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, operation
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 .
This workflow attempts to find clusters of similar sequences in a set of nanopore reads. It is intended to be used for samples that have been size selected for viral-like particles.
Overview
Clustering 1: windowed minimap2
The first pass uses minimap2 and mcl to find clusters of similar sequences.
Because an all-v-all comparison is system-taxing and we're looking for clusters of complete sequences of the same length, we split the reads into buckets by size. We use overlapping size windows so that we can combine the distance measures and run MCL on all the reads at once.
Clustering 2: lastal
Clusters with enough reads are processed with lastal and mcl to find subclusters. A cutoff of 85% identity is used to build the MCL distance matrix
Polishing
Subclusters with enough reads and a tight size distribution are polished with racon and medaka to produce consensus sequences.
Installation
The workflow is executed via snakemake, which can handle the installation of all required dependencies (with the help of conda).
NOTE: The commands in this section (Installation) are assumed to be run from the repo directory.
Conda
The simplest approach is to install conda if you don't have it, and then to use conda to install snakemake.
Mamba
Mamba is an add on for conda that can isntall programs much faster than conda. We recommend installing that, too, but you don't have to. If you do have mamba install, replace "conda" with "mamba" in the following command.
The snakemake env
Create a conda environment for running snakemake by using the provided configuration file:
$ conda env create -p ./env -f np_read_clustering/conda/snake.yaml
To use snakemake, you'll have to activate the environment in the shell (or script) from which you want to launch the workflow:
$ conda activate ./env
NOTE: The
-p ./env
option creates the environment in a folder named
env
in your current directory. You can use any name and location you wish. You can also use
-n
to name the environment and keep it in you conda installation location. See the conda documentation for how to name and activate environments for more detail.
A Test Run
That's it. Now you are ready to go. Test your setup (and pre-install the rest of the dependencies):
$ snakemake --configfile=config.yaml -j 2 -p --use-conda --conda-frontend mamba
Note: this assumes you are running from the repo directory.
You can also run the larger test file by overriding key config values:
$ snakemake --configfile=config.yaml -j 20 -p --use-conda --conda-frontend mamba \
--config all_fasta=test/test.fasta work_dir=test/outputs/nprc
Note: we increased the threasd count from 2 to 20, because this is a bigger dataset.
Running
Configuration
All of the configuration parameters are top level, so they can be supplied on the command line, but can be passed in by file as in the test example above.
Required Parameters
The only stricltly necessary input is:
- all_fasta: a fasta file of all the nanopore reads. This is assumed to be in {work_dir}/all.reads.fasta.
You may also want to specify:
-
work_dir: location to create all files (defaults to 'np_clustering'). This can be outside the repo.
-
name: naming prefix for the final sequences (defaults to 'SEQ')
-
pfam_hmm_path: the PFAM HMM file for gene annotation (pfam annotations are empty otherwise)
Other Parameters
See the example config.yaml for the rest of the parameters and their defaults
See the provided example and the snakeamke documentation for full information on snakemake configuration by file and command line, but the basic ideas are:
Command time config
Cofiguration options can be supplied by the command line with the --config option:
$ snakemake -s path/to/Snakefile --config name=my_name work_dir=my_Dir ...
Configfiles
Config value can be supplied by files. These can be formatted as JSON or YAML. Tell snakemake where to find the file with
--configfile=
$ snakemake -s /path/to/np_read_clustering/Snakefile --configfile=my.config.yaml
We suggest you copy the provided example into a new file and modify the files accordingly.
parallelization and performance
Single node
The
-j
flag tells snakemake how many threads are availabe on your computer, and it will run workflow steps in parallel as much as is possible (usually).
multithreaded steps
Some steps in the workflow are mutithreaded (EG: minimap2). You can configure how many threads these teps get in the configuration (EG: mapping_threads).
examples
Note, we don't obother to use the --conda-frontend flag here assuming the conda environments have already been created during the test run above. Mamba is only needed when creaating environments.
