org.Hs.egENSEMBL2EG) # Ensembl to EntrezID y <- org.Hs.egGENENAME y_mapped_genes <- mappedkeys(y) entrez.to.genename <- as.list(y[y_mapped_genes]) genename.to.entrez <- as.list(org.Hs.egGENENAME) z <- org.Hs.egSYMBOL z_mapped_genes <- mappedkeys(z) entrez.to.symbol <- as.list(z[z_mapped_genes]) symbol.to.entrez <- as.list(org.Hs.egSYMBOL) ## preds.df preds.df <- preds %>% mutate(EntrezID = xx.ensembl.to.entrez[preds$ENSGID %>% as.character] %>% sapply("[[",1)) %>% mutate(Gene.name = entrez.to.genename[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character) ## load all preds preds.combined.df <- read.delim(file=opt$combined.preds, stringsAsFactors=F) ## load cNMF features features <- read.delim(opt$cNMF.features, stringsAsFactors=F) ## load all features all.features <- readRDS(opt$all.features) features <- read.delim(opt$cNMF.features, stringsAsFactors=F) # ## load gene x feature importance score # load(file=paste0(OUTDIR, "/coefs.marginals.feature.outer.prod.RDS")) ## load coefs.defining.top.topic.df coefs.defining.top.topic.df <- readRDS(file=opt$coefs.defining.top.topic.RDS) # paste0(SCRATCH.OUTDIR, "/", PREFIX, "_coefs.defining.top.topic.RDS")) ## load PoPS importance score with key columns table PoPS_preds.importance.score.key <- read.delim(file=opt$preds.importance.score.key.columns, stringsAsFactors=F) #paste0(OUTDIR, "/PoPS_preds.importance.score.key.columns.txt"), stringsAsFactors = F) ################################################## ## Plots ## plot the list of topics and their PoPS component scores for gene of interest gene.set <- c("GOSR2", "TLNRD1", "EDN1", "NOS3", "KLF2", "ERG", "CCM2", "KRIT") pdf(paste0(FIGDIR, "/", PREFIX, "_", DENSITY.THRESHOLD.FILENAME, "_feature_x_gene.component.importance.score.coefs.pdf"), width=4, height=6) for (gene.here in gene.set) { toPlot <- coefs.defining.top.topic.df %>% subset(grepl(paste0("^",gene.here,"$"), Gene)) %>% select(topic, gene.feature_x_beta) p <- toPlot %>% ggplot(aes(x=reorder(topic, gene.feature_x_beta), y=gene.feature_x_beta)) + geom_col(fill="#38b4f7") + theme_minimal() + coord_flip() + xlab("Feature (Topic)") + ylab("Feature x Gene\nImportance Score") + ggtitle(paste0(gene.here)) + mytheme print(p) toPlot <- PoPS_preds.importance.score.key %>% subset(grepl(paste0("^",gene.here,"$"), Gene)) %>% select(Long_Name, gene.feature_x_beta) %>% slice(1:15) p <- toPlot %>% ggplot(aes(x=reorder(Long_Name, gene.feature_x_beta), y=gene.feature_x_beta)) + geom_col(fill="#38b4f7") + theme_minimal() + coord_flip() + xlab("Features") + ylab("Feature x Gene\nImportance Score") + ggtitle(paste0(gene.here)) + mytheme } dev.off() pdf(paste0(FIGDIR, "/", PREFIX, "_", DENSITY.THRESHOLD.FILENAME, "_Topic.coef.beta.pdf"), width=4, height=6) ## double check figures toPlot <- coefs.df %>% subset(grepl("zscore",parameter)) %>% arrange(desc(beta)) p <- toPlot %>% ggplot(aes(x=reorder(parameter, beta), y=beta)) + geom_col(fill="#38b4f7") + theme_minimal() + coord_flip() + xlab("Features (topic)") + ylab("Beta Score") + ggtitle(paste0("K = ", k, " Topics")) + mytheme print(p) dev.off() pdf(paste0(FIGDIR, "/all.coef.beta.pdf")) toPlot <- coefs.df %>% arrange(desc(beta)) %>% slice(1:15) p <- toPlot %>% ggplot(aes(x=reorder(Long_Name, beta), y=beta)) + geom_col(fill="#38b4f7") + theme_minimal() + coord_flip() + xlab("Features") + ylab("Beta Score") + ggtitle(paste0("K = ", k, " Topics")) + mytheme print(p) dev.off() pdf(paste0(FIGDIR, "/all.marginals.beta.pdf")) toPlot <- marginals %>% arrange(desc(beta)) %>% slice(1:15) p <- toPlot %>% ggplot(aes(x=reorder(Long_Name, beta), y=beta)) + geom_col(fill="#38b4f7") + theme_minimal() + coord_flip() + xlab("Features") + ylab("Beta Score") + ggtitle(paste0("K = ", k, " Topics, Ranked by marginals beta score")) + mytheme print(p) dev.off() ## figure pdf(paste0(FIGDIR, "/", PREFIX, "_", DENSITY.THRESHOLD.FILENAME, "_PoPS_score_list.pdf"), width=4, height=6) top.PoPS.genes <- preds.df %>% slice(1:20) toPlot <- data.frame(Gene= top.PoPS.genes %>% pull(Gene), Score=top.PoPS.genes %>% pull(PoPS_Score_with.cNMF)) p <- toPlot %>% ggplot(aes(x=reorder(Gene, Score), y=Score) ) + geom_col(fill="#38b4f7") + theme_minimal() p <- p + coord_flip() + xlab("Top 50 Genes") + ylab("PoPS Score") + ggtitle(paste0(SAMPLE, ", ", PREFIX)) + mytheme print(p) dev.off() ## PoPS before vs after score pdf(paste0(FIGDIR, "/", PREFIX, "_", DENSITY.THRESHOLD.FILENAME, "_before.vs.after.cNMF.pdf")) labels <- preds.combined.df %>% subset((PoPS_Score_with.cNMF > (2 * PoPS_Score_without.cNMF)) & PoPS_Score_with.cNMF > 1) p <- preds.combined.df %>% ggplot(aes(x=PoPS_Score_without.cNMF, y=PoPS_Score_with.cNMF)) + geom_point(size=0.5) + mytheme + xlab("PoPS Score (without cNMF features)") + ylab("PoPS Score(with cNMF features)") + geom_abline(slope=1, color="red") + geom_text_repel(data=labels, box.padding = 0.5, max.overlaps=30, aes(label=Gene), size=4, color="blue") print(p) dev.off() |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 | import pandas as pd import numpy as np import re import scipy.linalg import random import logging import argparse from sklearn.linear_model import LinearRegression, RidgeCV, LassoCV from sklearn.metrics import make_scorer from scipy.sparse import load_npz from numpy.linalg import LinAlgError ### --------------------------------- PROGRAM INPUTS --------------------------------- ### def get_pops_args(argv=None): parser = argparse.ArgumentParser(description='...') parser.add_argument("--gene_annot_path", help="...") parser.add_argument("--feature_mat_prefix", help="...") parser.add_argument("--num_feature_chunks", type=int, help="...") parser.add_argument("--magma_prefix", help="...") parser.add_argument('--use_magma_covariates', dest='use_magma_covariates', action='store_true') parser.add_argument('--ignore_magma_covariates', dest='use_magma_covariates', action='store_false') parser.set_defaults(use_magma_covariates=True) parser.add_argument('--use_magma_error_cov', dest='use_magma_error_cov', action='store_true') parser.add_argument('--ignore_magma_error_cov', dest='use_magma_error_cov', action='store_false') parser.set_defaults(use_magma_error_cov=True) parser.add_argument("--y_path", help="...") parser.add_argument("--y_covariates_path", help="...") parser.add_argument("--y_error_cov_path", help="...") parser.add_argument("--project_out_covariates_chromosomes", nargs="*", help="...") parser.add_argument('--project_out_covariates_remove_hla', dest='project_out_covariates_remove_hla', action='store_true') parser.add_argument('--project_out_covariates_keep_hla', dest='project_out_covariates_remove_hla', action='store_false') parser.set_defaults(project_out_covariates_remove_hla=True) parser.add_argument("--subset_features_path", help="...") parser.add_argument("--control_features_path", help="...") parser.add_argument("--feature_selection_chromosomes", nargs="*", help="...") parser.add_argument("--feature_selection_p_cutoff", type=float, default=0.05, help="...") parser.add_argument("--feature_selection_max_num", type=int, help="...") parser.add_argument("--feature_selection_fss_num_features", type=int, help="...") parser.add_argument('--feature_selection_remove_hla', dest='feature_selection_remove_hla', action='store_true') parser.add_argument('--feature_selection_keep_hla', dest='feature_selection_remove_hla', action='store_false') parser.set_defaults(feature_selection_remove_hla=True) parser.add_argument("--training_chromosomes", nargs="*", help="...") parser.add_argument('--training_remove_hla', dest='training_remove_hla', action='store_true') parser.add_argument('--training_keep_hla', dest='training_remove_hla', action='store_false') parser.set_defaults(training_remove_hla=True) parser.add_argument("--method", default="ridge", help="...") parser.add_argument("--out_prefix", help="...") parser.add_argument('--save_matrix_files', dest='save_matrix_files', action='store_true') parser.add_argument('--no_save_matrix_files', dest='save_matrix_files', action='store_false') parser.set_defaults(save_matrix_files=False) parser.add_argument("--random_seed", type=int, default=42, help="...") parser.add_argument('--verbose', dest='verbose', action='store_true') parser.add_argument('--no_verbose', dest='verbose', action='store_false') parser.set_defaults(verbose=False) return parser.parse_args(argv) ### --------------------------------- GENERAL --------------------------------- ### def natural_key(string_): """See https://blog.codinghorror.com/sorting-for-humans-natural-sort-order/""" return [int(s) if s.isdigit() else s for s in re.split(r'(\d+)', string_)] def get_hla_genes(gene_annot_df): sub_gene_annot_df = gene_annot_df[gene_annot_df.CHR == "6"] sub_gene_annot_df = sub_gene_annot_df[sub_gene_annot_df.TSS >= 20 * (10 ** 6)] sub_gene_annot_df = sub_gene_annot_df[sub_gene_annot_df.TSS <= 40 * (10 ** 6)] return sub_gene_annot_df.index.values ### Returns as vector of booleans of length len(Y_ids) def get_gene_indices_to_use(Y_ids, gene_annot_df, use_chrs, remove_hla): all_chr_genes_set = set(gene_annot_df[gene_annot_df.CHR.isin(use_chrs)].index.values) if remove_hla == True: hla_genes_set = set(get_hla_genes(gene_annot_df)) use_genes = [True if (g in all_chr_genes_set) and (g not in hla_genes_set) else False for g in Y_ids] else: use_genes = [True if g in all_chr_genes_set else False for g in Y_ids] return np.array(use_genes) def get_indices_in_target_order(ref_list, target_names): ref_to_ind_mapper = {} for i, e in enumerate(ref_list): ref_to_ind_mapper[e] = i return np.array([ref_to_ind_mapper[t] for t in target_names]) ### --------------------------------- READING DATA --------------------------------- ### def read_gene_annot_df(gene_annot_path): gene_annot_df = pd.read_csv(gene_annot_path, delim_whitespace=True).set_index("ENSGID") gene_annot_df["CHR"] = gene_annot_df["CHR"].astype(str) return gene_annot_df def read_magma(magma_prefix, use_magma_covariates, use_magma_error_cov): ### Get Y and Y_ids magma_df = pd.read_csv(magma_prefix + ".genes.out", delim_whitespace=True) Y = magma_df.ZSTAT.values Y_ids = magma_df.GENE.values if use_magma_covariates is not None or use_magma_error_cov is not None: ### Get covariates and error_cov sigmas, gene_metadata = munge_magma_covariance_metadata(magma_prefix + ".genes.raw") cov_df = build_control_covariates(gene_metadata) ### Process assert (cov_df.index.values == Y_ids).all(), "Covariate ids and Y ids don't match." covariates = cov_df.values error_cov = scipy.linalg.block_diag(*sigmas) if use_magma_covariates == False: covariates = None if use_magma_error_cov == False: error_cov = None return Y, covariates, error_cov, Y_ids def munge_magma_covariance_metadata(magma_raw_path): sigmas = [] gene_metadata = [] with open(magma_raw_path) as f: ### Get all lines lines = list(f)[2:] lines = [np.asarray(line.strip('\n').split(' ')) for line in lines] ### Check that chromosomes are sequentially ordered all_chroms = np.array([l[1] for l in lines]) all_seq_breaks = np.where(all_chroms[:-1] != all_chroms[1:])[0] assert len(all_seq_breaks) == len(set(all_chroms)) - 1, "Chromosomes are not sequentially ordered." ### Get starting chromosome and set up temporary variables curr_chrom = lines[0][1] curr_ind = 0 num_genes_in_chr = sum([1 for line in lines if line[1] == curr_chrom]) curr_sigma = np.zeros((num_genes_in_chr, num_genes_in_chr)) curr_gene_metadata = [] for line in lines: ### If we move to a new chromosome, we reset everything if line[1] != curr_chrom: ### Symmetrize and save sigmas.append(curr_sigma + curr_sigma.T + np.eye(curr_sigma.shape[0])) gene_metadata.append(curr_gene_metadata) ### Reset curr_chrom = line[1] curr_ind = 0 num_genes_in_chr = sum([1 for line in lines if line[1] == curr_chrom]) curr_sigma = np.zeros((num_genes_in_chr, num_genes_in_chr)) curr_gene_metadata = [] ### Add metadata; GENE, NSNPS, NPARAM, MAC curr_gene_metadata.append([line[0], float(line[4]), float(line[5]), float(line[7])]) if len(line) > 9: ### Add covariance gene_corrs = np.array([float(c) for c in line[9:]]) curr_sigma[curr_ind, curr_ind - gene_corrs.shape[0]:curr_ind] = gene_corrs curr_ind += 1 ### Save last piece sigmas.append(curr_sigma + curr_sigma.T + np.eye(curr_sigma.shape[0])) gene_metadata.append(curr_gene_metadata) gene_metadata = pd.DataFrame(np.vstack(gene_metadata), columns=["GENE", "NSNPS", "NPARAM", "MAC"]) gene_metadata.NSNPS = gene_metadata.NSNPS.astype(np.float64) gene_metadata.NPARAM = gene_metadata.NPARAM.astype(np.float64) gene_metadata.MAC = gene_metadata.MAC.astype(np.float64) return sigmas, gene_metadata def build_control_covariates(metadata): genesize = metadata.NPARAM.values genedensity = metadata.NPARAM.values/metadata.NSNPS.values inverse_mac = 1.0/metadata.MAC.values cov = np.stack((genesize, np.log(genesize), genedensity, np.log(genedensity), inverse_mac, np.log(inverse_mac)), axis=1) cov_df = pd.DataFrame(cov, columns=["gene_size", "log_gene_size", "gene_density", "log_gene_density", "inverse_mac", "log_inverse_mac"]) cov_df["GENE"] = metadata.GENE.values cov_df = cov_df.loc[:,["GENE", "gene_size", "log_gene_size", "gene_density", "log_gene_density", "inverse_mac", "log_inverse_mac"]] cov_df = cov_df.set_index("GENE") return cov_df def read_error_cov_from_y(y_error_cov_path, Y_ids): ### Will try to read in as a: scipy sparse .npz, numpy .npy error_cov = None try: error_cov = load_npz(y_error_cov_path) error_cov = np.array(error_cov.todense()) except AttributeError as ev: error_cov = np.load(y_error_cov_path) if error_cov is None: raise IOError("Error reading from {}. Make sure data is in scipy .npz or numpy .npy format.".format(y_error_cov_path)) assert error_cov.shape[0] == error_cov.shape[1], "Error covariance is not square." assert error_cov.shape[0] == len(Y_ids), "Error covariance does not match dimensions of Y." return error_cov def read_from_y(y_path, y_covariates_path, y_error_cov_path): ### Get Y and Y_ids y_df = pd.read_csv(y_path, sep="\t") Y = y_df.Score.values Y_ids = y_df.ENSGID.values ### Read in covariates and error_cov covariates = None error_cov = None if y_covariates_path is not None: covariates = pd.read_csv(y_covariates_path, sep="\t", index_col="ENSGID").astype(np.float64) covariates = covariates.loc[Y_ids].values if y_error_cov_path is not None: error_cov = read_error_cov_from_y(y_error_cov_path, Y_ids) return Y, covariates, error_cov, Y_ids ### --------------------------------- PROCESSING DATA --------------------------------- ### def block_Linv(A, block_labels): block_labels = np.array(block_labels) Linv = np.zeros(A.shape) for l in set(block_labels): subset_ind = (block_labels == l) sub_A = A[np.ix_(subset_ind, subset_ind)] Linv[np.ix_(subset_ind, subset_ind)] = np.linalg.inv(np.linalg.cholesky(sub_A)) return Linv def block_AB(A, block_labels, B): block_labels = np.array(block_labels) new_B = np.zeros(B.shape) for l in set(block_labels): subset_ind = (block_labels == l) new_B[subset_ind] = A[np.ix_(subset_ind, subset_ind)].dot(B[subset_ind]) return new_B def block_BA(A, block_labels, B): block_labels = np.array(block_labels) new_B = np.zeros(B.shape) for l in set(block_labels): subset_ind = (block_labels == l) new_B[:,subset_ind] = B[:,subset_ind].dot(A[np.ix_(subset_ind, subset_ind)]) return new_B def regularize_error_cov(error_cov, Y, Y_ids, gene_annot_df): Y_chr = gene_annot_df.loc[Y_ids].CHR.values min_lambda = 0 for c in set(Y_chr): subset_ind = Y_chr == c W = np.linalg.eigvalsh(error_cov[np.ix_(subset_ind, subset_ind)]) min_lambda = min(min_lambda, min(W)) ridge = abs(min(min_lambda, 0))+.05+.9*max(0, np.var(Y)-1) return error_cov + np.eye(error_cov.shape[0]) * ridge def project_out_covariates(Y, covariates, error_cov, Y_ids, gene_annot_df, project_out_covariates_Y_gene_inds): ### If covariates doesn't contain intercept, add intercept if not np.isclose(covariates.var(axis=0), 