Mercurial > repos > davidvanzessen > shm_csr
comparison change_o/DefineClones.py @ 0:c33d93683a09 draft
Uploaded
author | davidvanzessen |
---|---|
date | Thu, 13 Oct 2016 10:52:24 -0400 |
parents | |
children | 22dddabe3637 |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 0:c33d93683a09 |
---|---|
1 #!/usr/bin/env python3 | |
2 """ | |
3 Assign Ig sequences into clones | |
4 """ | |
5 # Info | |
6 __author__ = 'Namita Gupta, Jason Anthony Vander Heiden, Gur Yaari, Mohamed Uduman' | |
7 from changeo import __version__, __date__ | |
8 | |
9 # Imports | |
10 import os | |
11 import re | |
12 import sys | |
13 import numpy as np | |
14 from argparse import ArgumentParser | |
15 from collections import OrderedDict | |
16 from itertools import chain | |
17 from textwrap import dedent | |
18 from time import time | |
19 from Bio import pairwise2 | |
20 from Bio.Seq import translate | |
21 | |
22 # Presto and changeo imports | |
23 from presto.Defaults import default_out_args | |
24 from presto.IO import getFileType, getOutputHandle, printLog, printProgress | |
25 from presto.Multiprocessing import manageProcesses | |
26 from presto.Sequence import getDNAScoreDict | |
27 from changeo.Commandline import CommonHelpFormatter, getCommonArgParser, parseCommonArgs | |
28 from changeo.Distance import getDNADistMatrix, getAADistMatrix, \ | |
29 hs1f_model, m1n_model, hs5f_model, \ | |
30 calcDistances, formClusters | |
31 from changeo.IO import getDbWriter, readDbFile, countDbFile | |
32 from changeo.Multiprocessing import DbData, DbResult | |
33 | |
34 # Defaults | |
35 default_translate = False | |
36 default_distance = 0.0 | |
37 default_bygroup_model = 'hs1f' | |
38 default_hclust_model = 'chen2010' | |
39 default_seq_field = 'JUNCTION' | |
40 default_norm = 'len' | |
41 default_sym = 'avg' | |
42 default_linkage = 'single' | |
43 | |
44 # TODO: should be in Distance, but need to be after function definitions | |
45 # Amino acid Hamming distance | |
46 aa_model = getAADistMatrix(mask_dist=1, gap_dist=0) | |
47 | |
48 # DNA Hamming distance | |
49 ham_model = getDNADistMatrix(mask_dist=0, gap_dist=0) | |
50 | |
51 | |
52 # TODO: this function is an abstraction to facilitate later cleanup | |
53 def getModelMatrix(model): | |
54 """ | |
55 Simple wrapper to get distance matrix from model name | |
56 | |
57 Arguments: | |
58 model = model name | |
59 | |
60 Return: | |
61 a pandas.DataFrame containing the character distance matrix | |
62 """ | |
63 if model == 'aa': | |
64 return(aa_model) | |
65 elif model == 'ham': | |
66 return(ham_model) | |
67 elif model == 'm1n': | |
68 return(m1n_model) | |
69 elif model == 'hs1f': | |
70 return(hs1f_model) | |
71 elif model == 'hs5f': | |
72 return(hs5f_model) | |
73 else: | |
74 sys.stderr.write('Unrecognized distance model: %s.\n' % model) | |
75 | |
76 | |
77 def indexJunctions(db_iter, fields=None, mode='gene', action='first'): | |
78 """ | |
79 Identifies preclonal groups by V, J and junction length | |
80 | |
81 Arguments: | |
82 db_iter = an iterator of IgRecords defined by readDbFile | |
83 fields = additional annotation fields to use to group preclones; | |
84 if None use only V, J and junction length | |
85 mode = specificity of alignment call to use for assigning preclones; | |
86 one of ('allele', 'gene') | |
87 action = how to handle multiple value fields when assigning preclones; | |
88 one of ('first', 'set') | |
89 | |
90 Returns: | |
91 a dictionary of {(V, J, junction length):[IgRecords]} | |
92 """ | |
93 # Define functions for grouping keys | |
94 if mode == 'allele' and fields is None: | |
95 def _get_key(rec, act): | |
96 return (rec.getVAllele(act), rec.getJAllele(act), | |
97 None if rec.junction is None else len(rec.junction)) | |
98 elif mode == 'gene' and fields is None: | |