Running with a custom config:
$ snakemake --configfile=my.config.yaml -j 40 -p --use-conda
Or use the provided test config an override key values:
$ snakemake --configfile=config.yaml -j 40 -p --use-conda \
--config all_fasta=/path/to/reads.fasta work_dir=/path/to/output hmm_path=/dbs/PFAM.hmm
Code Snippets
1 2 3 4 5 6 7 8 9 10 11 12 13 | import os from hit_tables import agg_hit_table hit_table = str(snakemake.input) fmt = str(snakemake.params.format).upper() if os.path.getsize(hit_table) > 0: agg_hit_table(hit_table, format=fmt) \ .to_csv(str(snakemake.output), sep='\t', index=None) else: # input is empty, touch the output with open(str(snakemake.output), 'wt') as out_handle: pass |
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 | import pandas, numpy, os from collections import deque from itertools import cycle from scipy import stats from Bio import SeqIO # load the read lengths from the summary file read_lens = pandas.read_csv(snakemake.input.read_lens, sep='\t', names=['read_id','sequence_length_template'], index_col='read_id', header=None).sequence_length_template.to_dict() # process clusters to choose keepers cluster_data = [] read_clusters = {} sigma_cutoff = snakemake.params.sigma_cutoff count_cutoff = snakemake.params.min_cl_size # loop over clusters in mcl_file with open(str(snakemake.input.mcl)) as mcl_lines: for i,line in enumerate(mcl_lines): # get cluster read names reads = set(line.strip().split()) count = len(reads) # get cluster read length dist cluster_lens = numpy.array([read_lens[r] for r in reads]) counts, bins = numpy.histogram(cluster_lens, bins=100) X = numpy.array([numpy.mean((bins[j], bins[j-1])) for j in range(1,len(bins))]) mu, sigma = stats.norm.fit(cluster_lens) keep = (sigma <= sigma_cutoff and count >= count_cutoff) cluster_data.append(dict(num=i, count=count, sigma=sigma, mu=mu, keep=keep)) if keep: """ # write read list if not os.path.exists(str(snakemake.output.reads)): os.makedirs(str(snakemake.output.reads), exist_ok=True) with open(f"{output.reads}/cluster.{i}.reads", 'wt') as reads_out: reads_out.write("\n".join(reads) + "\n") """ # save cluster id for read in reads: read_clusters[read] = i cluster_table = pandas.DataFrame(cluster_data) ## assign groups # this serves 2 purposes: # 1) we limit the number of files in each folder (too many files can slow # down snakemake) # 2) we enable running the workflow in chunks (can perform better in some # cases) keepers = cluster_table.query('keep') num_keepers = keepers.shape[0] # we want the number of groups, but we can get it from group_size if 'group_size' in snakemake.config and 'num_groups' not in snakemake.config: group_size = snakemake.config['group_size'] n_groups = int(numpy.ceil(num_keepers/group_size)) else: n_groups = snakemake.config.get('num_groups', 100) # assigne a group to each cluster (round-robin) groups = cycle(range(n_groups)) cluster_groups = {c:next(groups) for c in keepers['num']} cluster_table['group'] = [cluster_groups.get(c,None) if k else None for c,k in cluster_table[['num','keep']].values] # write fasta files if not os.path.exists(str(snakemake.output.reads)): os.makedirs(str(snakemake.output.reads), exist_ok=True) # limit number of open files with n_open = 250 open_handle_ids = deque([]) handles = {} def open_cluster_fasta(i): """ checks for open handle for this scluster and returns it if found otherwise closes oldest handle and replaes with new handle for this cluster """ # return open handle if it exists try: return handles[i] except KeyError: pass # close handle(s) if we have too many while len(handles) > n_open - 1: # drop oldest drop_id = open_handle_ids.popleft() # close and delete handles[drop_id].close() del handles[drop_id] group = cluster_groups[i] fasta_file = f"{snakemake.output.reads}/group.{group}/cluster.