0).any(): covariates = np.hstack((covariates, np.ones((covariates.shape[0], 1)))) X_train, y_train = covariates[project_out_covariates_Y_gene_inds], Y[project_out_covariates_Y_gene_inds] if error_cov is not None: sub_error_cov = error_cov[np.ix_(project_out_covariates_Y_gene_inds, project_out_covariates_Y_gene_inds)] sub_error_cov_labels = gene_annot_df.loc[Y_ids[project_out_covariates_Y_gene_inds]].CHR.values Linv = block_Linv(sub_error_cov, sub_error_cov_labels) X_train, y_train = block_AB(Linv, sub_error_cov_labels, X_train), block_AB(Linv, sub_error_cov_labels, y_train) reg = LinearRegression(fit_intercept=False).fit(X_train, y_train) Y_proj = Y - reg.predict(covariates) return Y_proj def project_out_V(M, V): gram_inv = np.linalg.inv(V.T.dot(V)) moment = V.T.dot(M) betas = gram_inv.dot(moment) M_res = M - V.dot(betas) return M_res ### --------------------------------- FEATURE SELECTION --------------------------------- ### def batch_marginal_ols(Y, X): ### Save current error settings and set divide to ignore old_settings = np.seterr(divide='ignore') ### Does not include intercept; we assume that's been projected out already sum_sq_X = np.sum(np.square(X), axis=0) ### If near-constant to 0 then set to nan. Make a safe copy so we don't get divide by 0 errors. near_const_0 = np.isclose(sum_sq_X, 0) sum_sq_X_safe = sum_sq_X.copy() sum_sq_X_safe[near_const_0] = 1 betas = Y.dot(X) / sum_sq_X_safe mse = np.mean(np.square(Y.reshape(-1,1) - X * betas), axis=0) se = np.sqrt(mse / sum_sq_X_safe) z = betas / se chi2 = np.square(z) pvals = scipy.stats.chi2.sf(chi2, 1) r2 = 1 - (mse / np.var(Y)) ### Set everything that's near-constant to 0 to be nan betas[near_const_0] = np.nan se[near_const_0] = np.nan pvals[near_const_0] = np.nan r2[near_const_0] = np.nan ### Reset error settings to old np.seterr(**old_settings) return betas, se, pvals, r2 ### Accepts covariates, error_cov = None def compute_marginal_assoc(feature_mat_prefix, num_feature_chunks, Y, Y_ids, covariates, error_cov, gene_annot_df, feature_selection_Y_gene_inds): ### Get Y data feature_selection_genes = Y_ids[feature_selection_Y_gene_inds] sub_Y = Y[feature_selection_Y_gene_inds] ### Add intercept if no near-constant feature if covariates is not None and not np.isclose(covariates.var(axis=0), 0).any(): covariates = np.hstack((covariates, np.ones((covariates.shape[0], 1)))) elif covariates is None: ### If no covariates then make intercept as only covariate covariates = np.ones((Y.shape[0], 1)) sub_covariates = covariates[feature_selection_Y_gene_inds] if error_cov is not None: sub_error_cov = error_cov[np.ix_(feature_selection_Y_gene_inds, feature_selection_Y_gene_inds)] sub_error_cov_labels = gene_annot_df.loc[feature_selection_genes].CHR.values Linv = block_Linv(sub_error_cov, sub_error_cov_labels) sub_Y = block_AB(Linv, sub_error_cov_labels, sub_Y) sub_covariates = block_AB(Linv, sub_error_cov_labels, sub_covariates) ### Project covariates out of sub_Y sub_Y = project_out_V(sub_Y.reshape(-1,1), sub_covariates).flatten() ### Get X training indices rows = np.loadtxt(feature_mat_prefix + ".rows.txt", dtype=str).flatten() X_train_inds = get_indices_in_target_order(rows, feature_selection_genes) ### Loop through and get marginal association data marginal_assoc_data = [] all_cols = [] for i in range(num_feature_chunks): mat = np.load(feature_mat_prefix + ".mat.{}.npy".format(i)) mat = mat[X_train_inds] cols = np.loadtxt(feature_mat_prefix + ".cols.{}.txt".format(i), dtype=str).flatten() ### Apply error covariance transformation if available if error_cov is not None: mat = block_AB(Linv, sub_error_cov_labels, mat) ### Project out covariates mat = project_out_V(mat, sub_covariates) ### Compute marginal associations marginal_assoc_data.append(np.vstack(batch_marginal_ols(sub_Y, mat)).T) all_cols.append(cols) marginal_assoc_data = np.vstack(marginal_assoc_data) all_cols = np.hstack(all_cols) marginal_assoc_df = pd.DataFrame(marginal_assoc_data, columns=["beta", "se", "pval", "r2"], index=all_cols) return marginal_assoc_df ### Note that subset_features overrides control_features. ### That is: we do not include control features that are not contained in subset features ### Also, control features do not count toward feature_selection_max_num def select_features_from_marginal_assoc_df(marginal_assoc_df, subset_features_path, control_features_path, feature_selection_p_cutoff, feature_selection_max_num): ### Subset to subset_features if subset_features_path is not None: subset_features = np.loadtxt(subset_features_path, dtype=str).flatten() marginal_assoc_df = marginal_assoc_df.loc[subset_features] ### Get control_features contained in currently subsetted features, and set those aside if control_features_path is not None: control_features = np.loadtxt(control_features_path, dtype=str).flatten() control_df = marginal_assoc_df[marginal_assoc_df.index.isin(control_features)] marginal_assoc_df = marginal_assoc_df[~marginal_assoc_df.index.isin(control_features)] ### Subset to features that pass p-value cutoff if feature_selection_p_cutoff is not None: marginal_assoc_df = marginal_assoc_df[marginal_assoc_df.pval < feature_selection_p_cutoff] ### Enforce maximum number of features if feature_selection_max_num is not None: marginal_assoc_df = marginal_assoc_df.sort_values("pval").iloc[:feature_selection_max_num] ### Get selected features selected_features = list(marginal_assoc_df.index.values) ### Combine with control features if control_features_path is not None: selected_features = selected_features + list(control_df.index.values) return selected_features def load_feature_matrix(feature_mat_prefix, num_feature_chunks, selected_features): if selected_features is not None: selected_features_set = set(selected_features) rows = np.loadtxt(feature_mat_prefix + ".rows.txt", dtype=str).flatten() all_mats = [] all_cols = [] for i in range(num_feature_chunks): mat = np.load(feature_mat_prefix + ".mat.{}.npy".format(i)) cols = np.loadtxt(feature_mat_prefix + ".cols.{}.txt".format(i), dtype=str).flatten() if selected_features is not None: keep_inds = [True if c in selected_features_set else False for c in cols] mat = mat[:,keep_inds] cols = cols[keep_inds] all_mats.append(mat) all_cols.append(cols) mat = np.hstack(all_mats) cols = np.hstack(all_cols) return mat, cols, rows def add_feature_to_covariates(covariates, covariates_ids, feature_mat_prefix, num_feature_chunks, feature_name): ### Get X indices rows = np.loadtxt(feature_mat_prefix + ".rows.txt", dtype=str).flatten() X_inds = get_indices_in_target_order(rows, covariates_ids) for i in range(num_feature_chunks): cols = np.loadtxt(feature_mat_prefix + ".cols.{}.txt".format(i), dtype=str).flatten() if feature_name in cols: mat = np.load(feature_mat_prefix + ".mat.{}.npy".format(i))[X_inds] f = mat[:,np.where(cols == feature_name)[0]] break covariates = np.hstack((covariates, f)) return covariates def forward_stepwise_selection(feature_mat_prefix, num_feature_chunks, Y, Y_ids, covariates, error_cov, gene_annot_df, feature_selection_Y_gene_inds, num_features_to_select): if covariates is None: covariates = np.ones((Y.shape[0], 1)) selected_features = [] for i in range(num_features_to_select): logging.info("FORWARD STEPWISE SELECTION: {} features selected".format(len(selected_features))) marginal_assoc_df = compute_marginal_assoc(feature_mat_prefix, num_feature_chunks, Y, Y_ids, covariates, error_cov, gene_annot_df, feature_selection_Y_gene_inds) top_feature = marginal_assoc_df[~marginal_assoc_df.index.isin(selected_features)].sort_values("pval").index.values[0] selected_features.append(top_feature) covariates = add_feature_to_covariates(covariates, Y_ids, feature_mat_prefix, num_feature_chunks, top_feature) return selected_features ### --------------------------------- MODEL FITTING --------------------------------- ### def build_training(mat, cols, rows, Y, Y_ids, error_cov, gene_annot_df, training_Y_gene_inds, project_out_intercept=True): ### Get training Y training_genes = Y_ids[training_Y_gene_inds] sub_Y = Y[training_Y_gene_inds] intercept = np.ones((sub_Y.shape[0], 1)) ### Make intercept ### Get training X X_train_inds = get_indices_in_target_order(rows, training_genes) X = mat[X_train_inds] assert (rows[X_train_inds] == training_genes).all(), "Something went wrong. This shouldn't happen." ### Apply error covariance if error_cov is not None: sub_error_cov = error_cov[np.ix_(training_Y_gene_inds, training_Y_gene_inds)] sub_error_cov_labels = gene_annot_df.loc[training_genes].CHR.values Linv = block_Linv(sub_error_cov, sub_error_cov_labels) sub_Y = block_AB(Linv, sub_error_cov_labels, sub_Y) X = block_AB(Linv, sub_error_cov_labels, X) intercept = block_AB(Linv, sub_error_cov_labels, intercept) if project_out_intercept == True: ### Project out intercept sub_Y = project_out_V(sub_Y.reshape(-1,1), intercept).flatten() X = project_out_V(X, intercept) return X, sub_Y # def corr_score(Y, Y_pred): # score = scipy.stats.pearsonr(Y, Y_pred)[0] # return score def initialize_regressor(method, random_state): # scorer = make_scorer(corr_score) if method == "ridge": alphas = np.logspace(-2, 10, num=25) # reg = RidgeCV(fit_intercept=False, alphas=alphas, scoring=scorer) # logging.info("Model = RidgeCV with 25 alphas, generalized leave-one-out cross-validation, held-out Pearson correlation as scoring metric.") reg = RidgeCV(fit_intercept=False, alphas=alphas) logging.info("Model = RidgeCV with 25 alphas, generalized leave-one-out cross-validation, NMSE as scoring metric.") elif method == 'lasso': alphas = np.logspace(-2, 10, num=25) reg = LassoCV(fit_intercept=False, alphas=alphas, random_state=random_state, selection="random") logging.info("Model = LassoCV with 25 alphas, 5-fold cross-validation, mean-squared error as scoring metric.") elif method == 'linreg': ### Note that this solves using pseudo-inverse if # features > # samples, corresponding to minimum norm OLS reg = LinearRegression(fit_intercept=False) logging.info("Model = LinearRegression. Note that this solves using the pseudo-inverse if # features > # samples, corresponding to minimum norm OLS.") return reg ### A custom function to replace sklearn RidgeCV solver if needed. Solves using gesvd instead of gesdd def _svd_decompose_design_matrix_custom(self, X, y, sqrt_sw): # X already centered X_mean = np.zeros(X.shape[1], dtype=X.dtype) if self.fit_intercept: # to emulate fit_intercept=True situation, add a column # containing the square roots of the sample weights # by centering, the other columns are orthogonal to that one intercept_column = sqrt_sw[:, None] X = np.hstack((X, intercept_column)) U, singvals, _ = scipy.linalg.svd(X, full_matrices=0, lapack_driver="gesvd") singvals_sq = singvals ** 2 UT_y = np.dot(U.T, y) return X_mean, singvals_sq, U, UT_y ### Original function in _RidgeGCV def _svd_decompose_design_matrix_original(self, X, y, sqrt_sw): # X already centered X_mean = np.zeros(X.shape[1], dtype=X.dtype) if self.fit_intercept: # to emulate fit_intercept=True situation, add a column # containing the square roots of the sample weights # by centering, the other columns are orthogonal to that one intercept_column = sqrt_sw[:, None] X = np.hstack((X, intercept_column)) U, singvals, _ = scipy.linalg.svd(X, full_matrices=0) singvals_sq = singvals ** 2 UT_y = np.dot(U.T, y) return X_mean, singvals_sq, U, UT_y ### A custom function to replace sklearn LinearRegression fit if needed. Solves using gelss def _linear_regression_fit_custom(self, X, y, sample_weight=None): ### Importing all the base functions needed to run the monkey-patched solver from sklearn.linear_model._base import _check_sample_weight, _rescale_data, Parallel, delayed, optimize, sp, sparse, sparse_lsqr, linalg n_jobs_ = self.n_jobs accept_sparse = False if self.positive else ['csr', 'csc', 'coo'] X, y = self._validate_data(X, y, accept_sparse=accept_sparse, y_numeric=True, multi_output=True) if sample_weight is not None: sample_weight = _check_sample_weight(sample_weight, X, dtype=X.dtype) X, y, X_offset, y_offset, X_scale = self._preprocess_data( X, y, fit_intercept=self.fit_intercept, normalize=self.normalize, copy=self.copy_X, sample_weight=sample_weight, return_mean=True) if sample_weight is not None: # Sample weight can be implemented via a simple rescaling. X, y = _rescale_data(X, y, sample_weight) if self.positive: if y.ndim < 2: self.coef_, self._residues = optimize.nnls(X, y) else: # scipy.optimize.nnls cannot handle y with shape (M, K) outs = Parallel(n_jobs=n_jobs_)( delayed(optimize.nnls)(X, y[:, j]) for j in range(y.shape[1])) self.coef_, self._residues = map(np.vstack, zip(*outs)) elif sp.issparse(X): X_offset_scale = X_offset / X_scale def matvec(b): return X.dot(b) - b.dot(X_offset_scale) def rmatvec(b): return X.T.dot(b) - X_offset_scale * np.sum(b) X_centered = sparse.linalg.LinearOperator(shape=X.shape, matvec=matvec, rmatvec=rmatvec) if y.ndim < 2: out = sparse_lsqr(X_centered, y) self.coef_ = out[0] self._residues = out[3] else: # sparse_lstsq cannot handle y with shape (M, K) outs = Parallel(n_jobs=n_jobs_)( delayed(sparse_lsqr)(X_centered, y[:, j].ravel()) for j in range(y.shape[1])) self.coef_ = np.vstack([out[0] for out in outs]) self._residues = np.vstack([out[3] for out in outs]) else: self.coef_, self._residues, self.rank_, self.singular_ = \ linalg.lstsq(X, y, lapack_driver="gelss") self.coef_ = self.coef_.T if y.ndim == 1: self.coef_ = np.ravel(self.coef_) self._set_intercept(X_offset, y_offset, X_scale) return self ### Original function in LinearRegression def _linear_regression_fit_original(self, X, y, sample_weight=None): ### Importing all the base functions needed to run the monkey-patched solver from sklearn.linear_model._base import _check_sample_weight, _rescale_data, Parallel, delayed, optimize, sp, sparse, sparse_lsqr, linalg n_jobs_ = self.n_jobs accept_sparse = False if self.positive else ['csr', 'csc', 'coo'] X, y = self._validate_data(X, y, accept_sparse=accept_sparse, y_numeric=True, multi_output=True) if sample_weight is not None: sample_weight = _check_sample_weight(sample_weight, X, dtype=X.dtype) X, y, X_offset, y_offset, X_scale = self._preprocess_data( X, y, fit_intercept=self.fit_intercept, normalize=self.normalize, copy=self.copy_X, sample_weight=sample_weight, return_mean=True) if sample_weight is not None: # Sample weight can be implemented via a simple rescaling. X, y = _rescale_data(X, y, sample_weight) if self.positive: if y.ndim < 2: self.coef_, self._residues = optimize.nnls(X, y) else: # scipy.optimize.nnls cannot handle y with shape (M, K) outs = Parallel(n_jobs=n_jobs_)( delayed(optimize.nnls)(X, y[:, j]) for j in range(y.shape[1])) self.coef_, self._residues = map(np.vstack, zip(*outs)) elif sp.issparse(X): X_offset_scale = X_offset / X_scale def matvec(b): return X.dot(b) - b.dot(X_offset_scale) def rmatvec(b): return X.T.dot(b) - X_offset_scale * np.sum(b) X_centered = sparse.linalg.LinearOperator(shape=X.shape, matvec=matvec, rmatvec=rmatvec) if y.ndim < 2: out = sparse_lsqr(X_centered, y) self.coef_ = out[0] self._residues = out[3] else: # sparse_lstsq cannot handle y with shape (M, K) outs = Parallel(n_jobs=n_jobs_)( delayed(sparse_lsqr)(X_centered, y[:, j].ravel()) for j in range(y.shape[1])) self.coef_ = np.vstack([out[0] for out in outs]) self._residues = np.vstack([out[3] for out in outs]) else: self.coef_, self._residues, self.rank_, self.singular_ = \ linalg.lstsq(X, y) self.coef_ = self.coef_.T if y.ndim == 1: self.coef_ = np.ravel(self.coef_) self._set_intercept(X_offset, y_offset, X_scale) return self def compute_coefficients(X_train, Y_train, cols, method, random_state): if method not in ["ridge", "lasso", "linreg"]: raise ValueError("Invalid argument for \"method\". Must be one of \"ridge\", \"lasso\", or \"linreg\".") reg = initialize_regressor(method, random_state) logging.info("Fitting