99 def _get_key(rec, act): | |
100 return (rec.getVGene(act), rec.getJGene(act), | |
101 None if rec.junction is None else len(rec.junction)) | |
102 elif mode == 'allele' and fields is not None: | |
103 def _get_key(rec, act): | |
104 vdj = [rec.getVAllele(act), rec.getJAllele(act), | |
105 None if rec.junction is None else len(rec.junction)] | |
106 ann = [rec.toDict().get(k, None) for k in fields] | |
107 return tuple(chain(vdj, ann)) | |
108 elif mode == 'gene' and fields is not None: | |
109 def _get_key(rec, act): | |
110 vdj = [rec.getVGene(act), rec.getJGene(act), | |
111 None if rec.junction is None else len(rec.junction)] | |
112 ann = [rec.toDict().get(k, None) for k in fields] | |
113 return tuple(chain(vdj, ann)) | |
114 | |
115 start_time = time() | |
116 clone_index = {} | |
117 rec_count = 0 | |
118 for rec in db_iter: | |
119 key = _get_key(rec, action) | |
120 | |
121 # Print progress | |
122 if rec_count == 0: | |
123 print('PROGRESS> Grouping sequences') | |
124 | |
125 printProgress(rec_count, step=1000, start_time=start_time) | |
126 rec_count += 1 | |
127 | |
128 # Assigned passed preclone records to key and failed to index None | |
129 if all([k is not None and k != '' for k in key]): | |
130 #print key | |
131 # TODO: Has much slow. Should have less slow. | |
132 if action == 'set': | |
133 | |
134 f_range = list(range(2, 3 + (len(fields) if fields else 0))) | |
135 vdj_range = list(range(2)) | |
136 | |
137 # Check for any keys that have matching columns and junction length and overlapping genes/alleles | |
138 to_remove = [] | |
139 if len(clone_index) > (1 if None in clone_index else 0) and key not in clone_index: | |
140 key = list(key) | |
141 for k in clone_index: | |
142 if k is not None and all([key[i] == k[i] for i in f_range]): | |
143 if all([not set(key[i]).isdisjoint(set(k[i])) for i in vdj_range]): | |
144 for i in vdj_range: key[i] = tuple(set(key[i]).union(set(k[i]))) | |
145 to_remove.append(k) | |
146 | |
147 # Remove original keys, replace with union of all genes/alleles and append values to new key | |
148 val = [rec] | |
149 val += list(chain(*(clone_index.pop(k) for k in to_remove))) | |
150 clone_index[tuple(key)] = clone_index.get(tuple(key),[]) + val | |
151 | |
152 elif action == 'first': | |
153 clone_index.setdefault(key, []).append(rec) | |
154 else: | |
155 clone_index.setdefault(None, []).append(rec) | |
156 | |
157 printProgress(rec_count, step=1000, start_time=start_time, end=True) | |
158 | |
159 return clone_index | |
160 | |
161 | |
162 def distanceClones(records, model=default_bygroup_model, distance=default_distance, | |
163 dist_mat=None, norm=default_norm, sym=default_sym, | |
164 linkage=default_linkage, seq_field=default_seq_field): | |
165 """ | |
166 Separates a set of IgRecords into clones | |
167 | |
168 Arguments: | |
169 records = an iterator of IgRecords | |
170 model = substitution model used to calculate distance | |
171 distance = the distance threshold to assign clonal groups | |
172 dist_mat = pandas DataFrame of pairwise nucleotide or amino acid distances | |
173 norm = normalization method | |
174 sym = symmetry method | |
175 linkage = type of linkage | |
176 seq_field = sequence field used to calculate distance between records | |
177 | |
178 Returns: | |
179 a dictionary of lists defining {clone number: [IgRecords clonal group]} | |
180 """ | |
181 # Get distance matrix if not provided | |
182 if dist_mat is None: dist_mat = getModelMatrix(model) | |
183 | |
184 # Determine length of n-mers | |
185 if model in ['hs1f', 'm1n', 'aa', 'ham']: | |
186 nmer_len = 1 | |
187 elif model in ['hs5f']: | |
188 nmer_len = 5 | |
189 else: | |
190 sys.stderr.write('Unrecognized distance model: %s.\n' % model) | |
191 | |
192 # Define unique junction mapping | |