{i}.fasta" fd = os.path.dirname(fasta_file) if not os.path.exists(fd): os.makedirs(fd) handle = open(fasta_file, 'at') handles[i] = handle open_handle_ids.append(i) return handle # loop over all reads and write out skipped_read_count = 0 for read in SeqIO.parse(snakemake.input.fasta, 'fasta'): try: cluster = read_clusters[read.id] except KeyError: # skip if no cluster skipped_read_count += 1 continue open_cluster_fasta(cluster).write(read.format('fasta')) # add row for unclustered reads for k,v in dict(i=-1, count=skipped_read_count, keep=False).items(): cluster_table.loc[-1,k] = v # save cluster table cluster_table.to_csv(str(snakemake.output.stats), sep='\t', index=False) |
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 | import numpy import os import pandas import re from collections import Counter, defaultdict from functools import partial from itertools import zip_longest from scipy import stats import matplotlib matplotlib.use('pdf') from matplotlib import pyplot as plt, cm, colors from matplotlib.patches import Polygon from matplotlib.backends.backend_pdf import PdfPages from snakemake.rules import Namedlist from Bio import SeqIO from hit_tables import parse_blast_m8, BLAST_PLUS from edl import blastm8 BLACK = (0, 0, 0, 1) def grouper_trim(iterable, n): "Collect data into fixed-length chunks or blocks and trim last chunk (and all null values)" # grouper('ABCDEFG', 3, 'x') --> ABC DEF Gxx" args = [iter(iterable)] * n return ([i for i in group if i is not None] for group in zip_longest(*args, fillvalue=None)) def main(input, output, params): """ input output should be namedlists (from snakemake) params should be a dict (so we can fall back to defaults) """ # prodigal amino acid output faa_file = input.faa # cluster fasta reads fasta_file = input.fasta # PFAM results (non-overlapping) dom_tbl_U = input.domtbl # mcl results mcl_file = input.mcl # lastal table raw lastal_file = input.lastal # lastal table aggregated agg_file = input.agg ## load clusters (just a list of reads in each cluster, sorted by size) subclusters = load_clusters(mcl_file, params.get('min_sub_size', 10)) ## load the fasta, keeping dict of lengths cluster_reads = {r.id:r for r in SeqIO.parse(fasta_file, 'fasta')} # use just the first word in the read id as a short name read_lens = {r.id.split('-')[0]:len(r) for r in cluster_reads.values()} ## plot all cluster hitsts, applying sigma cutoff subcluster_ids = plot_cluster_hists(subclusters, read_lens, agg_file, output.stats, output.hist_pdf, params) ## make the synteny plots for each good subcluster plot_subcluster_synteny(subcluster_ids, subclusters, read_lens, lastal_file, faa_file, dom_tbl_U, output.gene_pdf, params) ## write out sub cluster fasta os.makedirs(str(output.fasta_dir), exist_ok=True) for subcluster_id in subcluster_ids: with open(str(output.fasta_dir) + f"/subcluster.{subcluster_id}.fasta", 'wt') as fasta_out: for read_id in subclusters[subcluster_id]: fasta_out.write(cluster_reads[read_id].format('fasta')) def plot_cluster_hists(subclusters, read_lens, agg_file, stats_file, pdf_file, params ): """ For each subcluster plot: * histogram aod all read-read mfracs * histogram of all read lengths with overlaid normal dist """ # open PDF file pdf = PdfPages(pdf_file) mx_len = max(read_lens.values()) mn_len = min(read_lens.values()) window = [mn_len, mx_len] sigma_cutoff = params.get('sigma_cutoff', -1) # first pass to chose subclusters to keep and plot cluster_stats = {} for i, subcluster in enumerate(subclusters): keep = True if len(subcluster) < params.get('min_sub_size', 10): break # calculate best normal fit to length dist cluster_lens = numpy.array([read_lens[r.split('-')[0]] for r in subcluster]) counts, bins = numpy.histogram(cluster_lens, bins=100, range=window) #from scipy import stats mu, sigma = stats.norm.fit(cluster_lens) if sigma_cutoff > 0 and sigma > sigma_cutoff: keep = False # calculate the stats X = numpy.array([numpy.mean((bins[i], bins[i-1])) for