model.") try: reg.fit(X_train, Y_train) except LinAlgError as err: if method == "ridge": logging.warning(("First ridge regression failed with LinAlgError. Will re-run once more. " "This is due to a rare but documented issue with LAPACK. " "To attempt to circumvent this issue, we monkey-patch sklearn's _RidgeGCV to call scipy.linalg.svd with lapack_driver=\"gesvd\" instead of \"gesdd\". " "This seems to solve the problem but behavior is not guaranteed. " "For more details, see " "https://mathematica.stackexchange.com/questions/143894/sporadic-numerical-convergence-failure-of-singularvaluedecomposition-message-s")) logging.info("Re-running ridge regression with monkey-patched solver.") ### Import module and monkey patch import sklearn.linear_model._ridge as sklm sklm._RidgeGCV._svd_decompose_design_matrix = _svd_decompose_design_matrix_custom ### Re-initialize regressor reg = initialize_regressor(method, random_state) ### Re-fit reg.fit(X_train, Y_train) logging.info("Restoring original solver to _RidgeGCV class.") sklm._RidgeGCV._svd_decompose_design_matrix = _svd_decompose_design_matrix_original elif method == "linreg": logging.warning(("First linear regression failed with LinAlgError. Will re-run once more. " "This is due to a rare but documented issue with LAPACK. " "To attempt to circumvent this issue, we monkey-patch sklearn's LinearRegression class to call scipy.linalg.lstsq with lapack_driver=\"gelss\". " "This seems to solve the problem but behavior is not guaranteed. " "For more details, see " "https://mathematica.stackexchange.com/questions/143894/sporadic-numerical-convergence-failure-of-singularvaluedecomposition-message-s")) logging.info("Re-running linear regression with monkey-patched solver.") ### Import module and monkey patch import sklearn.linear_model._base as sklm sklm.LinearRegression.fit = _linear_regression_fit_custom ### Re-initialize regressor reg = initialize_regressor(method, random_state) ### Re-fit reg.fit(X_train, Y_train) logging.info("Restoring original solver to LinearRegression class.") sklm.LinearRegression.fit = _linear_regression_fit_original else: raise err if method == "ridge": coefs_df = pd.DataFrame([["METHOD", "RidgeCV"], ["SELECTED_CV_ALPHA", reg.alpha_], ["BEST_CV_SCORE", reg.best_score_]]) coefs_df = pd.concat([coefs_df, pd.DataFrame([cols, reg.coef_]).T]) coefs_df.columns = ["parameter", "beta"] coefs_df = coefs_df.set_index("parameter") elif method == "lasso": best_score = reg.mse_path_[np.where(reg.alphas_ == reg.alpha_)[0][0]].mean() coefs_df = pd.DataFrame([["METHOD", "LassoCV"], ["SELECTED_CV_ALPHA", reg.alpha_], ["BEST_CV_SCORE", best_score]]) coefs_df = pd.concat([coefs_df, pd.DataFrame([cols, reg.coef_]).T]) coefs_df.columns = ["parameter", "beta"] coefs_df = coefs_df.set_index("parameter") elif method == "linreg": coefs_df = pd.DataFrame([["METHOD", "LinearRegression"]]) coefs_df = pd.concat([coefs_df, pd.DataFrame([cols, reg.coef_]).T]) coefs_df.columns = ["parameter", "beta"] coefs_df = coefs_df.set_index("parameter") return coefs_df def pops_predict(mat, rows, cols, coefs_df): pred = mat.dot(coefs_df.loc[cols].beta.values) preds_df = pd.DataFrame([rows, pred]).T preds_df.columns = ["ENSGID", "PoPS_Score"] return preds_df ### --------------------------------- MAIN --------------------------------- ### def main(config_dict): ### --------------------------------- Basic settings --------------------------------- ### ### Set logging settings if config_dict["verbose"]: logging.basicConfig(format="%(levelname)s: %(message)s", level=logging.DEBUG) logging.info("Verbose output enabled.") else: logging.basicConfig(format="%(levelname)s: %(message)s") ### Set random seeds np.random.seed(config_dict["random_seed"]) random.seed(config_dict["random_seed"]) ### Display configs logging.info("Config dict = {}".format(str(config_dict))) ### --------------------------------- Reading/processing data --------------------------------- ### gene_annot_df = read_gene_annot_df(config_dict["gene_annot_path"]) ### If chromosome arguments are None, replace their values in config_dict with all chromosomes all_chromosomes = sorted(gene_annot_df.CHR.unique(), key=natural_key) if config_dict["project_out_covariates_chromosomes"] is None: config_dict["project_out_covariates_chromosomes"] = all_chromosomes logging.info("--project_out_covariates_chromosomes is None, defaulting to all chromosomes") if config_dict["feature_selection_chromosomes"] is None: config_dict["feature_selection_chromosomes"] = all_chromosomes logging.info("--feature_selection_chromosomes is None, defaulting to all chromosomes") if config_dict["training_chromosomes"] is None: config_dict["training_chromosomes"] = all_chromosomes logging.info("--training_chromosomes is None, defaulting to all chromosomes") ### Make sure all chromosome arguments are fully contained in gene_annot_df's chromosome list assert set(config_dict["project_out_covariates_chromosomes"]).issubset(all_chromosomes), "Invalid --project_out_covariates_chromosomes argument." assert set(config_dict["feature_selection_chromosomes"]).issubset(all_chromosomes), "Invalid --feature_selection_chromosomes argument." assert set(config_dict["training_chromosomes"]).issubset(all_chromosomes), "Invalid --training_chromosomes argument." ### Read in scores if config_dict["magma_prefix"] is not None: logging.info("MAGMA scores provided, loading MAGMA.") Y, covariates, error_cov, Y_ids = read_magma(config_dict["magma_prefix"], config_dict["use_magma_covariates"], config_dict["use_magma_error_cov"]) if config_dict["use_magma_covariates"] == True: logging.info("Using MAGMA covariates.") else: logging.info("Ignoring MAGMA covariates.") if config_dict["use_magma_error_cov"] == True: logging.info("Using MAGMA error covariance.") else: logging.info("Ignoring MAGMA error covariance.") ### Regularize MAGMA error covariance if using if error_cov is not None: logging.info("Regularizing MAGMA error covariance.") error_cov = regularize_error_cov(error_cov, Y, Y_ids, gene_annot_df) elif config_dict["y_path"] is not None: logging.info("Reading scores from {}.".format(config_dict["y_path"])) if config_dict["y_covariates_path"] is not None: logging.info("Reading covariates from {}.".format(config_dict["y_covariates_path"])) if config_dict["y_error_cov_path"] is not None: logging.info("Reading error covariance from {}.".format(config_dict["y_error_cov_path"])) ### Note that we do not regularize covariance matrix provided in y_error_cov_path. It will be used as is. Y, covariates, error_cov, Y_ids = read_from_y(config_dict["y_path"], config_dict["y_covariates_path"], config_dict["y_error_cov_path"]) else: raise ValueError("At least one of --magma_prefix or --y_path must be provided (--magma_prefix overrides --y_path).") ### Get projection, feature selection, and training genes project_out_covariates_Y_gene_inds = get_gene_indices_to_use(Y_ids, gene_annot_df, config_dict["project_out_covariates_chromosomes"], config_dict["project_out_covariates_remove_hla"]) feature_selection_Y_gene_inds = get_gene_indices_to_use(Y_ids, gene_annot_df, config_dict["feature_selection_chromosomes"], config_dict["feature_selection_remove_hla"]) training_Y_gene_inds = get_gene_indices_to_use(Y_ids, gene_annot_df, config_dict["training_chromosomes"], config_dict["training_remove_hla"]) ### Project out covariates if using if covariates is not None: logging.info("Projecting {} covariates out of target scores using genes on chromosome {}. HLA region {}." .format(covariates.shape[1], ", ".join(sorted(gene_annot_df.loc[Y_ids[project_out_covariates_Y_gene_inds]].CHR.unique(), key=natural_key)), "removed" if config_dict["project_out_covariates_remove_hla"] else "included")) Y_proj = project_out_covariates(Y, covariates, error_cov, Y_ids, gene_annot_df, project_out_covariates_Y_gene_inds) else: Y_proj = Y ### --------------------------------- Feature selection --------------------------------- ### ### Compute marginal association data frame logging.info("Computing marginal association table using genes on chromosome {}. HLA region {}." .format(", ".join(sorted(gene_annot_df.loc[Y_ids[feature_selection_Y_gene_inds]].CHR.unique(), key=natural_key)), "removed" if config_dict["feature_selection_remove_hla"] else "included")) marginal_assoc_df = compute_marginal_assoc(config_dict["feature_mat_prefix"], config_dict["num_feature_chunks"], Y_proj, Y_ids, None, error_cov, gene_annot_df, feature_selection_Y_gene_inds) ### Either do FSS or filter marginal_assoc_df if config_dict["feature_selection_fss_num_features"] is not None: logging.info("--feature_selection_fss_num_features set to {}, so performing forward stepwise selection (overriding all other feature selection settings).".format(config_dict["feature_selection_fss_num_features"])) selected_features = forward_stepwise_selection(config_dict["feature_mat_prefix"], config_dict["num_feature_chunks"], Y_proj, Y_ids, None, error_cov, gene_annot_df, feature_selection_Y_gene_inds, config_dict["feature_selection_fss_num_features"]) marginal_assoc_df["selected"] = marginal_assoc_df.index.isin(selected_features) ### Annotate with selection rank marginal_assoc_df["selection_rank"] = np.nan for i in range(len(selected_features)): marginal_assoc_df.loc[selected_features[i], "selection_rank"] = i + 1 logging.info("Forward stepwise selection complete, {} features in model.".format(len(selected_features))) else: ### Filter features based on settings selected_features = select_features_from_marginal_assoc_df(marginal_assoc_df, config_dict["subset_features_path"], config_dict["control_features_path"], config_dict["feature_selection_p_cutoff"], config_dict["feature_selection_max_num"]) ### Annotate marginal_assoc_df with selected True/False marginal_assoc_df["selected"] = marginal_assoc_df.index.isin(selected_features) ### Explicitly set features with nan p-values to not-selected marginal_assoc_df["selected"] = marginal_assoc_df["selected"] & ~pd.isnull(marginal_assoc_df.pval) ### Redefine selected_features selected_features = marginal_assoc_df[marginal_assoc_df.selected].index.values ### Complex logging statement select_feat_logtxt_pieces = [] if config_dict["subset_features_path"] is not None: select_feat_logtxt_pieces.append("subsetting to features at {}".format(config_dict["subset_features_path"])) if config_dict["feature_selection_p_cutoff"] is not None: if config_dict["feature_selection_max_num"] is not None: select_feat_logtxt_pieces.append("filtering to top {} features with p-value < {}" .format(config_dict["feature_selection_max_num"], config_dict["feature_selection_p_cutoff"])) else: select_feat_logtxt_pieces.append("filtering to features with p-value < {}" .format(config_dict["feature_selection_p_cutoff"])) elif config_dict["feature_selection_max_num"] is not None: select_feat_logtxt_pieces.append("filtering to top {} features by p-value" .format(config_dict["feature_selection_max_num"])) if config_dict["control_features_path"] is not None: select_feat_logtxt_pieces.append("unioning with non-constant control features") ### Combine complex logging statement if len(select_feat_logtxt_pieces) == 0: select_feat_logtxt = ("{} features reamin in model.".format(len(selected_features))) if len(select_feat_logtxt_pieces) == 1: select_feat_logtxt = ("After {}, {} features remain in model." .format(select_feat_logtxt_pieces[0], len(selected_features))) elif len(select_feat_logtxt_pieces) == 2: select_feat_logtxt = ("After {} and {}, {} features remain in model." .format(select_feat_logtxt_pieces[0], select_feat_logtxt_pieces[1], len(selected_features))) elif len(select_feat_logtxt_pieces) == 3: select_feat_logtxt = ("After {}, {}, and {}, {} features remain in model." .format(select_feat_logtxt_pieces[0], select_feat_logtxt_pieces[1], select_feat_logtxt_pieces[2], len(selected_features))) logging.info(select_feat_logtxt) ### --------------------------------- Training --------------------------------- ### ### Load data ### Won't necessarily load in order of selected_features. Loads in order of matrix columns. ### Note: doesn't raise error if trying to select feature that isn't in columns mat, cols, rows = load_feature_matrix(config_dict["feature_mat_prefix"], config_dict["num_feature_chunks"], selected_features) logging.info("Building training X and Y using genes on chromosome {}. HLA region {}." .format(", ".join(sorted(gene_annot_df.loc[Y_ids[training_Y_gene_inds]].CHR.unique(), key=natural_key)), "removed" if config_dict["training_remove_hla"] else "included")) ### Build training X and Y ### Should be properly subsetted and have error_cov applied. We also explicitly project out intercept X_train, Y_train = build_training(mat, cols, rows, Y_proj, Y_ids, error_cov, gene_annot_df, training_Y_gene_inds, project_out_intercept=True) logging.info("X dimensions = {}. Y dimensions = {}".format(X_train.shape, Y_train.shape)) ### Compute coefficients ### Output should contain at least one row for every column and additional rows for any metadata like method, regularization chosen by CV, etc. coefs_df = compute_coefficients(X_train, Y_train, cols, config_dict["method"], config_dict["random_seed"]) ### Prediction logging.info("Computing PoPS scores.") preds_df = pops_predict(mat, rows, cols, coefs_df) ### Annotate Y, Y_proj, and gene used in feature selection + gene used in training preds_df = preds_df.merge(pd.DataFrame(np.array([Y_ids, Y]).T, columns=["ENSGID", "Y"]), how="left", on="ENSGID") if covariates is not None: preds_df = preds_df.merge(pd.DataFrame(np.array([Y_ids, Y_proj]).T, columns=["ENSGID", "Y_proj"]), how="left", on="ENSGID") preds_df["project_out_covariates_gene"] = preds_df.ENSGID.isin(Y_ids[project_out_covariates_Y_gene_inds]) preds_df["feature_selection_gene"] = preds_df.ENSGID.isin(Y_ids[feature_selection_Y_gene_inds]) preds_df["training_gene"] = preds_df.ENSGID.isin(Y_ids[training_Y_gene_inds]) ### --------------------------------- Save --------------------------------- ### logging.info("Writing output files.") preds_df.to_csv(config_dict["out_prefix"] + ".preds", sep="\t", index=False) coefs_df.to_csv(config_dict["out_prefix"] + ".coefs", sep="\t") marginal_assoc_df.to_csv(config_dict["out_prefix"] + ".marginals", sep="\t") if config_dict["save_matrix_files"] == True: logging.info("Saving matrix files as well.") pd.DataFrame(np.hstack((Y_train.reshape(-1,1), X_train)), index=Y_ids[training_Y_gene_inds], columns=["Y_train"] + list(cols)).to_csv(config_dict["out_prefix"] + ".traindata", sep="\t") pd.DataFrame(mat, index=rows, columns=cols).to_csv(config_dict["out_prefix"] + ".matdata", sep="\t") ### Main if __name__ == '__main__': args = get_pops_args() config_dict = vars(args) main(config_dict) |