193 seq_map = {} | |
194 for ig in records: | |
195 seq = ig.getSeqField(seq_field) | |
196 # Check if sequence length is 0 | |
197 if len(seq) == 0: | |
198 return None | |
199 | |
200 seq = re.sub('[\.-]','N', str(seq)) | |
201 if model == 'aa': seq = translate(seq) | |
202 | |
203 seq_map.setdefault(seq, []).append(ig) | |
204 | |
205 # Process records | |
206 if len(seq_map) == 1: | |
207 return {1:records} | |
208 | |
209 # Define sequences | |
210 seqs = list(seq_map.keys()) | |
211 | |
212 # Calculate pairwise distance matrix | |
213 dists = calcDistances(seqs, nmer_len, dist_mat, norm, sym) | |
214 | |
215 # Perform hierarchical clustering | |
216 clusters = formClusters(dists, linkage, distance) | |
217 | |
218 # Turn clusters into clone dictionary | |
219 clone_dict = {} | |
220 for i, c in enumerate(clusters): | |
221 clone_dict.setdefault(c, []).extend(seq_map[seqs[i]]) | |
222 | |
223 return clone_dict | |
224 | |
225 | |
226 def distChen2010(records): | |
227 """ | |
228 Calculate pairwise distances as defined in Chen 2010 | |
229 | |
230 Arguments: | |
231 records = list of IgRecords where first is query to be compared to others in list | |
232 | |
233 Returns: | |
234 list of distances | |
235 """ | |
236 # Pull out query sequence and V/J information | |
237 query = records.popitem(last=False) | |
238 query_cdr3 = query.junction[3:-3] | |
239 query_v_allele = query.getVAllele() | |
240 query_v_gene = query.getVGene() | |
241 query_v_family = query.getVFamily() | |
242 query_j_allele = query.getJAllele() | |
243 query_j_gene = query.getJGene() | |
244 # Create alignment scoring dictionary | |
245 score_dict = getDNAScoreDict() | |
246 | |
247 scores = [0]*len(records) | |
248 for i in range(len(records)): | |
249 ld = pairwise2.align.globalds(query_cdr3, records[i].junction[3:-3], | |
250 score_dict, -1, -1, one_alignment_only=True) | |
251 # Check V similarity | |
252 if records[i].getVAllele() == query_v_allele: ld += 0 | |
253 elif records[i].getVGene() == query_v_gene: ld += 1 | |
254 elif records[i].getVFamily() == query_v_family: ld += 3 | |
255 else: ld += 5 | |
256 # Check J similarity | |
257 if records[i].getJAllele() == query_j_allele: ld += 0 | |
258 elif records[i].getJGene() == query_j_gene: ld += 1 | |
259 else: ld += 3 | |
260 # Divide by length | |
261 scores[i] = ld/max(len(records[i].junction[3:-3]), query_cdr3) | |
262 | |
263 return scores | |
264 | |
265 | |
266 def distAdemokun2011(records): | |
267 """ | |
268 Calculate pairwise distances as defined in Ademokun 2011 | |
269 | |
270 Arguments: | |
271 records = list of IgRecords where first is query to be compared to others in list | |
272 | |
273 Returns: | |
274 list of distances | |
275 """ | |
276 # Pull out query sequence and V family information | |
277 query = records.popitem(last=False) | |
278 query_cdr3 = query.junction[3:-3] | |
279 query_v_family = query.getVFamily() | |
280 # Create alignment scoring dictionary | |
281 score_dict = getDNAScoreDict() | |
282 | |
283 scores = [0]*len(records) | |
284 for i in range(len(records)): | |
285 | |
286 if abs(len(query_cdr3) - len(records[i].junction[3:-3])) > 10: | |
287 scores[i] = 1 | |
288 elif query_v_family != records[i].getVFamily(): | |
289 scores[i] = 1 | |
290 else: | |
291 ld = pairwise2.align.globalds(query_cdr3, records[i].junction[3:-3], | |
292 score_dict, -1, -1, one_alignment_only=True) | |
293 scores[i] = ld/min(len(records[i].junction[3:-3]), query_cdr3) | |
294 | |
295 return scores | |
296 | |
297 | |
298 def hierClust(dist_mat, method='chen2010'): | |
299 """ | |
300 Calculate hierarchical clustering | |
301 | |
302 Arguments: | |
303 dist_mat = square-formed distance matrix of pairwise CDR3 comparisons | |
304 | |
305 Returns: | |
306 list of cluster ids | |