i in range(1,len(bins))]) tot_in, tot_out, n_in, n_out = numpy.zeros(4) for x, count in zip(X, counts): if x < mu - sigma or x > mu + sigma: tot_out += count n_out += 1 else: tot_in += count n_in += 1 mean_in = tot_in / n_in mean_out = tot_out / n_out if n_out > 0 else 0 ratio = mean_in / mean_out n_ratio = n_in / (n_out + n_in) cluster_stats[i] = dict(zip( ['mu', 'sigma', 'ratio', 'n_ratio', 'N', 'keep', 'counts', 'bins', 'X'], [mu, sigma, ratio, n_ratio, len(subcluster), keep, counts, bins, X] )) # build cluster stats table write_cols = ['mu', 'sigma', 'ratio', 'n_ratio', 'N', 'keep'] cl_st_table = pandas.DataFrame([[i,] + [d[k] for k in write_cols] for i,d in cluster_stats.items()], columns=['index'] + write_cols) # write stats to file cl_st_table.to_csv(stats_file, sep='\t', index=None) # pull out list of good subclusters subcluster_ids = list(cl_st_table.query('keep').index) # load agg hits agg_table = pandas.read_csv(agg_file, sep='\t') # max 8 per page mx_rows = 8 for page_sc_ids in grouper_trim(cluster_stats.keys(), mx_rows): N = len(page_sc_ids) fig, axes = plt.subplots(N, 4, figsize=[11 * N / mx_rows, 8.5], sharey="col", sharex="col", squeeze=False) fig.subplots_adjust(hspace=.7, wspace=.6) ax_rows = iter(axes) for i, subcluster_id in enumerate(page_sc_ids): axs = next(ax_rows) # remove axes from top and right for ax in axs: for side in ['top', 'right']: ax.spines[side].set_visible(False) ax_sc_mf, ax_sc_id, ax_h_mf, ax_h_ln = axs # get the subset of the agg table for this subcluster subcluster = set(subclusters[subcluster_id]) sub_slice = (agg_table['query'].apply(lambda q: q in subcluster) & agg_table.hit.apply(lambda h: h in subcluster)) agg_hits_cluster = agg_table[sub_slice] \ .eval('mean_len = (hlen + qlen) / 2') \ .eval('frac = mlen / mean_len') mfrac_dict = agg_hits_cluster.set_index(['query','hit']).mfrac.to_dict() # scatter plot mfrac and mean length ax_sc_mf.scatter(agg_hits_cluster.mfrac.values, agg_hits_cluster.mean_len.values, marker='.', alpha=.5 ) ax_sc_mf.set_ylabel ('mean_len') # scatter plot of pctid and matched fraction ax_sc_id.scatter(agg_hits_cluster.pctid.values, agg_hits_cluster.frac.values, marker='.', alpha=.5 ) ax_sc_id.set_ylabel ('frac aln') # plot hist of pairwise mfracs h = ax_h_mf.hist(get_mfracs(subcluster, mfrac_dict=mfrac_dict), bins=100, range=[50,100]) # plot hist of read lens sc_stats = cluster_stats[subcluster_id] counts = sc_stats['counts'] X = sc_stats['X'] # recreate histogram from counts and X ax_h_ln.bar(X, counts, color='blue') # overlay norm dist best_fit_line = stats.norm.pdf(X, sc_stats['mu'], sc_stats['sigma']) best_fit_line = best_fit_line * counts.sum() / best_fit_line.sum() p = ax_h_ln.plot(X, best_fit_line, color='red', alpha=.5) ax_h_mf.set_ylabel(f"s.cl: {subcluster_id}") ax_h_ln.set_ylabel(f"{len(subcluster)} {int(sc_stats['sigma'])}") if i == N - 1: xl = ax_sc_mf.set_xlabel("score") xl = ax_h_ln.set_xlabel("length") xl = ax_sc_id.set_xlabel ('match %ID') xl = ax_h_mf.set_xlabel ('score') # close plot and go to next pdf page pdf.savefig(bbox_inches='tight') plt.close() pdf.close() # save stats to file, but drop extra data first write_cols = ['mu', 'sigma', 'ratio', 'n_ratio', 'N'] pandas.DataFrame([[i,] + [d[k] for k in write_cols] for i,d in cluster_stats.items()], columns=['index'] + write_cols).to_csv(stats_file, sep='\t', index=None) return subcluster_ids def get_N_colors(N, cmap_name='Dark2'): """ given N and a colormap, get N evenly spaced colors""" color_map=plt.get_cmap(cmap_name) return [color_map(c) for c in numpy.linspace(0, 1, N)] def get_scaled_color(value, minv=0, maxv=1, alpha=.75, reverse=False, cmap_name='cool'): colormap = plt.get_cmap(cmap_name) if reverse: maxv, minv = minv, maxv rangev = maxv - minv color = colormap((value - minv) / rangev) return color[:3] + (alpha,) def get_mfracs(reads, mfrac_dict): return [mfrac_dict.get((r1, r2), 0) for r1 in reads for r2 in reads if r2 > r1 ] def plot_subcluster_synteny(subcluster_ids, subclusters, read_lens, lastal_file, faa_file, dom_tbl_U, pdf_file, params ): """ For each subcluster: * identify the N genes that appear in the most reads * identify the M reads that have the most of the top genes * plot """ ## load the gene annotations # first get positions from faa headers read_genes = {} for gene in SeqIO.parse(faa_file, 'fasta'): gene_id, start, end, strand, _ = [b.strip() for b in gene.description.split("#")] read, name, gene_no = re.search(r'^((\w+)-[^_]+)_(\d+)', gene_id).groups() read_genes.setdefault(name, []).append(dict( gene_id=gene_id, start=int(start), end=int(end), strand=int(strand), num=int(gene_no), pfam=None, )) # convert to dict of DataFrames from dict of lists of dicts read_genes_tables = {read:pandas.DataFrame(genes).set_index('gene_id') for read, genes in read_genes.items()} # and add PFAM annotations for read, hits in blastm8.generate_hits(dom_tbl_U, format='hmmsearchdom'): read_id = read.split("-")[0] read_genes_table = read_genes_tables[read_id] for hit in hits: gene_id = hit.read # only assign PFAm if it's the first hit for the gene if pandas.isna(read_genes_table.loc[gene_id, 'pfam']): pfam = hit.hit read_genes_table.loc[gene_id, 'pfam'] = pfam # load all the read to read hits read_hits = parse_blast_m8(lastal_file, format=BLAST_PLUS) # now open the PDF file pdf = PdfPages(pdf_file) # for each good subcluster for subcluster_id in subcluster_ids: subcluster = set(subclusters[subcluster_id]) subcluster_names = {r.split('-')[0]:r for r in subcluster} fig = plot_subcluster_genes(subcluster_id, subcluster_names, read_genes_tables, read_hits, read_lens, params) # close plot and go to next pdf page pdf.savefig(bbox_inches='tight') plt.close() pdf.close() def plot_subcluster_genes(subcluster_id, subcluster_names, read_genes_tables, read_hits, read_lens, params): """ make a plot of gene positions: ax1 has a scatter plot of mean position by pfam ax2 has aligned genomes with top pfams colored """ # get the positions of the named PFAMs pf_positions = defaultdict(list) for read, gene_table in read_genes_tables.items(): if read in subcluster_names: # do we want to flip the read dir? (too many strand < 1) reverse = gene_table.eval('glen = strand * (end - start)').glen.sum() < 1 for start, end, pfam in gene_table[['start','end','pfam']].values: if pandas.isna(pfam): continue if reverse: start, end = [read_lens[read] - p for p in (start, end)] # add mean post to list for this pfam pf_positions[pfam].append((end + start) / 2) # chose which genes to color N = params.get('max_colored_genes', 8) sorted_genes = sorted(pf_positions.keys(), key=lambda k: len(pf_positions[k]), reverse=True) top_N_pfams = sorted_genes[:N] gene_color_dict = dict(zip(top_N_pfams, get_N_colors(N, cmap_name=params.get('gene_cmap', 'Dark2')))) # chose which reads to draw M = params.get('max_synteny_reads', 20) def count_top_pfams_in_read(read): if read in read_genes_tables: return sum(1 for p in read_genes_tables[read].pfam.values if p in top_N_pfams) return 0 top_M_reads = sorted(subcluster_names, key=count_top_pfams_in_read, reverse=True, )[:M] m = len(top_M_reads) # calculate the sizes necessary to draw genes using the matplotlib arrow function align_height = (7 * (m-1) / (M-1)) #use up to 7 in figsize = [8.5, 4 + align_height] fig, axes = plt.subplots(2,1, figsize=figsize, gridspec_kw={'height_ratios':[4,align_height]}, sharex='col') fig.subplots_adjust(hspace=.1,) ## draw gene positions ax = axes[0] ax.set_title(f'PFAM annotations in subcluster {subcluster_id}') n = params.get('max_plotted_genes', 18) sorted_pf = sorted([p for p in sorted_genes[:n] if len(pf_positions[p]) > 1], key=lambda p: numpy.mean(list(pf_positions[p]))) for i, p in enumerate(sorted_pf): x,y = zip(*((gp,i) for gp in