14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 | library(conflicted) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") packages <- c("optparse","dplyr","data.table","reshape2","ggplot2","ggpubr","conflicted", "cluster","textshape","readxl","writexl","tidyr","org.Hs.eg.db","stats", "gplots", "stringi" # heatmap.2 ) xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("replace_na", "dplyr") conflict_prefer("filter", "dplyr") ## optparse option.list <- list( make_option("--input.GWAS.table", type="character", default="/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/overlap/MAP/MAP_GWAS_gene_incl_ubq_genes.txt", help="Input CAD GWAS Table"), make_option("--cNMF.table", type="character", default="", help="Table with cNMF program result"), make_option("--outdir", type="character", default="./outputs/MAP/", help="Output directory"), make_option("--figdir", type="character", default="./figures/", help="Output directory"), make_option("--sampleName", type="character", default="2kG.library", help="Name of the sample"), make_option("--celltype", type="character", default="EC", help="Cell type in GWAS table"), make_option("--K.val", type="numeric", default=60, help="The value of K"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--outdirsample", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/", help="path to cNMF analysis results"), ## or for 2n1.99x: "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211116_snakemake_dup4_cells/analysis/all_genes/Perturb_2kG_dup4/K60/threshold_0_2/" make_option("--num.tests", type="numeric", default=2, help="number of statistical test to do from the top of statistical.test.df.txt"), make_option("--trait.name", type="character", default="MAP", help="name of the trait"), make_option("--coding.variant.df", type="character", default="/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/SBP/SBPvariant.list.1.coordinate.txt", help="Data frame with coding variant information for GWAS trait"), make_option("--regulator.analysis.type", type="character", default="GWASWide", help="path to statistical test recipe"), make_option("--perturbSeq", type="logical", default=FALSE, help="Whether this is a Perturb-seq experiment"), make_option("--TPM.table", type="character", default="", help="Path to TPM table for the correct cell type") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## sdev for K562 gwps 2k overdispersed genes K=80 ## opt$input.GWAS.table <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/overlap/RBC/RBC_GWAS_gene_incl_ubq_genes.txt" ## opt$cNMF.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K80/threshold_0_2/prepare_compute_enrichment.txt" ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K80/program_prioritization_GenomeWide/RBC/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K80/threshold_0_2/program_prioritization_GenomeWide/RBC/" ## opt$trait.name <- "RBC" ## opt$celltype <- "K562" ## opt$K.val <- 80 ## opt$coding.variant.df <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/RBC/RBCvariant.list.1.coordinate.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K80/threshold_0_2/" ## opt$perturbSeq <- TRUE ## opt$regulator.analysis.type <- "GenomeWide" ## ## sdev for K562 gwps 2k overdispersed genes K=90 RBC ## opt$input.GWAS.table <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/overlap/RBC/RBC_GWAS_gene_incl_ubq_genes.txt" ## opt$cNMF.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/prepare_compute_enrichment.txt" ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K90/program_prioritization_GenomeWide/RBC/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/program_prioritization_GenomeWide/RBC/" ## opt$trait.name <- "RBC" ## opt$celltype <- "K562" ## opt$K.val <- 90 ## opt$coding.variant.df <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/RBC/RBCvariant.list.1.coordinate.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/" ## opt$perturbSeq <- TRUE ## opt$regulator.analysis.type <- "GenomeWide" ## opt$TPM.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/230512_TPM/outputs/K562.TPM.df.txt" ## ## sdev for K562 gwps 2k overdispersed genes K=90 Plt ## opt$input.GWAS.table <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/overlap/Plt/Plt_GWAS_gene_incl_ubq_genes.txt" ## opt$cNMF.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/prepare_compute_enrichment.txt" ## opt$sampleName <- "WeissmanK562gwps" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K90/program_prioritization_GenomeWide/Plt/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/program_prioritization_GenomeWide/Plt/" ## opt$trait.name <- "Plt" ## opt$celltype <- "K562" ## opt$K.val <- 90 ## opt$coding.variant.df <- "/oak/stanford/groups/engreitz/Users/rosaxma/2111_pipeline_output/UKB/Plt/Pltvariant.list.1.coordinate.txt" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/" ## opt$perturbSeq <- TRUE ## opt$regulator.analysis.type <- "GenomeWide" ## opt$TPM.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/230512_TPM/outputs/K562.TPM.df.txt" ## ## sdev for EC Perturb-seq K=60 ## opt$input.GWAS.table <- "/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/GWAS_tables/CAD_GWAS_gene_incl_ubq_genes_aragam_all_harst.txt" ## opt$cNMF.table <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/prepare_compute_enrichment.txt" ## opt$sampleName <- "2kG.library" ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/2kG.library/K60/program_prioritization_GWASWide/CAD/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/program_prioritization_GWASWide/CAD/" ## opt$trait.name <- "CAD" ## opt$celltype <- "EC" ## opt$K.val <- 60 ## opt$coding.variant.df <- "" ## opt$outdirsample <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/" ## opt$perturbSeq <- TRUE ## opt$regulator.analysis.type <- "GWASWide" ## Directories FIGDIR <- opt$figdir OUTDIR <- opt$outdir DATADIR <- "/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/" ## update this path check.dir <- c(FIGDIR, OUTDIR) invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x, recursive=T) })) ########################################################################################## ## Load Data ## load cNMF results to get the list of input genes to cNMF (need theta) k <- opt$K.val SAMPLE <- opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) SUBSCRIPT.SHORT=paste0("k_", k, ".dt_", DENSITY.THRESHOLD) OUTDIRSAMPLE <- opt$outdirsample ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/" celltype <- opt$celltype cNMF.result.file <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT, ".RData") print(cNMF.result.file) if(file.exists(cNMF.result.file)) { print("loading cNMF result file") load(cNMF.result.file) } ## load 10X reference gtf.10X.df <- readRDS(paste0("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/refdata-cellranger-arc-GRCh38-2020-A_genes.gtf_df.RDS")) ## load 10X gtf file Gene.ENSEMBL.10X.df <- gtf.10X.df %>% mutate(ENSGID = gene_id, Gene10X = gene_name) %>% select(Gene10X, ENSGID) %>% unique ## gtf <- importGTF("/home/groups/engreitz/Software/cellranger-arc-1.0.1/refdata-cellranger-arc-GRCh38-2020-A/genes/genes.gtf") db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) ## load the appropriate database ## helper function to map between ENSGID and SYMBOL map.ENSGID.SYMBOL <- function(df) { ## need column `Gene` to be present in df ## detect gene data type (e.g. ENSGID, Entrez Symbol) gene.type <- ifelse(nrow(df) == sum(as.numeric(grepl("^ENS", df$Gene))), "ENSGID", "Gene") if(gene.type == "ENSGID") { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "ENSEMBL", column = "SYMBOL") df <- df %>% mutate(ENSGID = Gene, Gene = mapped.genes) } else { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "SYMBOL", column = "ENSEMBL") df <- df %>% mutate(ENSGID = mapped.genes) } df <- df %>% mutate(Gene = ifelse(is.na(Gene), "NA", Gene), ENSGID = ifelse(is.na(ENSGID), "NA", ENSGID)) df <- df %>% merge(Gene.ENSEMBL.10X.df, by="ENSGID", all.x=T) df <- df %>% mutate(Gene = ifelse(Gene == "NA", Gene10X, Gene)) notMatched.df <- df %>% subset(is.na(Gene10X)) if(nrow(notMatched.df) > 0) { notMatched.index <- notMatched.df %>% rownames match.df <- merge(notMatched.df %>% select(-ENSGID, -Gene10X), Gene.ENSEMBL.10X.df, by.x="Gene", by.y="Gene10X") %>% mutate(Gene10X = Gene) %>% select(all_of(df %>% colnames)) matched.index <- notMatched.df %>% subset(Gene %in% c(match.df$Gene %>% unique)) %>% rownames %>% as.numeric df <- rbind(df[-matched.index,], match.df) } df <- df %>% mutate(OriginalGene = Gene, Gene = Gene10X) ## toconvert <- df %>% subset(Gene != Gene10X) %>% select(ENSGID, Gene, Gene10X) ## toconvert.index <- toconvert %>% ## df %>% subset(Gene %in% Gene.ENSEMBL.10X.df$Gene10X & is.na(ENSGID)) return(df) } if(grepl("2kG.library", SAMPLE)) { ## Perturb-seq vs 10X names (from Gavin) ## ptb10xNames.df <- read_xlsx("../data/Perturbation 10X names.xlsx") %>% as.data.frame ptb10xNames.df <- read.delim(paste0(DATADIR, "/data/220627_add_Perturbation 10X names.txt"), stringsAsFactors=F, check.names=F) ## write.table(ptb10xNames.df,"../data/220627_add_Perturbation 10X names.220628.txt", quote=F, row.names=F, sep="\t") perturbseq.gene.names.to10X <- function(Gene) stri_replace_all_regex(Gene, pattern = ptb10xNames.df$Symbol, replace = ptb10xNames.df$`Name used by CellRanger`, vectorize=F) tenX.gene.names.toperturbseq <- function(Gene) stri_replace_all_regex(Gene, pattern = ptb10xNames.df$`Name used by CellRanger`, replace = ptb10xNames.df$Symbol, vectorize=F) } ## gtf.10X.df <- readRDS(paste0(DATADIR, "/data/refdata-cellranger-arc-GRCh38-2020-A_genes.gtf_df.RDS")) ## load 10X gtf file message(paste0("Loading input GWAS table file from ", opt$input.GWAS.table)) GWAS.df <- GWAS.df.original <- read.delim(opt$input.GWAS.table, stringsAsFactors=F) %>% mutate(original.gene = gene) %>% select(-ProgramsInWhichGeneIsInTop100ZScoreSpecificGenes, -ProgramsInWhichGeneIsInTop300ZScoreSpecificGenes, -ProgramsInWhichGeneIsInTop500ZScoreSpecificGenes, -PoPS_Score, -PoPS.Rank, -Top5ProgramsThatContributeToPoPSScore) %>% mutate(GeneInGWASTable = TRUE, Gene = gene) %>% map.ENSGID.SYMBOL if(grepl("2kG.library", SAMPLE)){ GWAS.df <- GWAS.df %>% ## mutate(original.gene = gene) %>% mutate(gene = gene %>% perturbseq.gene.names.to10X) ## add columns on_2kG_lib and TeloHAEC_ctrl_TPM genes.on.2kG.lib <- read_xlsx(paste0(DATADIR, "/data/Table.S1.1 revised 220722.xlsx")) %>% mutate(original.gene = `Symbol (for library design)`, gene = `Symbol (in CellRanger)`) RNAseq_CITV <- read.delim(paste0(DATADIR, "/CAD_SNP_INFO/RNASeq_Telo_Eahy_pm_IL1b_TNF_VEGF.txt"), stringsAsFactors=F) TPM.df <- RNAseq_CITV %>% select(Gene_symbol, TeloHAEC.Ctrl_avg) %>% `colnames<-`(c("gene", "TeloHAEC_ctrl_TPM")) ## GWAS.df <- GWAS.df %>% ## mutate(on_2kG_lib = gene %in% genes.on.2kG.lib$gene) %>% ## merge(TPM.df, by="gene", all.x=T) } perturbed_gene_ary <- barcode.names$Gene %>% unique ## Load data for all genes relation to the topics (not limited to CAD GWAS genes) message(paste0("Loading input G2P table file from ", opt$cNMF.table)) narrow.df <- read.delim(opt$cNMF.table, header=T, stringsAsFactors=F)## %>% select(-Top5FeaturesThatContributeToPoPSScore) ## narrow.df <- narrow.df[narrow.df$Gene!="NULL", ] if(grepl("2kG.library", SAMPLE)) { narrow.df.TPM <- merge(TPM.df, narrow.df %>% select(-ENSGID) %>% unique, by.x=c("gene"), by.y=c("Gene"), all.y=T) narrow.df.TPM$gene[narrow.df.TPM$gene == "MESDC1"] <- "TLNRD1" ## one time fix MESDC1 -> TLNRD1 } else { narrow.df.TPM <- narrow.df %>% mutate(gene = Gene) } if(!grepl("2kG.library", SAMPLE)) { coding_variant.df <- read.delim(opt$coding.variant.df, stringsAsFactors=F) GeneWithCodingVariant <- coding_variant.df$CodingVariantGene %>% unique GWAS.df <- GWAS.df %>% mutate(GeneContainsCodingVariant = (gene %in% GeneWithCodingVariant) | (original.gene %in% GeneWithCodingVariant), CodingVariantGene = ifelse(GeneContainsCodingVariant, gene, NA)) } else { all.cs.aragam <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/CAD_SNP_INFO/Aragam2021/all.cs.txt", stringsAsFactors=F) all.cs.harst <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/CAD_SNP_INFO/Harst2017/all.cs.txt", stringsAsFactors=F) addCodingGenes <- function(df, all.cs) { codingGenes <- all.cs %>% filter(AnyCoding) %>% select(CredibleSet, CodingVariantGene) df <- df %>% merge(codingGenes, all.x=TRUE) %>% rowwise() %>% mutate(GeneContainsCodingVariant=ifelse(is.na(CodingVariantGene), FALSE, (gene %in% strsplit(CodingVariantGene,";")[[1]]))) %>% ungroup() %>% as.data.frame() return(df) } GWAS.df <- GWAS.df %>% ## mutate(Resource=ifelse(Resource=="Aragam_2021","Aragam2021","Harst2017"), ## CredibleSet=paste0(Resource,"-",Lead_SNP_rsID)) %>% addCodingGenes(rbind(all.cs.aragam,all.cs.harst)) %>% mutate(CredibleSet = ifelse(CredibleSet %in% c("Aragam2021-rs768453105", "Harst2017-rs138120077"), "Aragam2021-rs768453105_Harst2017-rs138120077", CredibleSet)) } ## include TPM table if(!("TPM" %in% colnames(GWAS.df))) { TPM.df <- read.delim(opt$TPM.table, stringsAsFactors=F) GWAS.tmp.df <- merge(GWAS.df, TPM.df, by.x="gene", by.y="Gene", all.x=T) GWAS.df <- GWAS.tmp.df } ######################################################################################## ## Define key variables for analysis GWAS.df <- GWAS.df %>% mutate( Expressed = (TPM >= 1), InPlausibleCellTypeSpecificLocus = ( !LipidLevelsAssociated & (!is.na(get(paste0("RankOfDistanceTo", celltype, "PeakWithVariant"))) | !is.na(get(paste0("MaxABC.Rank.", celltype, "Only"))) | !is.na(CodingVariantGene)) ), ## Also require that locus is not associated with lipid levels InPlausibleCellTypeSpecificLocus_includeLipid = (!is.na(get(paste0("RankOfDistanceTo", celltype, "PeakWithVariant"))) | !is.na(get(paste0("MaxABC.Rank.", celltype, "Only"))) | !is.na(CodingVariantGene)), InPlausibleLocus = ( !LipidLevelsAssociated & (!is.na(rank_SNP_to_TSS) | !is.na(MaxABC.Rank) | !is.na(CodingVariantGene)) ), InPlausibleLocus_includeLipid = !is.na(rank_SNP_to_TSS) | !is.na(MaxABC.Rank) | !is.na(CodingVariantGene), TopCandidate = (get(paste0("RankOfDistanceTo", celltype, "PeakWithVariant")) <= 2 | MaxABC.Rank.ECOnly <= 2 | GeneContainsCodingVariant) %>% replace_na(FALSE), TopCandidateInCellTypeSpecificLocus = TopCandidate & InPlausibleCellTypeSpecificLocus ## !!(paste0("ExpressedTopCandidateIn", celltype, "Locus")) := ((get(paste0("RankOfDistanceTo", celltype, "PeakWithVariant")) <= 2 | get(paste0("MaxABC.Rank.", celltype, "Only")) <= 2 | GeneContainsCodingVariant) %>% replace_na(FALSE)) & get(paste0("InPlausible", celltype, "Locus")) & Expressed, ## TopCandidateInCellTypeSpecificLocus_includeLipid = ((get(paste0("RankOfDistanceTo", celltype, "PeakWithVariant")) <= 2 | get(paste0("MaxABC.Rank.", celltype, "Only")) <= 2 | GeneContainsCodingVariant) %>% replace_na(FALSE)) & get(paste0("InPlausible", celltype, "Locus_includeLipid")) ## TopCandidate = ((rank_SNP_to_TSS <= 2 | MaxABC.Rank <= 2 | GeneContainsCodingVariant) %>% replace_na(FALSE)) & InPlausibleLocus, ## TopCandidate_includeLipid = ((rank_SNP_to_TSS <= 2 | MaxABC.Rank <= 2 | GeneContainsCodingVariant) %>% replace_na(FALSE)) & InPlausibleLocus_includeLipid ) %>% mutate(perturbed_gene = (gene %in% perturbed_gene_ary)) %>% as.data.frame() ########################################################################################## ## quick statistics on GWAS.df ## Count of V2G linked genes cat("Number of genes with V2G links: ", GWAS.df %>% filter(TopCandidateInCellTypeSpecificLocus) %>% pull(gene) %>% unique %>% length, "\n") ## Count of credible sets that count as "plausibleCellTypeSpecificLocus": cat("Number of credible sets that count as 'in a plausible CellTypeSpecific locus': ", GWAS.df %>% filter(InPlausibleCellTypeSpecificLocus) %>% pull(CredibleSet) %>% unique() %>% length(), " out of ", GWAS.df %>% pull(CredibleSet) %>% unique() %>% length(), "\n") ## Count of credible sets that count as "plausibleLocus": cat("Number of credible sets that count as 'in a plausible locus': ", GWAS.df %>% filter(InPlausibleLocus) %>% pull(CredibleSet) %>% unique() %>% length(), " out of ", GWAS.df %>% pull(CredibleSet) %>% unique() %>% length(), "\n") ## merge CAD.GWAS.df into narrow.df.TPM to get all expressed gene's information remove_na <- function(x) x[is.na(x)] <- NULL overlapping.colnames <- intersect(narrow.df.TPM %>% colnames, GWAS.df %>% colnames) narrow.df.TPM <- merge(narrow.df.TPM, # %>% GWAS.df %>% select(-one_of((overlapping.colnames[!