307 """ | |
308 if method == 'chen2010': | |
309 clusters = formClusters(dist_mat, 'average', 0.32) | |
310 elif method == 'ademokun2011': | |
311 clusters = formClusters(dist_mat, 'complete', 0.25) | |
312 else: clusters = np.ones(dist_mat.shape[0]) | |
313 | |
314 return clusters | |
315 | |
316 # TODO: Merge duplicate feed, process and collect functions. | |
317 def feedQueue(alive, data_queue, db_file, group_func, group_args={}): | |
318 """ | |
319 Feeds the data queue with Ig records | |
320 | |
321 Arguments: | |
322 alive = a multiprocessing.Value boolean controlling whether processing continues | |
323 if False exit process | |
324 data_queue = a multiprocessing.Queue to hold data for processing | |
325 db_file = the Ig record database file | |
326 group_func = the function to use for assigning preclones | |
327 group_args = a dictionary of arguments to pass to group_func | |
328 | |
329 Returns: | |
330 None | |
331 """ | |
332 # Open input file and perform grouping | |
333 try: | |
334 # Iterate over Ig records and assign groups | |
335 db_iter = readDbFile(db_file) | |
336 clone_dict = group_func(db_iter, **group_args) | |
337 except: | |
338 #sys.stderr.write('Exception in feeder grouping step\n') | |
339 alive.value = False | |
340 raise | |
341 | |
342 # Add groups to data queue | |
343 try: | |
344 #print 'START FEED', alive.value | |
345 # Iterate over groups and feed data queue | |
346 clone_iter = iter(clone_dict.items()) | |
347 while alive.value: | |
348 # Get data from queue | |
349 if data_queue.full(): continue | |
350 else: data = next(clone_iter, None) | |
351 # Exit upon reaching end of iterator | |
352 if data is None: break | |
353 #print "FEED", alive.value, k | |
354 | |
355 # Feed queue | |
356 data_queue.put(DbData(*data)) | |
357 else: | |
358 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
359 % os.getpid()) | |
360 return None | |
361 except: | |
362 #sys.stderr.write('Exception in feeder queue feeding step\n') | |
363 alive.value = False | |
364 raise | |
365 | |
366 return None | |
367 | |
368 | |
369 def feedQueueClust(alive, data_queue, db_file, group_func=None, group_args={}): | |
370 """ | |
371 Feeds the data queue with Ig records | |
372 | |
373 Arguments: | |
374 alive = a multiprocessing.Value boolean controlling whether processing continues | |
375 if False exit process | |
376 data_queue = a multiprocessing.Queue to hold data for processing | |
377 db_file = the Ig record database file | |
378 | |
379 Returns: | |
380 None | |
381 """ | |
382 # Open input file and perform grouping | |
383 try: | |
384 # Iterate over Ig records and order by junction length | |
385 records = {} | |
386 db_iter = readDbFile(db_file) | |
387 for rec in db_iter: | |
388 records[rec.id] = rec | |
389 records = OrderedDict(sorted(list(records.items()), key=lambda i: i[1].junction_length)) | |
390 dist_dict = {} | |
391 for __ in range(len(records)): | |
392 k,v = records.popitem(last=False) | |
393 dist_dict[k] = [v].append(list(records.values())) | |
394 except: | |
395 #sys.stderr.write('Exception in feeder grouping step\n') | |
396 alive.value = False | |
397 raise | |
398 | |
399 # Add groups to data queue | |
400 try: | |
401 # print 'START FEED', alive.value | |
402 # Iterate over groups and feed data queue | |
403 dist_iter = iter(dist_dict.items()) | |
404 while alive.value: | |
405 # Get data from queue | |
406 if data_queue.full(): continue | |
407 else: data = next(dist_iter, None) | |
408 # Exit upon reaching end of iterator | |
409 if data is None: break | |
410 #print "FEED", alive.value, k | |
411 | |
412 # Feed queue | |
413 data_queue.put(DbData(*data)) | |
414 else: | |
415 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
416 % os.getpid()) | |
417 return None | |