pf_positions[p])) ax.scatter(x,y, c=len(y) * [gene_color_dict.get(p, BLACK)], ec=None, alpha=.5) yt = ax.set_yticks(range(len(sorted_pf))) ytl = ax.set_yticklabels(sorted_pf) for label in ytl: label.set_color(gene_color_dict.get(label.get_text(), BLACK)) ## draw alignments ax = axes[-1] min_x = 0 max_x = max(read_lens[r] for r in subcluster_names) range_x = max_x - min_x range_y = M thickness = .5 head_length = range_x * (thickness / range_y) * (figsize[1] / figsize[0]) cmap = params.get('read_cmap','cool') min_pctid = read_hits.pctid.min() pctid_range = 100 - min_pctid get_conn_color = partial(get_scaled_color, minv=min_pctid, maxv=100, alpha=.75, cmap_name=cmap) y = 0 pad = .1 prev_read = None for name in top_M_reads: read = subcluster_names[name] read_length = read_lens[name] if name in read_genes_tables: gene_table = read_genes_tables[name] # do we want to flip the read dir? (too many strand < 1) reverse = gene_table.eval('glen = strand * (end - start)').glen.sum() < 1 # draw genes for start, end, strand, pfam in gene_table[['start','end','strand','pfam']].values: if reverse: strand = -1 * strand start = read_length - start end = read_length - end strand = int(strand) hl = min(head_length, end-start) al = max((end - start) - hl, .0001) * strand ast = start if al > 0 else end color = gene_color_dict.get(pfam, 'k') plt.arrow(ast, y, al, 0, fc=color, ec=color, lw=0, width=thickness, head_width=thickness, head_length=hl, head_starts_at_zero=(int(strand) > 0)) else: reverse=False # connect matched segments for read pairs if prev_read is not None: # get hits between reads pair_hits = read_hits.query(f'(hit == "{read}" and query == "{prev_read}") or ' f'(query == "{read}" and hit == "{prev_read}")') \ .query('hit != query') \ .sort_values('score', ascending=True) # loop over hits cols = ['query', 'hit', 'qstart', 'qend', 'hstart', 'hend', 'pctid'] for query, hit, qstart, qend, hstart, hend, pctid in pair_hits[cols].values: # if hit was recorded the other way, flip hit/query if query == prev_read: qstart, qend, hstart, hend = hstart, hend, qstart, qend # if either read is reversed, flip x coordinates if reverse: qstart = read_length - qstart qend = read_length - qend if prev_rev: hstart = prev_len - hstart hend = prev_len - hend # draw connecting paralellogram color = get_conn_color(pctid, alpha=.9) xy = numpy.array([(hstart, y-1+pad), (qstart, y-pad), (qend, y-pad), (hend, y-1+pad)]) ax.add_patch(Polygon(xy, fc=(.6,.6,.6,.2), ec=color)) # save read info for next one prev_read = read prev_rev = reverse prev_len = read_length # increment y value y += 1 x = plt.xlim(min_x - 50, max_x + 50) y = plt.ylim(-.5, y - .5) plt.yticks(list(range(m)), top_M_reads) plt.xlabel('read position') cax = plt.axes([0.95, 0.15, 0.025, 0.4 * (align_height / 7)]) plt.colorbar(mappable=cm.ScalarMappable(norm=colors.Normalize(min_pctid, 100), cmap=cmap), cax=cax) cl = cax.set_ylabel('alignment %ID') return fig def load_clusters(mcl_file, size_cutoff=10): with open(mcl_file) as mcl_lines: return [c for c in [line.strip().split() for line in mcl_lines] if len(c) >= size_cutoff] # scriptify if __name__ == "__main__": try: # assume this is called from snakemake input = snakemake.input output = snakemake.output params = dict(snakemake.params.items()) except NameError: # TODO: fallback to argparse if we call from the command line (for testing) import argparse raise Exception("Currently only works from snakemake, sorry") main(input, output, params) |