(overlapping.colnames %in% c("gene"))]))), by=c("gene"), all=T ) %>% rowwise() %>% mutate(ProgramsLinkedToGene= c(strsplit(ProgramsRegulatedByThisGene,",")[[1]], strsplit(ProgramsInWhichGeneIsInTop300ZScoreSpecificGenes, ",")[[1]]) %>% unique() %>% paste(collapse=',')) %>% ## mutate(perturbed_gene = (gene %in% perturbed_gene_ary) | (OriginalGene %in% perturbed_gene_ary)) %>% ## ## mutate(ProgramsLinkedToGene= c(strsplit(ProgramsRegulatedByThisGene,",")[[1]], strsplit(ProgramsInWhichGeneIsInTop300ZScoreSpecificGenes, ",")[[1]]) %>% unique() %>% paste(collapse=',')) %>% as.data.frame ## because CAD.GWAS.df has dupicated genes, TopicsInWhichGeneIsInTop100ZScoreSpecificGenes sum up to > 100 narrow.df.TPM[is.na(narrow.df.TPM)] <- FALSE ## head(narrow.df.TPM) ## print(colnames(narrow.df.TPM)) ## GenesIncNMF <- c(theta.zscore.rank.df$Gene %>% unique) narrow.df.TPM <- narrow.df.TPM %>% ## rowwise %>% mutate(AllGenesIncNMFInput = IncNMFAnalysis) %>% as.data.frame ## write GWAS table to file message("write GWAS.df to file") write.table(GWAS.df, paste0(OUTDIR, opt$trait.name, ".GWAS.df.txt"), sep="\t", quote=F, row.names=F) write.table(narrow.df.TPM, paste0(OUTDIR, opt$trait.name, ".narrow.df.TPM.txt"), sep="\t", quote=F, row.names=F) # narrow.df.TPM <- read.delim(paste0(OUTDIR, opt$trait.name, ".narrow.df.TPM.txt"), stringsAsFactors=F) ######################################################################################## ## Run statistical Tests ## Create subset dfs based on background conditions ## read table that specifies which gene set is for alternative hypothesis and which set of topics to use message("reading statistical test list") statistical.test.list.df <- read.delim(paste0("workflow/scripts/program_prioritization/All_GWAS_traits.statistical.test.list_", opt$regulator.analysis.type, ".txt"), stringsAsFactors=F) ## statistical.test.list.df <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230119_apply_V2G2P_to_more_GWAS_traits/All_GWAS_traits.statistical.test.list.txt"), stringsAsFactors=F) message("double check statistical test criteria") statistical.test.list.df[1:2,] ## double check test criteria num.tests <- opt$num.tests ## actual number of tests requested by the user statistical.test.list.df <- statistical.test.list.df[1:num.tests,] ## num.tests <- nrow(statistical.test.list.df) ## specify number of statistical tests to conduct df.list.to.test <- vector("list", num.tests) ## initiate background data storage list names(df.list.to.test) <- statistical.test.list.df$test.name ## remove batch topics from test if(grepl("2kG.library", SAMPLE)) { ## load topic summary TopicSummary.df <- read.delim(paste0(DATADIR, "/data/210730_cNMF_topic_model_analysis.xlsx - TopicCatalogNEW.tsv"), stringsAsFactors=F) ## batch effect topic batch.topics <- TopicSummary.df %>% filter(ProgramCategoryLabel == "Batch") %>% pull(ProgramID) %>% unique } else { message("reading batch topics") filename = paste0(opt$outdirsample, "/batch.topics.txt") empty = file.size(filename) == 0L if(empty) { message(paste0("There is no batch topics in K = ", k)) batch.topics = character(0) } else { batch.topics <- read.delim(filename, stringsAsFactors=F, header=F) %>% mutate(ProgramID = paste0("K", k, "_", gsub("topic_", "", V1))) %>% pull(ProgramID) %>% unique } } ## topics_to_test <- setdiff(c(1:k), batch.topics %>% gsub(paste0("K", k, "_"), "", .)) ## do not remove batch topic from the test topics_to_test = c(1:k) all.test.results.background.list <- vector("list", num.tests) for(i in 1:num.tests) { background.name <- names(df.list.to.test)[i] ## get the name of the background filter subset.df <- eval(parse(text = paste0("subset.df <- ", statistical.test.list.df$background.subset.command[i]))) all.test.results.list <- vector("list", k) ## initialize variable ## loop over every topic t for(t in topics_to_test) { topic <- paste0("K", k, "_", t) ## new scratch for test that only looks at 'expressed genes in a topic' genes.to.test <- statistical.test.list.df$genes.to.test[i] topics.to.test <- statistical.test.list.df$topics.to.test[i] list.to.test <- subset.df %>% rowwise %>% mutate(hypothesis.gene = ifelse(eval(parse(text = genes.to.test)), 1, 0), LinkedToTopic.gene = ifelse(topic %in% (get(topics.to.test) %>% strsplit("\\|") %>% unlist %>% as.character), 1, 0)) %>% ## debug ## mutate(gene = paste0(gene, "_", Gene, "_", OriginalGene)) %>% select(gene, hypothesis.gene, LinkedToTopic.gene) %>% unique %>% group_by(gene) %>% summarize(hypothesis.gene = ifelse(hypothesis.gene %>% sum > 0, "CandidateGene", "NotCandidateGene"), LinkedToTopic.gene = ifelse(LinkedToTopic.gene %>% sum > 0, "LinkedToTopic", "NotLinkedToTopic")) %>% as.data.frame table.to.test <- table(list.to.test$hypothesis.gene, list.to.test$LinkedToTopic.gene) ## count frequencies if(ncol(table.to.test) == 1) { columns.needed <- c("LinkedToTopic", "NotLinkedToTopic") column.to.add <- columns.needed[!(columns.needed %in% colnames(table.to.test))] table.to.add <- matrix(c(0,0), nrow=2, dimnames=list(rownames(table.to.test),c(column.to.add))) if(column.to.add == "NotLinkedToTopic") table.to.test <- cbind(table.to.test, table.to.add) else table.to.test <- cbind(table.to.add, table.to.test) } else if (nrow(table.to.test) == 1) { rows.needed <- c("CandidateGene", "NotCandidateGene") row.to.add <- row.needed[!(row.needed %in% rownames(table.to.test))] table.to.add <- matrix(c(0,0), ncol=2, dimnames=list(c(row.to.add), colnames(table.to.test))) if(row.to.add == "NotCandidateGene") table.to.test <- rbind(table.to.test, table.to.add) else table.to.test <- rbind(table.to.add, table.to.test) } flattened.table.to.store <- table.to.test %>% matrix(nrow=1, ncol=4) %>% ## flatten the table as.data.frame %>% `colnames<-`(c("LinkedToTopic_CandidateGene", "LinkedToTopic_NotCandidateGene", "NotLinkedToTopic_CandidateGene", "NotLinkedToTopic_NotCandidateGene")) %>% ## assign the table values to the corresponding column names mutate(enrichment = ( LinkedToTopic_CandidateGene / (LinkedToTopic_CandidateGene + NotLinkedToTopic_CandidateGene) ) / ( LinkedToTopic_NotCandidateGene / ( LinkedToTopic_NotCandidateGene + NotLinkedToTopic_NotCandidateGene) ), ## calculate enrichment enrichment.log2fc = log2(enrichment), ## log2fc of the enrichment Gene_LinkedToTopic_CandidateGene = list.to.test %>% ## list the genes that fall into each of the four categories in the contingency table subset(hypothesis.gene == "CandidateGene" & LinkedToTopic.gene == "LinkedToTopic") %>% pull(gene) %>% paste0(collapse=","), Gene_LinkedToTopic_NotCandidateGene = list.to.test %>% subset(hypothesis.gene == "NotCandidateGene" & LinkedToTopic.gene == "LinkedToTopic") %>% pull(gene) %>% paste0(collapse=","), Gene_NotLinkedToTopic_CandidateGene = list.to.test %>% subset(hypothesis.gene == "CandidateGene" & LinkedToTopic.gene == "NotLinkedToTopic") %>% pull(gene) %>% paste0(collapse=",")#, ## Gene_NotLinkedToTopic_NotCandidateGene = list.to.test %>% ## subset(hypothesis.gene == "NotCandidateGene" & LinkedToTopic.gene == "NotLinkedToTopic") %>% ## pull(gene) %>% ## paste0(collapse=",") ) ## null distribution probability p.binomial.null <- (list.to.test %>% subset(LinkedToTopic.gene == "LinkedToTopic") %>% nrow) / (list.to.test %>% nrow) p.binomial.test <- binom.test(list.to.test %>% subset(hypothesis.gene == "CandidateGene" & LinkedToTopic.gene == "LinkedToTopic") %>% nrow, ## number of candidate genes that link to the topic list.to.test %>% subset(hypothesis.gene == "CandidateGene") %>% nrow, ## number of candidate genes p = p.binomial.null)$p.value ## binomial test if (dim(table.to.test)[1] == 2 & dim(table.to.test) [2] == 2) { fisher.p.val <- fisher.test(table.to.test, alternative="greater")$p.value } else { fisher.p.val = NA ## add a row of 0? } all.test.results.list[[t]] <- data.frame(Topic = topic, fisher.p.value = fisher.p.val, binomial.p.value = p.binomial.test) %>% cbind(flattened.table.to.store) ## put all results for one topic in one line to store } all.test.results.background.list[[i]] <- do.call(rbind, all.test.results.list) %>% mutate(fisher.p.adjust = p.adjust(fisher.p.value), binomial.p.adjust = p.adjust(binomial.p.value), background = background.name, .after="binomial.p.value") } all.test.results <- do.call(rbind, all.test.results.background.list) %>% as.data.frame %>% mutate(Gene_NotLinkedToTopic_CandidateGene = ifelse(NotLinkedToTopic_CandidateGene > 300, "", Gene_NotLinkedToTopic_CandidateGene), Gene_LinkedToTopic_CandidateGene = ifelse(LinkedToTopic_CandidateGene > 150, "", Gene_LinkedToTopic_CandidateGene)) #Gene_LinkedToTopic_NotCandidateGene)) all.test.results.nonbatch <- all.test.results %>% subset(Topic %in% paste0("K", k, "_", setdiff(c(1:k), batch.topics %>% gsub(paste0("K", k, "_"), "", .)))) %>% group_by(background) %>% mutate(fisher.p.adjust = p.adjust(fisher.p.value, method="fdr"), binomial.p.adjust = p.adjust(binomial.p.value, method="fdr"), .after="binomial.p.value") %>% as.data.frame ## output statistical test results write.table(all.test.results, file=paste0(OUTDIR, opt$trait.name, ".fisher.exact.test.to.prioritize.topics.txt"), sep="\t", quote=F, row.names=F) write.table(all.test.results.nonbatch, file=paste0(OUTDIR, opt$trait.name, ".fisher.exact.test.to.prioritize.topics.nonbatch.txt"), sep="\t", quote=F, row.names=F) head(all.test.results) ## ## reload data ## all.test.results.nonbatch <- read.delim(paste0(OUTDIR, opt$trait.name, ".fisher.exact.test.to.prioritize.topics.nonbatch.txt"), stringsAsFactors=F) regulator.programGene.test.pairs <- data.frame(regulator = statistical.test.list.df$test.name[seq(2, num.tests, by=2)], programGene = statistical.test.list.df$test.name[seq(1, num.tests-1, by=2)], name = statistical.test.list.df[1:num.tests,] %>% separate(col="test.name", sep="_geneSet.", into = c("regulator.or.program.gene", "toProcess")) %>% separate(col="toProcess", sep="_", into=c("test.handle", "background")) %>% pull(test.handle) %>% unique, command = statistical.test.list.df$genes.to.test[seq(1, num.tests-1, by=2)], regulator.selection.command = statistical.test.list.df$background.subset.command[seq(2, num.tests, by=2)]) regulator.programGene.combined.pval.nonbatch.list <- vector("list", nrow(regulator.programGene.test.pairs)) for(i in 1:nrow(regulator.programGene.test.pairs)) { regulator.results.nonbatch <- all.test.results.nonbatch %>% subset(background == regulator.programGene.test.pairs$regulator[i]) %>% `colnames<-`(paste0("Regulator_", colnames(.))) %>% dplyr::rename("ProgramID" = "Regulator_Topic") programGene.results.nonbatch <- all.test.results.nonbatch %>% subset(background == regulator.programGene.test.pairs$programGene[i]) %>% `colnames<-`(paste0("ProgramGene_", colnames(.))) %>% dplyr::rename("ProgramID" = "ProgramGene_Topic") ## calculate expected number of genes ## adapted from 221115_compute_enrichment.R mean.Regulators = regulator.results.nonbatch %>% mutate(nRegulators=Regulator_LinkedToTopic_CandidateGene + Regulator_LinkedToTopic_NotCandidateGene) %>% pull(nRegulators) %>% mean ## nPerturbedGeneInCADGWASLoci = GWAS.df %>% filter(perturbed_gene == 1) %>% pull(gene) %>% unique %>% length ## nPerturbedGeneCandidate = GWAS.df %>% filter(perturbed_gene == 1 & eval(parse(text = regulator.programGene.test.pairs$command[i] %>% as.character))) %>% pull(gene) %>% unique %>% length nPerturbedGene = eval(parse(text = regulator.programGene.test.pairs$regulator.selection.command[i] %>% as.character)) %>% pull(gene) %>% unique %>% length ## old: narrow.df.TPM %>% filter(perturbed_gene == 1) %>% pull(gene) %>% unique %>% length nPerturbedGeneCandidate = eval(parse(text = regulator.programGene.test.pairs$regulator.selection.command[i] %>% as.character)) %>% filter(eval(parse(text = regulator.programGene.test.pairs$command[i] %>% as.character))) %>% pull(gene) %>% unique %>% length ## old: narrow.df.TPM %>% filter(perturbed_gene == 1 & eval(parse(text = regulator.programGene.test.pairs$command[i] %>% as.character))) %>% pull(gene) %>% unique %>% length expected.Regulators = mean.Regulators * nPerturbedGeneCandidate / nPerturbedGene nProgramGeneBackground = narrow.df.TPM %>% filter(IncNMFAnalysis == 1) %>% pull(gene) %>% unique %>% length nProgramGeneCandidate = narrow.df.TPM %>% subset(IncNMFAnalysis == 1 & eval(parse(text = regulator.programGene.test.pairs$command[i] %>% as.character))) %>% pull(gene) %>% unique %>% length expected.ProgramGenes = 300 * nProgramGeneCandidate / nProgramGeneBackground expected.nLinkedGenes = expected.Regulators + expected.ProgramGenes regulator.programGene.combined.pval.nonbatch.list[[i]] <- merge(regulator.results.nonbatch, programGene.results.nonbatch, by="ProgramID") %>% mutate(LinkedGenes_ChiSquareTestStatistic = -2 * (log(Regulator_fisher.p.value) + log(ProgramGene_fisher.p.value)), LinkedGenes_ChiSquare.p.value = pchisq(LinkedGenes_ChiSquareTestStatistic, df = 4, lower.tail=F), LinkedGenes_fisher.p.value = Regulator_fisher.p.value * ProgramGene_fisher.p.value, LinkedGenes_binomial.p.value = Regulator_binomial.p.value * ProgramGene_binomial.p.value, LinkedGenes_fisher.p.adjust = p.adjust(LinkedGenes_fisher.p.value, method="fdr"), LinkedGenes_binomial.p.adjust = p.adjust(LinkedGenes_binomial.p.value, method="fdr"), LinkedGenes_ChiSquare.p.adjust = p.adjust(LinkedGenes_ChiSquare.p.value, method="fdr"), LinkedGenes_fisherNegLog10FDR = -log10(LinkedGenes_fisher.p.adjust), LinkedGenes_binomialNegLog10FDR = -log10(LinkedGenes_binomial.p.adjust), LinkedGenes_ChiSquareNegLog10FDR = -log10(LinkedGenes_ChiSquare.p.adjust), ## LinkedGenes_LinkedToTopic_CandidateGene = Regulator_LinkedToTopic_CandidateGene + ProgramGene_LinkedToTopic_CandidateGene, .after = "ProgramID") %>% ## adapted from 211115_compute_enrichment.R rowwise %>% mutate(LinkedGenes_LinkedToTopic_CandidateGene = c(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% length, LinkedGenes_Gene_LinkedToTopic_CandidateGene = c(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% paste0(collapse=","), InGeneSet_Regulator_and_ProgramGene = intersect(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% length, Gene_InGeneSet_Regulator_and_ProgramGene = intersect(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% paste0(collapse=","), Unique_Regulator_InGeneSet_CandidateGene = setdiff(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% length, Unique_ProgramGene_InGeneSet_CandidateGene = setdiff(ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% length, Unique_Regulator_Gene_InGeneSet_CandidateGene = setdiff(Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% paste0(collapse=","), Unique_ProgramGene_Gene_InGeneSet_CandidateGene = setdiff(ProgramGene_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character, Regulator_Gene_LinkedToTopic_CandidateGene %>% strsplit(split=",") %>% unlist %>% as.matrix %>% as.character) %>% unique %>% paste0(collapse=","), .after = "LinkedGenes_binomialNegLog10FDR") %>% mutate(.after = "LinkedGenes_LinkedToTopic_CandidateGene", LinkedGenes_Expected = expected.nLinkedGenes, LinkedGenes_Enrichment = LinkedGenes_LinkedToTopic_CandidateGene / LinkedGenes_Expected) %>% mutate(.before = "Regulator_enrichment", Regulator_Expected = expected.Regulators) %>% mutate(.before = "ProgramGene_enrichment", ProgramGene_Expected = expected.ProgramGenes) %>% as.data.frame %>% ## end of adapted code ## arrange(LinkedGenes_binomial.p.adjust) %>% mutate(test.name = regulator.programGene.test.pairs$name[i], .after = "ProgramID") } regulator.programGene.combined.pval.nonbatch <- do.call(rbind, regulator.programGene.combined.pval.nonbatch.list) %>% arrange(LinkedGenes_ChiSquare.p.adjust, desc(LinkedGenes_LinkedToTopic_CandidateGene)) write.table(regulator.programGene.combined.pval.nonbatch, file=paste0(OUTDIR, "/", opt$trait.name, ".program_prioritization.txt"), sep="\t", quote=F, row.names=F) ## create a table with V2G2P prioritized genes for Program Genes fdr.thr <- 0.05 prioritized.program.genes.df <- regulator.programGene.combined.pval.nonbatch %>% subset(ProgramGene_fisher.p.adjust < fdr.thr) %>% select(ProgramID, ProgramGene_Gene_LinkedToTopic_CandidateGene) %>% separate_rows(ProgramGene_Gene_LinkedToTopic_CandidateGene, sep=",", convert=F) %>% `colnames<-`(c("ProgramID", "V2G2P Program Gene")) %>% as.data.frame #################################################################################################### ## plots ## adapted from TeloHAEC_Perturb-seq_2kG/221115_stimulation_condition_V2G/221115_compute_enrichment.R source('/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/figures/helper_scripts/plot_helper_functions.R') ## reload data regulator.programGene.combined.pval.nonbatch <- read.delim(paste0(OUTDIR, "/", opt$trait.name, ".program_prioritization.txt"), stringsAsFactors=F) ## regulator.programGene.combined.pval.nonbatch <- read.delim(paste0(OUTDIR, opt$trait.name, ".combinedRegulatorProgramGeneEnrichmentTest.to.prioritize.topics.nonbatch.txt"), stringsAsFactors=F) theme.here <- theme(legend.key.size=unit(0.1, unit="in"), legend.text = element_text(size=5), legend.position = "bottom", legend.direction="vertical", legend.margin = margin(1,1,1,1,unit="pt"), panel.spacing = unit(1, units="pt")) mytheme <- theme_classic() + theme(axis.text = element_text(size = 5), axis.title = element_text(size = 6), plot.title = element_text(hjust = 0.5, face = "bold", size=6), axis.line = element_line(color = "black", size = 0.25), axis.ticks = element_line(color = "black", size = 0.25)) if(opt$perturbSeq) { category_ary <- c("LinkedGenes", "ProgramGene", "Regulator") } else { category_ary <- c("ProgramGene") } fdr.thr <- 0.05 for(category in category_ary) { expectedNumEnrichedGenes <- regulator.programGene.combined.pval.nonbatch %>% pull(get(paste0(category, "_Expected"))) %>% unique toplot <- regulator.programGene.combined.pval.nonbatch %>% mutate(significant = ifelse(get(paste0(category, "_", ifelse(category == "LinkedGenes", "fisher", "fisher"), ".p.adjust")) < fdr.thr, ifelse(get(paste0(category, "_", ifelse(category == "LinkedGenes", "fisher", "fisher"), ".p.adjust")) < (fdr.thr / 10), ifelse(get(paste0(category, "_", ifelse(category == "LinkedGenes", "fisher", "fisher"), ".p.adjust")) < (fdr.thr / 100), "***", "**"), ## change to "ChiSquare" in the true slot if you want to combine regulator and program genes by Fisher's method "*"), "")) %>% ## add.df.Program.name %>% ## todo (generalize) arrange(desc(get(paste0(category, "_LinkedToTopic_CandidateGene")))) ## numSignificantPrograms <- nrow(toplot) if(grepl("2kG.library", SAMPLE)) { toplot <- toplot %>% add.df.Program.name } else { toplot <- toplot %>% mutate(truncatedLabel = ProgramID) } if(category == "LinkedGenes") { toplot <- toplot %>% arrange(desc(LinkedGenes_LinkedToTopic_CandidateGene), desc(InGeneSet_Regulator_and_ProgramGene), desc(Unique_ProgramGene_InGeneSet_CandidateGene), desc(Unique_Regulator_Gene_InGeneSet_CandidateGene)) } label.order <- toplot$truncatedLabel %>% rev toplot <- toplot %>% mutate(truncatedLabel = factor(truncatedLabel, levels=label.order)) if(category == "LinkedGenes") { maxNumGenes <- toplot$LinkedGenes_LinkedToTopic_CandidateGene %>% max toplot.here <- toplot %>% select(truncatedLabel, InGeneSet_Regulator_and_ProgramGene, Unique_Regulator_InGeneSet_CandidateGene, Unique_ProgramGene_InGeneSet_CandidateGene, significant) %>% melt(id.vars=c("truncatedLabel", "significant"), value.name="numGenes", variable.name="LinkType") %>% mutate(LinkTypeText = ifelse(grepl("_and_", LinkType), "Regulator and Co-regulated Genes", ifelse(grepl("Regulator", LinkType), "Regulators", "Co-regulated Genes")), LinkTypeText = factor(LinkTypeText, levels=c("Co-regulated Genes", "Regulators", "Regulator and Co-regulated Genes") %>% rev)) p <- toplot.here %>% ggplot(aes(x=truncatedLabel, y=numGenes, fill=LinkTypeText)) + geom_col() + coord_flip() + mytheme + geom_text(aes(label=significant, y=maxNumGenes*1.1), nudge_x=-0.5, size=3, color='gray') + scale_fill_manual(name = "", values = c("gray30", "#38b4f7", "#0141a8") %>% rev) + xlab("Programs") + ylab("# Genes Linked to Program") + ggtitle(paste0(opt$trait.name, " GWAS trait")) + theme.here + geom_hline(yintercept=expectedNumEnrichedGenes, linetype="dashed", color="gray") } else { toplot.here <- toplot %>% select(truncatedLabel, paste0(category, "_LinkedToTopic_CandidateGene"), significant) p <- toplot.here %>% ggplot(aes(x=truncatedLabel, y=get(paste0(category, "_LinkedToTopic_CandidateGene")))) + geom_col(fill="gray30") + coord_flip() + mytheme + geom_text(aes(label=significant, y=max(get(paste0(category, "_LinkedToTopic_CandidateGene")))*1.1), nudge_x=-0.5, size=3, color='gray') + xlab("Programs") + ylab("# Genes Linked to Program") + ggtitle(paste0(opt$trait.name, " GWAS trait")) + theme.here + geom_hline(yintercept=expectedNumEnrichedGenes, linetype="dashed", color="gray") } ## if(opt$perturbSeq) p <- p + scale_fill_manual(name = "", values = c("gray30", "#38b4f7", "#0141a8") %>% rev) filename <- paste0(FIGDIR, "/", SAMPLE, "_K", k, "_dt_", DENSITY.THRESHOLD, "_", opt$trait.name, "_", category, "_GeneCountBarPlot") pdf(paste0(filename, ".pdf"), width=2.5, height=3/60*k+1) print(p) dev.off() ## Enrichment Plot column.here <- ifelse(category == "LinkedGenes", "LinkedGenes_Enrichment", paste0(category, "_enrichment")) toplot <- toplot %>% arrange(desc(get(column.here))) label.order <- toplot$truncatedLabel %>% rev toplot <- toplot %>% mutate(truncatedLabel = factor(truncatedLabel, levels=label.order)) p <- toplot %>% ggplot(aes(x=truncatedLabel, y=get(column.here))) + geom_col(fill="gray30") + coord_flip() + mytheme + geom_text(aes(label=significant, y=max(get(column.here))*1.1), nudge_x=-0.5, size=3, color='gray') + xlab("Program") + ylab(paste0(category, "Enrichment")) + ggtitle(paste0(opt$trait.name, " GWAS trait")) + theme.here filename <- paste0(FIGDIR, "/", SAMPLE, "_K", k, "_dt_", DENSITY.THRESHOLD, "_", opt$trait.name, "_", category, "_EnrichmentBarPlot") pdf(paste0(filename, ".pdf"), width=2.5, height=3/60*k+1) print(p) dev.off() } ## output V2G2P genes fdr.thr <- 0.05 for(key in category_ary) { significantGenes.df <- regulator.programGene.combined.pval.nonbatch %>% subset(get(paste0(key, "_fisher.p.adjust")) < fdr.thr) %>% select(ProgramID, paste0(key, "_Gene_LinkedToTopic_CandidateGene")) %>% as.data.frame write.table(significantGenes.df, paste0(OUTDIR, "/significant", key, ".df.txt"), sep="\t", quote=F, row.names=F) significantGenes <- significantGenes.df %>% pull(get(paste0(key, "_Gene_LinkedToTopic_CandidateGene"))) %>% paste0(collapse=",") %>% strsplit(split=",") %>% unlist %>% unique write.table(significantGenes, paste0(OUTDIR, "/significant", key, ".txt"), sep="\n", quote=F, row.names=F, col.names=F) significantGenes.formatted.df <- significantGenes.df %>% separate_rows(paste0(key, "_Gene_LinkedToTopic_CandidateGene")) %>% mutate(t = gsub(paste0("K", k, "_"), "", ProgramID)) %>% arrange(t) %>% group_by(get(paste0(key, "_Gene_LinkedToTopic_CandidateGene"))) %>% summarize(ProgramID = paste0(ProgramID, collapse=",")) %>% `colnames<-`(c(paste0(key,"_Gene_LinkedToTopic_CandidateGene"), "ProgramID")) %>% as.data.frame write.table(significantGenes.formatted.df, paste0(OUTDIR, "/significant", key, ".formatted.df.txt"), sep="\t", quote=F, row.names=F) } ## add program gene and regulator membership to linked genes table if(opt$perturbSeq) { linkedSignificantGenes.formatted.df <- read.delim(paste0(OUTDIR, "/significantLinkedGenes.formatted.df.txt"), stringsAsFactors=F) %>% `colnames<-`(c("Gene", "LinkedPrograms")) for(key in c("ProgramGene", "Regulator")) { assign(paste0(key, "SignificantGenes.formatted.df"), regulator.programGene.combined.pval.nonbatch %>% subset(get(paste0("LinkedGenes_fisher.p.adjust")) < fdr.thr) %>% select(ProgramID, paste0(key, "_Gene_LinkedToTopic_CandidateGene")) %>% separate_rows(paste0(key, "_Gene_LinkedToTopic_CandidateGene")) %>% mutate(t = gsub(paste0("K", k, "_"), "", ProgramID)) %>% arrange(t) %>% group_by(get(paste0(key, "_Gene_LinkedToTopic_CandidateGene"))) %>% summarize(ProgramID = paste0(ProgramID, collapse=",")) %>% `colnames<-`(c("Gene", key)) %>% as.data.frame) } significantGenes.formatted.df <- merge(linkedSignificantGenes.formatted.df, ProgramGeneSignificantGenes.formatted.df, by="Gene", all.x=T) %>% merge(RegulatorSignificantGenes.formatted.df, by="Gene", all.x=T) %>% `colnames<-`(c("Gene", "LinkedProgram", "PartOfProgram", "RegulatorOfProgram")) write.table(significantGenes.formatted.df, paste0(OUTDIR, "/significantLinkedGenes.formatted.df.txt"), sep="\t", quote=F, row.names=F) } |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 | library(conflicted) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") packages <- c("optparse","dplyr", "cowplot", "ggplot2", "gplots", "data.table", "reshape2", "tidyr", "grid", "gtable", "gridExtra","ggrepel","ramify", "ggpubr","gridExtra", "org.Hs.eg.db","limma","fgsea", "conflicted", "cluster","textshape","readxl", "ggdist", "gghalves", "Seurat", "writexl", "stringi") xfun::pkg_attach(packages) conflict_prefer("select","dplyr") # multiple packages have select(), prioritize dplyr conflict_prefer("melt", "reshape2") conflict_prefer("slice", "dplyr") conflict_prefer("summarize", "dplyr") conflict_prefer("filter", "dplyr") conflict_prefer("combine", "dplyr") conflict_prefer("list", "base") conflict_prefer("desc", "dplyr") conflict_prefer("Position", "ggplot2") conflict_prefer("first", "dplyr") # setwd("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/figures/helper_scripts/") # source("../../figures/helper_scripts/plot_helper_functions.R") # source("../../figures/helper_scripts/load_data.R") # source("../../figures/helper_scripts/load_edgeR.R") # setwd("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/sample_cNMF_PoPS_input_table/") source("./workflow/scripts/helper_functions.R") option.list <- list( make_option("--figdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/figures/all_genes/2kG.library.ctrl.only/K25/threshold_0_2/", help="Figure directory"), make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211206_ctrl_only_snakemake/analysis/all_genes/2kG.library.ctrl.only/K25/threshold_0_2/", help="Output directory"), ## make_option("--olddatadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/data/", help="Input 10x data directory"), make_option("--datadir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/", help="Input 10x data directory"), make_option("--topic.model.result.dir", type="character", default="/scratch/groups/engreitz/Users/kangh/Perturb-seq_CAD/210625_snakemake_output/top3000VariableGenes_acrossK/2kG.library/", help="Topic model results directory"), make_option("--sampleName", type="character", default="2kG.library.ctrl.only", help="Name of Samples to be processed, separated by commas"), make_option("--K.val", type="numeric", default=25, help="K value to analyze"), ## make_option("--cell.count.thr", type="numeric", default=2, help="filter threshold for number of cells per guide (greater than the input number)"), ## make_option("--guide.count.thr", type="numeric", default=1, help="filter threshold for number of guide per perturbation (greater than the input number)"), make_option("--density.thr", type="character", default="0.2", help="concensus cluster threshold, 2 for no filtering"), make_option("--perturbSeq", type="logical", default=F, help="T for Perturb-seq experiment, F for no perturbation"), ## PoPS results make_option("--preds_with_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6_cNMF60.preds", help="PoPS Score with cNMF input"), make_option("--preds_without_cNMF", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/CAD_aug6.preds", help="PoPS Score with cNMF input"), ## summary plot parameters make_option("--test.type", type="character", default="per.guide.wilcoxon", help="Significance test to threshold perturbation results"), make_option("--adj.p.value.thr", type="numeric", default=0.1, help="adjusted p-value threshold"), make_option("--recompute", type="logical", default=F, help="T for recomputing statistical tests and F for not recompute") ) opt <- parse_args(OptionParser(option_list=option.list)) ## ## all genes directories (for sdev) ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/figures/2kG.library/all_genes/2kG.library/K60/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/" ## opt$K.val <- 60 ## opt$sampleName <- "2kG.library" ## opt$perturbSeq <- TRUE ## ## K562 gwps 2k overdispersed genes ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K80/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K80/threshold_0_2/" ## opt$K.val <- 80 ## opt$sampleName <- "WeissmanK562gwps" ## opt$perturbSeq <- TRUE ## ## K562 gwps 2k overdispersed genes ## opt$figdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/figures/top2000VariableGenes/WeissmanK562gwps/K90/" ## opt$outdir <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K90/threshold_0_2/" ## opt$K.val <- 90 ## opt$sampleName <- "WeissmanK562gwps" ## opt$perturbSeq <- TRUE OUTDIRSAMPLE <- OUTDIR <- opt$outdir DATADIR <- opt$datadir SAMPLE=opt$sampleName DENSITY.THRESHOLD <- gsub("\\.","_", opt$density.thr) k <- opt$K.val ## OUTDIRSAMPLE=paste0(OUTDIR, SAMPLE, "/K",k,"/threshold_", DENSITY.THRESHOLD, "/") SUBSCRIPT=paste0("k_", k,".dt_",DENSITY.THRESHOLD,".minGuidePerPtb_",opt$guide.count.thr,".minCellPerGuide_", opt$cell.count.thr) SUBSCRIPT.SHORT=paste0("k_", k,".dt_",DENSITY.THRESHOLD) if(!dir.exists(OUTDIR)) dir.create(OUTDIR) fdr.thr <- 0.05 db <- ifelse(grepl("mouse", SAMPLE), "org.Mm.eg.db", "org.Hs.eg.db") library(!!db) ## load the appropriate database if(grepl("2kG.library", SAMPLE)) { ## map ids ## slow version x <- org.Hs.egENSEMBL mapped_genes <- mappedkeys(x) entrez.to.ensembl <- as.list(x[mapped_genes]) # EntrezID to Ensembl ensembl.to.entrez <- as.list(org.Hs.egENSEMBL2EG) # Ensembl to EntrezID y <- org.Hs.egGENENAME y_mapped_genes <- mappedkeys(y) entrez.to.genename <- as.list(y[y_mapped_genes]) genename.to.entrez <- as.list(org.Hs.egGENENAME) ## map between EntrezID and Gene Symbol z <- org.Hs.egSYMBOL z_mapped_genes <- mappedkeys(z) entrez.to.symbol <- as.list(z[z_mapped_genes]) entrez.to.symbol <- as.list(org.Hs.egSYMBOL) symbol.to.entrez <- as.list(org.Hs.egSYMBOL2EG) } ## load 10X reference gtf.10X.df <- readRDS(paste0("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/refdata-cellranger-arc-GRCh38-2020-A_genes.gtf_df.RDS")) ## load 10X gtf file Gene.ENSEMBL.10X.df <- gtf.10X.df %>% mutate(ENSGID = gene_id, Gene10X = gene_name) %>% select(Gene10X, ENSGID) %>% unique ## gtf <- importGTF("/home/groups/engreitz/Software/cellranger-arc-1.0.1/refdata-cellranger-arc-GRCh38-2020-A/genes/genes.gtf") ## helper function to map between ENSGID and SYMBOL map.ENSGID.SYMBOL <- function(df) { ## need column `Gene` to be present in df ## detect gene data type (e.g. ENSGID, Entrez Symbol) gene.type <- ifelse(nrow(df) == sum(as.numeric(grepl("^ENS", df$Gene))), "ENSGID", "Gene") if(gene.type == "ENSGID") { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "ENSEMBL", column = "SYMBOL") df <- df %>% mutate(ENSGID = Gene, Gene = mapped.genes) } else { mapped.genes <- mapIds(get(db), keys=df$Gene, keytype = "SYMBOL", column = "ENSEMBL") df <- df %>% mutate(ENSGID = mapped.genes) } df <- df %>% mutate(Gene = ifelse(is.na(Gene), "NA", Gene), ENSGID = ifelse(is.na(ENSGID), "NA", ENSGID)) df <- df %>% merge(Gene.ENSEMBL.10X.df, by="ENSGID", all.x=T) df <- df %>% mutate(Gene = ifelse(Gene == "NA", Gene10X, Gene)) notMatched.df <- df %>% subset(is.na(Gene10X)) if(nrow(notMatched.df) > 0) { notMatched.index <- notMatched.df %>% rownames match.df <- merge(notMatched.df %>% select(-ENSGID, -Gene10X), Gene.ENSEMBL.10X.df, by.x="Gene", by.y="Gene10X") %>% mutate(Gene10X = Gene) %>% select(all_of(df %>% colnames)) matched.index <- notMatched.df %>% subset(Gene %in% c(match.df$Gene %>% unique)) %>% rownames %>% as.numeric df <- rbind(df[-matched.index,], match.df) } df <- df %>% mutate(OriginalGene = Gene, Gene = Gene10X) ## toconvert <- df %>% subset(Gene != Gene10X) %>% select(ENSGID, Gene, Gene10X) ## toconvert.index <- toconvert %>% ## df %>% subset(Gene %in% Gene.ENSEMBL.10X.df$Gene10X & is.na(ENSGID)) return(df) } ## ## load Perturb-seq analysis results ## OUTDIRSAMPLE <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210707_snakemake_maxParallel/analysis/2kG.library/all_genes/2kG.library/K60/threshold_0_2/" ## SUBSCRIPT <- "k_60.dt_0_2.minGuidePerPtb_1.minCellPerGuide_2" file.name <- paste0(OUTDIRSAMPLE,"/cNMF_results.",SUBSCRIPT.SHORT,".RData") print(file.name) if(file.exists((file.name))) { print(paste0("loading ",file.name)) load(file.name) } if(opt$perturbSeq) { MAST.file.name <- paste0(OUTDIRSAMPLE, "/", SAMPLE, "_MAST_DEtopics.txt") print(paste0("loading ", MAST.file.name)) MAST.df <- read.delim(MAST.file.name, stringsAsFactors=F, check.names=F) if(grepl("2kG.library", SAMPLE)){ MAST.df <- MAST.df %>% subset(zlm.model.name == "batch.correction") multiTarget.genes <- MAST.df %>% subset(grepl("multiTarget", perturbation)) %>% pull(perturbation) %>% unique %>% gsub("_multiTarget", "", .) MAST.df <- MAST.df %>% mutate(ProgramID = gsub("topic_", "K60_", primerid)) %>% subset(!