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 | import os, re, pandas from Bio import SeqIO def main(input, output, params): stats = {} files = {} groups = {} logger.debug("Collecting polished sequences") # get fasta files and gene stats for sc_comp_file in input: sc_dir = os.path.dirname(sc_comp_file) group, cluster, subcluster = \ re.search(r'group.(\d+).+cluster\.(\d+).+subcluster\.(\d+)', sc_dir).groups() cl_sc_id = f"{cluster}_{subcluster}" files[cl_sc_id] = dict( fasta=sc_dir + "/medaka.fasta", faa =sc_dir + "/medaka.faa" ) gene_stats = \ pandas.read_csv(sc_dir + "/medaka.v.drafts.gene.lengths", sep='\t', index_col=0) \ .loc['medaka'] stats.setdefault(cluster, {})[subcluster] = \ {f"gene_{k}": v for k,v in gene_stats.items()} groups[cluster] = group logger.debug("Found {len(files)} poilished seqs from {len(stats)} clusters") # get read len stats for subclusters for cluster in stats: sc_tsv = (f"{params.work_dir}/refine_lastal/group.{groups[cluster]}" f"/cluster.{cluster}/subclusters/cluster_stats.tsv") sc_stats = pandas.read_csv(sc_tsv, sep='\t', index_col=0) for sc, row in sc_stats.iterrows(): sc = str(sc) if sc not in stats[cluster]: continue stats[cluster][sc].update(dict( read_len_mean=row['mu'], read_len_dev=row['sigma'], read_count=row['N'])) # convert stats to table df = pandas.DataFrame({f"{cl}_{sc}": sc_stats for cl, cl_stats in stats.items() for sc, sc_stats in cl_stats.items()},) \ .T # polsihed fasta with open(str(output.fasta), 'wt') as fasta_out: for cl_sc_id in files: for read in SeqIO.parse(files[cl_sc_id]['fasta'], 'fasta'): np_read = read.id read.id = f"{params.name}_{cl_sc_id}" df.loc[cl_sc_id, 'length'] = len(read) N = df.loc[cl_sc_id, 'read_count'] read.description = f"n_reads={N};rep_read={np_read}" fasta_out.write(read.format('fasta')) # write out stats table df.to_csv(str(output.stats), sep='\t') with open(str(output.faa), 'wt') as faa_out: for cl_sc_id in files: N = 0 for gene in SeqIO.parse(files[cl_sc_id]['faa'], 'fasta'): N += 1 read.id = f"{params.name}_{cl_sc_id}_{N}" faa_out.write(read.format('fasta')) if __name__ == "__main__": main(snakemake.input, snakemake.output, snakemake.params) |
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 | import pandas, numpy def main(input, output, params): with open(str(output.report), 'wt') as output_handle: cluster_stats = pandas.read_csv(str(input.mcl_stats), sep='\t', index_col=0) n_reads = cluster_stats['count'].sum() n_clusters = cluster_stats.shape[0] n_gt_size_cutoff = cluster_stats.query(f'count >= {params.min_cl_size}').shape[0] n_kept = cluster_stats.query('keep').shape[0] output_handle.write(f"Cluster Search Results:\n" f" minimap2 clusters:\n" f" reads: {n_reads}\n" f" clusters: {n_clusters}\n" f" gt_{params.min_cl_size}: {n_gt_size_cutoff}\n" f" kept_clusters: {n_kept}\n\n") # count raw subclusters in mcl files n_scs, n_sc_gt_cutoff = 0, 0 for sc_mcl_file in input.sc_mcls: with open(sc_mcl_file) as mcl_lines: for line in mcl_lines: n_scs += 1 if len(line.strip().split()) > params.min_cl_size: n_sc_gt_cutoff += 1 # get stats from polished subclusters pol_stats = pandas.read_csv(str(input.pol_stats), sep='\t', index_col=0) pol_lens = pol_stats[pol_stats['length'].notna()]['length'].values n_sc_kept = len(pol_lens) output_handle.write(f" lastal subclusters:\n" f" subclusters: {n_scs}\n" f" gt_{params.min_cl_size}: {n_sc_gt_cutoff}\n" f" kept_subclusters: {n_sc_kept}\n\n") output_handle.write(f" polished seqs:\n" f" count: {len(pol_lens)}\n" f" mean: {pol_lens.mean()}\n" f" max: {pol_lens.max()}\n" f" min: {pol_lens.min()}\n" f" median: {numpy.median(pol_lens)}\n" f" stddev: {pol_lens.std()}\n\n") if __name__ == "__main__": main(snakemake.input, snakemake.output, snakemake.params) |