(perturbation %in% multiTarget.genes) & ## for multiTarget genes, swap in the set where cells have GeneA and GeneA-and-GeneB guides. !grepl("-and-", perturbation)) %>% ## remove the set where cells only have GeneA-and-GeneB guides. mutate(perturbation = gsub("_multiTarget", "", perturbation)) %>% group_by(zlm.model.name) %>% mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method="fdr")) %>% subset(zlm.model.name %in% c("batch.correction")) %>% select(-zlm.model.name) %>% group_by(ProgramID) %>% arrange(desc(coef)) %>% mutate(coef_rank = 1:n()) %>% as.data.frame } } ## make theta.zscore.rank.df ## if(!("theta.zscore.rank.df" %in% ls())) { theta.zscore.rank.df <- theta.zscore %>% as.data.frame %>% mutate(Gene = rownames(.)) %>% melt(id.vars="Gene", variable.name="ProgramID", value.name="zscore.specificity") %>% mutate(ProgramID = paste0("K", k, "_", ProgramID)) %>% group_by(ProgramID) %>% arrange(desc(zscore.specificity)) %>% mutate(zscore.specificity.rank = 1:n()) %>% ungroup %>% arrange(ProgramID, zscore.specificity.rank) %>% as.data.frame ## } theta.zscore.rank.df <- theta.zscore.rank.df %>% map.ENSGID.SYMBOL if(grepl("2kG.library", SAMPLE)) { ## load PoPS outputs preds <- read.table(file=opt$preds_without_cNMF,header=T, stringsAsFactors=F, sep="\t") ## load PoPS processed gene x feature score ## load("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/coefs.marginals.feature.outer.prod.RDS") ## takes a while OUTDIRPOPS <- "/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/" ## PoPS_Score.coefs.all.outer <- read.delim(paste0(OUTDIRPOPS, "/coefs.all.feature.outer.prod.txt"), stringsAsFactors=F) ## also takes a long time to load ## top.genes.in.top.features <- read.delim(file=paste0(OUTDIRPOPS, "/top.genes.in.top.features.coefs.txt"), stringsAsFactors=F) ## load top features defining each gene's PoPS score PoPS_preds.importance.score.key <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/210831_PoPS/211108_withoutBBJ/outputs/PoPS_preds.importance.score.key.columns.txt"), stringsAsFactors=F) ## load top topic defining each gene's PoPS score PREFIX <- paste0("CAD_aug6_cNMF", k) load(paste0(OUTDIRPOPS, "/", PREFIX, "_coefs.defining.top.topic.RDS")) coefs <- read.delim(paste0(OUTDIRPOPS, "/", PREFIX, ".coefs"), header=T, stringsAsFactors=F) coefs.df <- coefs[4:nrow(coefs),] coefs.cNMF.df <- coefs.df %>% subset(grepl("zscore",parameter)) coefs.cNMF.names <- coefs.cNMF.df %>% pull(parameter) ## load EdgeR log2fcs and p-values log2fc.edgeR <- read.table(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/EdgeR/ALL_log2fcs_dup4_s4n3.99x.txt"), header=T, stringsAsFactors=F) p.value.edgeR <- read.table(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/data/EdgeR/ALL_Pvalues_dup4_s4n3.99x.txt"), header=T, stringsAsFactors=F) ## load known CAD gene set params.known.CAD.Gene.set <- read.delim("/oak/stanford/groups/engreitz/Users/kangh/ECPerturbSeq2021-Analysis/data/known_CAD_gene_set.txt", stringsAsFactors=F, header=F) %>% as.matrix %>% as.character } ## Helper function to add EntrezID and ENSGID add.EntrezID.ENSGID <- function(df) { return ( df %>% mutate(EntrezID = symbol.to.entrez[.$Gene %>% as.character] %>% sapply("[[",1) %>% as.character) %>% mutate(ENSGID = entrez.to.ensembl[.$EntrezID %>% as.character] %>% sapply("[[", 1) %>% as.character) ) } ## ## MAST results ## MAST.df.4n.original <- read.delim(paste0("/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220217_MAST/2kG.library4n3.99x_MAST.txt"), stringsAsFactors=F, check.names=F) ## MAST.df <- MAST.df.4n.original %>% ## subset(!grepl("multiTarget", perturbation)) %>% ## group_by(zlm.model.name) %>% ## mutate(fdr.across.ptb = p.adjust(`Pr(>Chisq)`, method="fdr")) %>% ## subset(zlm.model.name %in% c("batch.correction")) %>% ## select(-zlm.model.name) %>% ## mutate(ProgramID = gsub("topic_", "K60_", primerid)) %>% ## group_by(ProgramID) %>% ## arrange(desc(coef)) %>% ## mutate(coef_rank = 1:n()) %>% ## as.data.frame ## sigPerturbationsProgram <- MAST.df %>% ## subset(fdr.across.ptb < fdr.thr) %>% ## as.data.frame if(grepl("2kG.library", SAMPLE)) { ## curated gene name conversion list geneNameConversion <- read.delim(paste0(DATADIR, "/heterogenous_geneName_conversion_toMatchENSGID.txt"), stringsAsFactors=F) %>% separate(col=Conversion, into=c("fromGene", "toGene"), sep=" -> ", remove=F) geneNameConversionBackToPerturbseq <- geneNameConversion %>% filter(!KeepForGWASTable) ## helper function to conver gene names convert.gene.names.toMatchENSGID <- function(Gene) stri_replace_all_regex(Gene, pattern = geneNameConversion$fromGene, replace = geneNameConversion$toGene, vectorize=F) convert.gene.names.toMatchPerturbseq <- function(Gene) stri_replace_all_regex(Gene, pattern = geneNameConversionBackToPerturbseq$toGene, replace = geneNameConversionBackToPerturbseq$fromGene, vectorize=F) } ## new gene name conversion method ## 220628 ptb10xNames.df <- read.delim(paste0(DATADIR, "/220627_add_Perturbation 10X names.txt"), stringsAsFactors=F, check.names=F) perturbseq.gene.names.to10X <- function(Gene) { if(Gene %in% ptb10xNames.df$Symbol) { out <- stri_replace_all_regex(Gene, pattern = ptb10xNames.df$Symbol, replace = ptb10xNames.df$`Name used by CellRanger`, vectorize=F) } else { out <- Gene } return(out) } tenX.gene.names.toperturbseq <- function(Gene) stri_replace_all_regex(Gene, pattern = ptb10xNames.df$`Name used by CellRanger`, replace = ptb10xNames.df$Symbol, vectorize=F) print("loaded all prerequisite data") ########################################################################################## ptb.topic.thr <- 0.05 ## FDR threshold ## Column: ProgramsRegulatedByThisGene if(opt$perturbSeq) { ProgramsRegulatedByThisGene.df <- MAST.df %>% mutate(OriginalGene = perturbation, Gene = perturbseq.gene.names.to10X(perturbation), t = gsub("topic_", "", primerid) %>% as.numeric, ProgramID = paste0("K", k, "_", t)) %>% mutate(significant = fdr.across.ptb < ptb.topic.thr & (coef > log(1.1) | coef < log(0.9)), ProgramID = ifelse(significant, ProgramID, "NA")) %>% ## subset(fdr.across.ptb < ptb.topic.thr) %>% ## mutate(Gene = convert.gene.names.toMatchENSGID(perturbation), map.ENSGID.SYMBOL ## mutate(EntrezID = symbol.to.entrez[.$Gene %>% as.character] %>% sapply("[[",1) %>% as.character) %>% ## mutate(ENSGID = entrez.to.ensembl[.$EntrezID %>% as.character] %>% sapply("[[", 1) %>% as.character) ProgramsRegulatedByThisGene.tokeep <- ProgramsRegulatedByThisGene.df %>% arrange(t) %>% ## select(Gene, OriginalGene, ENSGID, EntrezID, ProgramID) %>% select(Gene, OriginalGene, Gene10X, ENSGID, ProgramID) %>% unique %>% ## group_by(ENSGID, Gene, OriginalGene, EntrezID) %>% group_by(ENSGID, Gene, Gene10X, OriginalGene) %>% summarize(ProgramsRegulatedByThisGene=paste0(ProgramID, collapse="|")) %>% mutate(ProgramsRegulatedByThisGene = gsub("NA[|]|[|]NA", "", ProgramsRegulatedByThisGene), ProgramsRegulatedByThisGene = gsub("[|]NA[|]", "|", ProgramsRegulatedByThisGene), perturbed_gene = TRUE) %>% ## clean up "NA"s ## mutate(Gene = convert.gene.names.toMatchENSGID(rownames(.))) %>% ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% as.data.frame ## if(grepl("2kG.library", SAMPLE)) { ## ProgramsRegulatedByThisGene.df <- ProgramsRegulatedByThisGene.df %>% mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(perturbation)) ## ProgramsRegulatedByThisGene.tokeep <- ProgramsRegulatedByThisGene.df %>% ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## as.data.frame ## } } ## ## ## ## Column: ProgramsInWhichGeneIsInTop100ZScoreSpecificGenes ## ## ## params.top.n.genes.in.topic <- 100 ## if(grepl("2kG.library", SAMPLE)) { ## ProgramsInWhichGeneIsExpressed.df <- theta.zscore.rank.df %>% ## mutate(OriginalGene = Gene, ## PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene), ## Gene = perturbseq.gene.names.to10X(Gene)) %>% ## ## add.EntrezID.ENSGID %>% ## as.data.frame ## } else { ## ProgramsInWhichGeneIsExpressed.df <- theta.zscore.rank.df %>% ## ## mutate(OriginalGene = Gene, ## ## PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene), ## ## Gene = perturbseq.gene.names.to10X(Gene)) %>% ## ## add.EntrezID.ENSGID %>% ## as.data.frame ## } IncNMFAnalysis.tokeep <- theta.zscore.rank.df %>% select(Gene, Gene10X, ENSGID) %>% unique %>% mutate(IncNMFAnalysis = TRUE) %>% as.data.frame params.top.n.genes.in.topic.list <- c(100, 300, 500) ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep <- Reduce( function(x, y, ...) full_join(x, y, by = c("Gene", "Gene10X", "ENSGID"), ...), ## want to join all the data frames that has columns c("Gene", "ProgramsInWhichGeneIsInTopNZScoreSpecificGenes") out <- lapply(params.top.n.genes.in.topic.list, function (params.top.n.genes.in.topic) { # for each selected number of top genes we want to include column.name <- paste0("ProgramsInWhichGeneIsInTop", params.top.n.genes.in.topic, "ZScoreSpecificGenes") # store the column name in a variable ## out <- ProgramsInWhichGeneIsExpressed.df %>% ## group_by(ProgramID) %>% ## arrange(desc(Topic.zscore)) %>% # sort the z-score within each topic ## slice(1:params.top.n.genes.in.topic) %>% # select the top N genes out <- theta.zscore.rank.df %>% mutate(IsProgramGene = zscore.specificity.rank <= params.top.n.genes.in.topic, ProgramID = ifelse(IsProgramGene, ProgramID, "NA")) %>% ## subset(zscore.specificity.rank <= params.top.n.genes.in.topic) %>% ## ungroup %>% group_by(Gene, Gene10X, ENSGID) %>% # for each gene summarize(!!column.name := paste0(ProgramID, collapse="|")) %>% # paste the topics each gene links to by "," mutate(!!column.name := gsub("NA[|]|[|]NA", "", get(column.name)), !!column.name := gsub("[|]NA[|]", "|", get(column.name))) %>% ## clean up "NA"s as.data.frame })) ## if(grepl("2kG.library", SAMPLE)) { ## ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep <- ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep %>% ## ## mutate(Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(OriginalGene = Gene) %>% ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## add.EntrezID.ENSGID %>% ## as.data.frame ## } ## ## Column: PoPSEnrichedProgramsRegulatedByThisGene ## FDR < 0.05 and in PoPS prioritized features ## if(opt$perturbSeq & grepl("2kG.library", SAMPLE)){ ## cNMF.PoPS.enriched.topics <- coefs.cNMF.names %>% gsub("zscore_K60_topic", "K60_", .) ## get the list of PoPS prioritized cNMF topics and reformat ## PoPSEnrichedProgramsRegulatedByThisGene.tokeep <- ProgramsRegulatedByThisGene.df %>% ## select(Gene, OriginalGene, PreviousConversionGene, ENSGID, ProgramID) %>% ## subset(ProgramID %in% cNMF.PoPS.enriched.topics) %>% ## group_by(ENSGID, Gene, OriginalGene, PreviousConversionGene) %>% ## summarize(PoPSEnrichedProgramsRegulatedByThisGene=paste0(ProgramID, collapse="|")) %>% ## ## mutate(Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## as.data.frame ## } ## ## Column: PoPSEnrichedProgramsInWhichGeneIsInTop100ZScoreSpecificGenes ## cNMF.PoPS.enriched.topics <- coefs.cNMF.names %>% gsub("zscore_K60_topic", "K60_", .) ## get the list of PoPS prioritized cNMF topics and reformat ## params.top.n.genes.in.topic <- 300 ## PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes.tokeep <- ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep %>% ## ProgramsInWhichGeneIsExpressed.df %>% ## subset(ProgramID %in% cNMF.PoPS.enriched.topics) %>% ## ## group_by(ProgramID) %>% ## ## arrange(desc(Topic.zscore)) %>% ## ## slice(1:params.top.n.genes.in.topic) %>% ## ## ungroup %>% ## subset(zscore.specificity.rank <= params.top.n.genes.in.topic) %>% ## group_by(Gene, PreviousConversionGene, OriginalGene, ENSGID) %>% ## summarize(PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes = paste0(ProgramID, collapse="|")) %>% ## ## mutate(Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## ## add.EntrezID.ENSGID %>% ## as.data.frame ## ## PoPS.Score ## PoPS.Score.tokeep <- preds %>% select(ENSGID, PoPS_Score) %>% ## mutate(EntrezID = ensembl.to.entrez[.$ENSGID %>% as.character] %>% sapply("[[",1) %>% as.character) %>% ## mutate(Gene = entrez.to.symbol[.$EntrezID %>% as.character] %>% sapply("[[",1) %>% as.character, ## ## Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(OriginalGene = Gene) %>% ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## subset(!(EntrezID == "NULL" & Gene == "NULL" & PreviousConversionGene == "NULL" & OriginalGene == "NULL")) %>% ## as.data.frame ## ## ## PoPS.Rank [rank of the score among genes near this CredibleSet/GWAS signal] ## ## create later when we merge in CredibleSet/GWAS signal ## ## ## Top5ProgramsThatContributeToPoPSScore ## ## Top5ProgramsThatContributeToPoPSScore.tokeep <- coefs.defining.top.topic.df %>% ## ## mutate(ProgramID = gsub("zscore_K60_topic", "K60_", topic), ## ## Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## ## group_by(Gene) %>% ## ## arrange(desc(gene.feature_x_beta)) %>% ## ## slice(1:5) %>% ## ## summarize(Top5ProgramsThatContributeToPoPSScore = paste0(ProgramID, collapse="|")) %>% ## ## add.EntrezID.ENSGID %>% ## ## as.data.frame ## ## Top5FeaturesThatContributeToPoPSScore ## Top5FeaturesThatContributeToPoPSScore.tokeep <- PoPS_preds.importance.score.key %>% ## group_by(Gene) %>% ## arrange(desc(gene.feature_x_beta)) %>% ## slice(1:5) %>% ## mutate(to.display=paste0(pathway, ":", Long_Name)) %>% ## summarize(Top5FeaturesThatContributeToPoPSScore = paste0(to.display, collapse="|")) %>% ## ## mutate(Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(OriginalGene = Gene) %>% ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## add.EntrezID.ENSGID %>% ## as.data.frame ## } ## ## DoesThisGeneWhenPerturbedRegulateAKnownCADGene ## if(opt$perturbSeq & grepl("2kG.library", SAMPLE)){ ## CAD.gene.regulator.log2fc <- log2fc.edgeR %>% ## subset(grepl(paste0(params.known.CAD.Gene.set, ":ENSG") %>% ## paste0(collapse="|"), .$gene)) ## CAD.gene.regulator.p.value <- p.value.edgeR %>% ## subset(grepl(paste0(params.known.CAD.Gene.set, ":ENSG") %>% ## paste0(collapse="|"), .$gene)) ## params.EdgeR.p.value.thr <- "0.05" ## CAD.gene.regulator.df <- CAD.gene.regulator.p.value %>% ## melt(id.vars="genes", value.name="p.value", variable.name="perturbation") %>% ## subset(p.value < params.EdgeR.p.value.thr) ## DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep <- CAD.gene.regulator.df %>% ## separate(genes, into=c("Gene", "ENSGID"), sep=":") %>% ## group_by(perturbation) %>% ## summarize(DoesThisGeneWhenPerturbedRegulateAKnownCADGene = paste0(Gene, collapse="|")) ## colnames(DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep)[colnames(DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep) == "perturbation"] <- "Gene" ## DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep <- DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep %>% ## ## mutate(Gene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(OriginalGene = Gene) %>% ## mutate(PreviousConversionGene = convert.gene.names.toMatchENSGID(Gene)) %>% ## mutate(Gene = perturbseq.gene.names.to10X(Gene)) %>% ## add.EntrezID.ENSGID %>% ## as.data.frame ## } ## put together all if(opt$perturbSeq){ ## if(grepl("2kG.library", SAMPLE)) { ## list.all <- list(ProgramsRegulatedByThisGene = ProgramsRegulatedByThisGene.tokeep, ## ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes = ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep, ## PoPSEnrichedProgramsRegulatedByThisGene = PoPSEnrichedProgramsRegulatedByThisGene.tokeep, ## PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes = PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes.tokeep, ## PoPS.Score = PoPS.Score.tokeep, ## DoesThisGeneWhenPerturbedRegulateAKnownCADGene = DoesThisGeneWhenPerturbedRegulateAKnownCADGene.tokeep, ## IncNMFAnalysis = IncNMFAnalysis.tokeep) ## caveat: Not all perturbations are in common Gene Symbol format (e.g. FAM212A, Icam2, Slc9a3r2) ## } else { list.all <- list(ProgramsRegulatedByThisGene = ProgramsRegulatedByThisGene.tokeep, ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes = ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep, IncNMFAnalysis = IncNMFAnalysis.tokeep) ## caveat: Not all perturbations are in common Gene Symbol format (e.g. FAM212A, Icam2, Slc9a3r2) ## } } else { list.all <- list(ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes = ProgramsInWhichGeneIsInTopNListZScoreSpecificGenes.tokeep, ## PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes = PoPSEnrichedProgramsInWhichGeneIsInTop300ZScoreSpecificGenes.tokeep, ## PoPS.Score = PoPS.Score.tokeep, IncNMFAnalysis = IncNMFAnalysis.tokeep) } ## if(grepl("2kG.library", SAMPLE)) { ## df.all <- Reduce(function(x, y, ...) full_join(x, y %>% select(-OriginalGene, -PreviousConversionGene), by = c("Gene", "ENSGID", "EntrezID"), ...), list.all) ## } else { df.all <- Reduce(function(x, y, ...) full_join(x, y, by = c("Gene", "Gene10X", "ENSGID"), ...), list.all) ## } df.all[df.all == "NULL"] <- NA na.row.index <- which(df.all$Gene %>% is.na & df.all$ENSGID %>% is.na & df.all$PoPS_Score %>% is.na) ## Perturb-seq specific trimming of duplicated Gene names if(opt$perturbSeq) { ## if(grepl("2kG.library", SAMPLE)) { ## df.all <- df.all %>% subset(!(is.na(Gene) & is.na(ProgramsRegulatedByThisGene) & !(Gene %in% c(theta.zscore %>% rownames, MAST.df$perturbation %>% unique)) & is.na(ENSGID))) %>% ## combine duplicated Gene names' rows ## group_by(Gene, ENSGID) %>% ## consider implement match with 10X ## summarize_all(., function(x) { ## out <- paste0(x %>% unique, collapse="|") ## if(grepl("TRUE", out %>% as.character)) return("TRUE") else return(out) ## }) %>% ## as.data.frame ## } else { df.all <- df.all %>% subset(!(is.na(Gene) & is.na(ProgramsRegulatedByThisGene) & !(Gene %in% c(theta.zscore %>% rownames, MAST.df$perturbation %>% unique)) & is.na(ENSGID))) %>% ## combine duplicated Gene names' rows group_by(Gene, Gene10X, ENSGID) %>% summarize_all(., function(x) { out <- paste0(x %>% unique, collapse="|") if(grepl("TRUE", out %>% as.character)) return("TRUE") else return(out) }) %>% mutate(IncNMFAnalysis = ifelse(IncNMFAnalysis == "TRUE", TRUE, FALSE), perturbed_gene = ifelse(perturbed_gene == "TRUE", TRUE, FALSE)) %>% as.data.frame ## } } ## write table print("writing prepared table to file") write.table(df.all, file=paste0(OUTDIRSAMPLE, "/prepare_compute_enrichment.txt"), row.names=F, quote=F, sep="\t") |
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 | suppressPackageStartupMessages(library(optparse)) option.list <- list( # make_option("--outdir", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2106_FT007_Analysis/outputs/", help="Output directory"), # make_option("--datadir",type="character", default="/oak/stanford/groups/engreitz/Users/kangh/process_sequencing_data/210611_FT007_CM_CMO/gex_FT007_50k/outs/filtered_feature_bc_matrix/", help="Data directory"), # make_option("--project",type="character",default="/oak/stanford/groups/engreitz/Users/kangh/heart_atlas/2106_FT007_Analysis/",help="Project Directory"), make_option("--sampleName",type="character",default="gex_FT007_50k", help="Sample name"), make_option("--inputSeuratObject", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/data/FT010_fresh_3min.SeuratObject.RDS", help="Path to the Seurat Object"), make_option("--output_h5ad", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/data/FT010_fresh_3min.h5ad"), make_option("--output_gene_name_txt", type="character", default="/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/211011_Perturb-seq_Analysis_Pipeline_scratch/analysis/data/FT010_fresh_3min.h5ad.all.genes.txt"), make_option("--minUMIsPerCell", type="numeric", default=200), make_option("--minUniqueGenesPerCell", type="numeric", default=200) # make_option("--recompute", type="logical", default=F, help="T for recomputing UMAP from 10x count matrix") ) opt <- parse_args(OptionParser(option_list=option.list)) suppressPackageStartupMessages(library(SeuratObject)) suppressPackageStartupMessages(library(Seurat)) suppressPackageStartupMessages(library(reticulate)) suppressPackageStartupMessages(library(dplyr)) suppressPackageStartupMessages(library(ggplot2)) suppressPackageStartupMessages(library(data.table)) suppressPackageStartupMessages(library(tidyr)) suppressPackageStartupMessages(library(Matrix)) # suppressPackageStartupMessages(library(readxl)) # suppressPackageStartupMessages(library(ggrepel)) # mytheme <- theme_classic() + theme(axis.text = element_text(size = 13), axis.title = element_text(size = 15), plot.title = element_text(hjust = 0.5)) # source("/oak/stanford/groups/engreitz/Users/kangh/2009_endothelial_perturbseq_analysis/topicModelAnalysis.functions.R") ## ## sdev for mouse ENCODE adrenal data ## opt$inputSeuratObject <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230117_snakemake_mouse_ENCODE_adrenal/analysis/data/mouse_ENCODE_adrenal.SeuratObject.RDS" ## ## sdev for IGVF_b01_LeftCortex ## opt$inputSeuratObject <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/data/IGVF_b01_LeftCortex.SeuratObject.RDS" ## opt$output_h5ad <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/data/IGVF_b01_LeftCortex.h5ad" ## opt$output_gene_name_txt <- "/oak/stanford/groups/engreitz/Users/kangh/IGVF/Cellular_Programs_Networks/230706_snakemake_igvf_b01_LeftCortex/analysis/data/IGVF_b01_LeftCortex.h5ad.all.genes.txt" ## opt$sampleName <- "IGVF_b01_LeftCortex" ####################################################################### ## Constants # PROJECT=opt$project # DATADIR=opt$datadir # OUTDIR=opt$outdir # FIGDIR= paste0(PROJECT, "/figures/") SAMPLE=opt$sampleName # OUTDIRSAMPLE=paste0(OUTDIR,"/",SAMPLE,"/") # FIGDIRSAMPLE=paste0(FIGDIR,SAMPLE,"/") # palette = colorRampPalette(c("#38b4f7", "white", "red"))(n=100) # # create dir if not already # check.dir <- c(OUTDIR, OUTDIRSAMPLE) # invisible(lapply(check.dir, function(x) { if(!dir.exists(x)) dir.create(x) })) ## convert Seurat Object to h5ad anndata <- import("anndata", convert = FALSE) ## load AnnData module ## load Seruat Object s <- readRDS(opt$inputSeuratObject) print("finished loading Seurat Object") ## set assay to RNA first to avoid error such as "SCT is not an assay present in the given object. Available assays are: RNA" (could happen when getting gene names). DefaultAssay(object = s) <- "RNA" ## only keep RNA counts and metadata. Remove any other items (e.g. PCA, SCTransform, UMAP) to avoid errors. s.meta <- s[[]] s.count <- s@assays$RNA@counts s <- CreateSeuratObject(counts = s.count, project = SAMPLE, meta.data = s.meta) ## filter genes and cells again, for the cases when input file is Seurat Object and create_seurat_object step is skipped ## remove non-protein coding genes and genes detected in fewer than 10 cells tokeep <- which(!(grepl("^LINC|^[A-Za-z][A-Za-z][0-9][0-9][0-9][0-9][0-9][0-9]\\.|^Gm[0-9]|[0-9]Rik$|-ps", s %>% rownames))) s.subset <- s[tokeep,] print('finished subsetting to remove non-coding genes') s.subset <- subset(s.subset, subset= nCount_RNA > opt$minUMIsPerCell & nFeature_RNA > opt$minUniqueGenesPerCell) # remove cells with less than predefined number of UMIs (e.g. 200 UMIs) and less than a number of genes (e.g. 200 genes) print('removed cells with less than 200 UMIs and less than 200 genes') ## tokeep <- which(s.subset@assays$RNA@counts %>% apply(1, sum) > 10) # keep genes detected in more than 10 UMIs ## tokeep <- tryCatch( ## { ## which(s.subset@assays$RNA@counts %>% apply(1, function(x) ((x > 0) %>% as.numeric %>% sum > 10))) ## }, ## error = function(cond) { ## message("cannot genes expressed in less than 10 cells") ## return(seq(1, s.subset %>% nrow(), by=1)) ## }, ## finally = { ## message('removed genes expressed in less than 10 cells') ## } ## ) ## s <- s.subset[tokeep,] s <- s.subset adata <- anndata$AnnData( X = t(GetAssayData(object = s)), obs = data.frame(s@meta.data), var = s %>% rownames %>% as.data.frame %>% `rownames<-`(s %>% rownames) %>% `colnames<-`("Gene") ) ## adata <- anndata$AnnData( ## X = t(GetAssayData(object = s) %>% as.matrix), ## obs = data.frame(s@meta.data), ## var = s %>% rownames %>% as.data.frame %>% `rownames<-`(s %>% rownames) %>% `colnames<-`("Gene") ## ) anndata$AnnData$write(adata, opt$output_h5ad) ## write gene names gene.names <- s %>% rownames write.table(gene.names, file=opt$output_gene_name_txt, col.names=F, row.names=F, quote=F, sep="\t") |
7
of
scripts/seurat_to_h5ad.R
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | import numpy as np import scipy import pandas as pd # import scipy.sparse as sp import scanpy as sc import argparse import os import re from cnmf import cNMF from sklearn.decomposition import PCA ## argparse parser = argparse.ArgumentParser() ## add arguments parser.add_argument('--path_to_topics', type=str, help='path to the topic (cNMF directory) to project data on', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes_acrossK/') parser.add_argument('--topic_sampleName', type=str, help='sample name for topics to project on, use the same sample name as used for the cNMF directory', default='2kG.library_overdispersedGenes') parser.add_argument('--X_normalized', type=str, help='path to normalized input cell x gene matrix from cNMF pipeline', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes_acrossK/2kG.library_overdispersedGenes/cnmf_tmp/2kG.library_overdispersedGenes.norm_counts.h5ad') parser.add_argument('--outdir', dest = 'outdir', type=str, help = 'path to output directory', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/220716_snakemake_overdispersedGenes/analysis/top2000VariableGenes/2kG.library_overdispersedGenes/K60/threshold_0_2/') parser.add_argument('--k', dest = 'k', type=int, help = 'number of components', default='60') parser.add_argument('--density_threshold', dest = 'density_threshold', type=float, help = 'component spectra clustering threshold, 2 for no filtering, recommend 0_2 (means 0.2)', default="0.2") # ## sdev for K562 gwps # parser.add_argument('--path_to_topics', type=str, help='path to the topic (cNMF directory) to project data on', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes_acrossK/') # parser.add_argument('--topic_sampleName', type=str, help='sample name for topics to project on, use the same sample name as used for the cNMF directory', default='WeissmanK562gwps') # parser.add_argument('--X_normalized', type=str, help='path to normalized input cell x gene matrix from cNMF pipeline', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes_acrossK/WeissmanK562gwps/cnmf_tmp/WeissmanK562gwps.norm_counts.h5ad') # parser.add_argument('--outdir', dest = 'outdir', type=str, help = 'path to output directory', default='/oak/stanford/groups/engreitz/Users/kangh/TeloHAEC_Perturb-seq_2kG/230104_snakemake_WeissmanLabData/analysis/top2000VariableGenes/WeissmanK562gwps/K35/threshold_0_2/') # parser.add_argument('--k', dest = 'k', type=int, help = 'number of components', default='35') # parser.add_argument('--density_threshold', dest = 'density_threshold', type=float, help = 'component spectra clustering threshold, 2 for no filtering, recommend 0_2 (means 0.2)', default="0.2") args = parser.parse_args() # ## sdev debug for Disha's error # args.X_normalized = "/oak/stanford/groups/engreitz/Users/kangh/scratch_space/230612_debug_cNMF_pipeline_variance_explained/Merge_Pauletal_subset_SMC.norm_counts.h5ad" # args.outdir = "/oak/stanford/groups/engreitz/Users/kangh/scratch_space/230612_debug_cNMF_pipeline_variance_explained/" sample = args.topic_sampleName # output_sample = args.output_sampleName # tpm_counts_path = args.tpm_counts_path OUTDIR = args.outdir selected_K = args.k density_threshold = args.density_threshold output_directory = args.path_to_topics run_name = args.topic_sampleName if not os.path.exists(OUTDIR): raise Exception("Output directory does not exist") cnmf_obj = cNMF(output_dir=output_directory, name=run_name) usage_norm, gep_scores, gep_tpm, topgenes = cnmf_obj.load_results(K=selected_K, density_threshold=density_threshold) # X_original = sc.read_h5ad(args.tpm_counts_path) X_norm = sc.read_h5ad(args.X_normalized) # sc.pp.normalize_per_cell(X_original, counts_per_cell_after=1e6) ## normalize X to TPM # X_original.X[0:10,:].todense().sum(axis=1) ## check normalization results ## functions def compute_Var(X): if scipy.sparse.issparse(X): return np.sum(np.var(X.todense(), axis=0, ddof=1)) else: return np.sum(np.var(X, axis=0, ddof=1)) # ## first turn X into TPM # X_tpm_dense = X_original.X.todense() # X_tpm_dense[0:10,].sum(axis=1) ## check TPM normalization # X = X_norm.X.todense() ## 221203 X = X_norm.X H_path = cnmf_obj.paths['consensus_spectra__txt'] % (selected_K, '0_2') ## median_spectra_file H_df = pd.read_csv(H_path, sep='\t', index_col=0).T H = H_df.to_numpy() H = (H/H.sum(0)) W_path = cnmf_obj.paths['consensus_usages__txt'] % (selected_K, '0_2') ## median_spectra_file W_df = pd.read_csv(W_path, sep='\t', index_col=0) W = W_df.to_numpy() WH = W @ H.T # diff = X - WH # X_col_sd = np.std(X_tpm_dense, axis=0, ddof=1) ## column normalization, shape: (1,17472) # type(gep_tpm) ## data frame # gep_tpm.shape ## (17472, 60) # H_tmp = gep_tpm.to_numpy().T ## shape: (60, 17472) # H = gep_tpm.to_numpy().T / X_col_sd ## normalize TPM spectra matrix (gene x component) # X = X_tpm_dense / X_col_sd # X[:,0:10].std(axis=0) ## check if standard deviation of gene expression is 1 # W = usage_norm.to_numpy() # W[0:10,].sum(axis=1) ## to check if sum of each cell's usage is 1 # WH = W @ H diff = X - WH diff_sumOfSquaresError = (np.asarray(diff)**2).sum() # X_sumOfSquares = (np.asarray(X)**2).sum() # WH_sumOfSquares = (np.asarray(WH)**2).sum() Var_diff = compute_Var(diff) Var_X = compute_Var(X) TotalVarianceExplained = 1 - Var_diff / Var_X def computeVarianceExplained(X, H, Var_X, i): if not isinstance(H, (pd.DataFrame)): B_k = X @ H[i,:].T / np.sqrt((np.asarray(H[i,:])**2).sum()) numerator = compute_Var(X - np.outer(B_k, H[i,:])) else: B_k = X @ H.iloc[i,:] / np.sqrt((H.iloc[i,:]**2).sum()) numerator = compute_Var(X - np.outer(B_k, H.iloc[i,:])) return (1 - numerator / Var_X) ## initialize storage variable V_k = np.empty([selected_K]) for i in range(selected_K): print(i) V_k[i] = computeVarianceExplained(X, H.T, Var_X, i) ProgramID = ['K' + str(selected_K) + '_' + str(i+1) for i in range(selected_K)] metrics_df = pd.DataFrame({'VarianceExplained': V_k, 'ProgramID': ProgramID }) metrics_summary = pd.DataFrame({'Sum' : metrics_df['VarianceExplained'].sum(), 'Median' : metrics_df['VarianceExplained'].median(), 'Max' : metrics_df['VarianceExplained'].max(), 'Total' : TotalVarianceExplained}, index = [0]) metrics_df.to_csv(os.path.join(OUTDIR, "metrics.varianceExplained.df.txt"), index = None, sep="\t") metrics_summary.to_csv(os.path.join(OUTDIR, "summary.varianceExplained.df.txt"), index = None, sep="\t") |
Support
- Future updates
Related Workflows