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 | import os from hit_tables import agg_hit_table hit_table = str(snakemake.input) fmt = str(snakemake.params.format).upper() # identify sequences that hit longer sequences at > 95% subseqs = set() if os.path.getsize(hit_table) > 0: aggs = agg_hit_table(hit_table, format=fmt) count = 0 values = aggs.query('hit != query')[['query', 'hit', 'matches', 'qlen', 'hlen']].values for query, hit, matches, qlen, hlen in values: if hlen < qlen: query, hit, qlen, hlen = hit, query, hlen, qlen mfracq = matches / qlen if mfracq > .95: if query.split('_')[1] != hit.split('_')[1]: subseqs.add(query) count += 1 # write fragment IDs to list with open(str(snakemake.output), 'wt') as out_handle: for seq_id in subseqs: out_handle.write(f"{seq_id}\n") |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | import pandas from hit_tables import parse_blast_m8, PAF from Bio import SeqIO # pick a best read hits = parse_blast_m8(str(snakemake.input.paf),format=PAF) hit_matches = hits.groupby(['hit','query']).agg({'matches':sum}) mean_matches = {r:hit_matches.query(f'hit != query and (hit == "{r}" or query == "{r}")').matches.mean() for r in set(i[0] for i in hit_matches.index).union(i[1] for i in hit_matches.index)} best_matches = sorted(mean_matches.keys(), key=lambda r: mean_matches[r], reverse=True) ref_read = best_matches[0] # write out to 2 files with open(str(snakemake.output.ref), 'wt') as ref_out: with open(str(snakemake.output.others), 'wt') as others_out: for read in SeqIO.parse(str(snakemake.input.fasta), 'fasta'): if read.id == ref_read: ref_out.write(read.format('fasta')) else: others_out.write(read.format('fasta')) |
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 | import pandas, numpy, os from scipy import stats import matplotlib matplotlib.use('pdf') from matplotlib import pyplot as plt from matplotlib.backends.backend_pdf import PdfPages sigma_cutoff = snakemake.params.sigma_cutoff count_cutoff = snakemake.params.min_cl_size # load the read lengths from the summary file read_lens = pandas.read_csv(snakemake.input.read_lens, sep='\t', names=['read_id','sequence_length_template'], index_col='read_id', header=None).sequence_length_template.to_dict() # load the all.v.all mfrac values (but just map pairs where q > h) mfrac_dict = {tuple(sorted(i)):m for i,m in pandas.read_csv(snakemake.input.abc, sep='\t', names=['q', 'h', 'mfrac'], header=None, index_col=['q','h']) \ .mfrac.items()} def get_mfracs(reads, mfrac_dict=mfrac_dict): return [mfrac_dict.get((r1, r2), 0) for r1 in reads for r2 in reads if r2 > r1 ] # load the clusters with open(snakemake.input.mcl) as mcl_lines: all_clusters = [line.strip().split() for line in mcl_lines] # plots pdf=PdfPages(snakemake.output.pdf) ROWS = 20 COLS = 5 N0=0 while True: clusters = all_clusters[N0:N0+ROWS*COLS] if len(clusters) == 0: break rows = int(numpy.ceil(len(clusters)/COLS)) fig, axes = plt.subplots(rows, COLS*2, figsize=[COLS*4,rows], sharex=False, squeeze=False) fig.subplots_adjust(hspace=.7, wspace=.6) cluster_iter = enumerate(clusters, start=N0) axes_list = axes.flatten() for i, cluster in enumerate(clusters): j = i*2 ax1, ax2 = axes_list[j:j+2] # plot hist of pairwise mfracs h = ax1.hist(get_mfracs(cluster, mfrac_dict=mfrac_dict), bins=100, range=[0,100]) # plot hist of read lens cluster_lens = numpy.array([read_lens[r] for r in cluster]) counts, bins, h_line = ax2.hist(cluster_lens, bins=100, histtype='step') X = numpy.array([numpy.mean((bins[j], bins[j-1])) for j in range(1,len(bins))]) mu, sigma = stats.norm.fit(cluster_lens) # overlay norm dist best_fit_line = stats.norm.pdf(X, mu, sigma) best_fit_line = best_fit_line * counts.sum() / best_fit_line.sum() p = ax2.plot(X, best_fit_line, color='red', alpha=.5) keep = (sigma <= sigma_cutoff and len(cluster) >= count_cutoff) if keep: ax1.set_ylabel('keep') ax2.set_ylabel(f"c{i} n={len(cluster)}") # only put xlabels on bottom plots if i >= len(clusters) - COLS: xl = ax1.set_xlabel("score") xl = ax2.set_xlabel("length") # remove axes from top and right for ax in [ax1, ax2]: for side in ['top', 'right']: ax.spines[side].set_visible(False) # hide unused axes (on last page) for i in range(j+2, len(axes_list)): fig.delaxes(axes_list[i]) pdf.savefig(bbox_inches='tight') plt.close() N0 += ROWS * COLS if len(cluster) < count_cutoff: break pdf.close() |
213 | shell: 'rm {input}' |
Support
- Future updates
Related Workflows