418 except: | |
419 #sys.stderr.write('Exception in feeder queue feeding step\n') | |
420 alive.value = False | |
421 raise | |
422 | |
423 return None | |
424 | |
425 | |
426 def processQueue(alive, data_queue, result_queue, clone_func, clone_args): | |
427 """ | |
428 Pulls from data queue, performs calculations, and feeds results queue | |
429 | |
430 Arguments: | |
431 alive = a multiprocessing.Value boolean controlling whether processing continues | |
432 if False exit process | |
433 data_queue = a multiprocessing.Queue holding data to process | |
434 result_queue = a multiprocessing.Queue to hold processed results | |
435 clone_func = the function to call for clonal assignment | |
436 clone_args = a dictionary of arguments to pass to clone_func | |
437 | |
438 Returns: | |
439 None | |
440 """ | |
441 try: | |
442 # Iterator over data queue until sentinel object reached | |
443 while alive.value: | |
444 # Get data from queue | |
445 if data_queue.empty(): continue | |
446 else: data = data_queue.get() | |
447 # Exit upon reaching sentinel | |
448 if data is None: break | |
449 | |
450 # Define result object for iteration and get data records | |
451 records = data.data | |
452 result = DbResult(data.id, records) | |
453 | |
454 # Check for invalid data (due to failed indexing) and add failed result | |
455 if not data: | |
456 result_queue.put(result) | |
457 continue | |
458 | |
459 # Add V(D)J to log | |
460 result.log['ID'] = ','.join([str(x) for x in data.id]) | |
461 result.log['VALLELE'] = ','.join(set([(r.getVAllele() or '') for r in records])) | |
462 result.log['DALLELE'] = ','.join(set([(r.getDAllele() or '') for r in records])) | |
463 result.log['JALLELE'] = ','.join(set([(r.getJAllele() or '') for r in records])) | |
464 result.log['JUNCLEN'] = ','.join(set([(str(len(r.junction)) or '0') for r in records])) | |
465 result.log['SEQUENCES'] = len(records) | |
466 | |
467 # Checking for preclone failure and assign clones | |
468 clones = clone_func(records, **clone_args) if data else None | |
469 | |
470 # import cProfile | |
471 # prof = cProfile.Profile() | |
472 # clones = prof.runcall(clone_func, records, **clone_args) | |
473 # prof.dump_stats('worker-%d.prof' % os.getpid()) | |
474 | |
475 if clones is not None: | |
476 result.results = clones | |
477 result.valid = True | |
478 result.log['CLONES'] = len(clones) | |
479 else: | |
480 result.log['CLONES'] = 0 | |
481 | |
482 # Feed results to result queue | |
483 result_queue.put(result) | |
484 else: | |
485 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
486 % os.getpid()) | |
487 return None | |
488 except: | |
489 #sys.stderr.write('Exception in worker\n') | |
490 alive.value = False | |
491 raise | |
492 | |
493 return None | |
494 | |
495 | |
496 def processQueueClust(alive, data_queue, result_queue, clone_func, clone_args): | |
497 """ | |
498 Pulls from data queue, performs calculations, and feeds results queue | |
499 | |
500 Arguments: | |
501 alive = a multiprocessing.Value boolean controlling whether processing continues | |
502 if False exit process | |
503 data_queue = a multiprocessing.Queue holding data to process | |
504 result_queue = a multiprocessing.Queue to hold processed results | |
505 clone_func = the function to call for calculating pairwise distances between sequences | |
506 clone_args = a dictionary of arguments to pass to clone_func | |
507 | |
508 Returns: | |
509 None | |
510 """ | |
511 | |
512 try: | |
513 # print 'START WORK', alive.value | |
514 # Iterator over data queue until sentinel object reached | |
515 while alive.value: | |
516 # Get data from queue | |
517 if data_queue.empty(): continue | |
518 else: data = data_queue.get() | |
519 # Exit upon reaching sentinel | |
520 if data is None: break | |
521 # print "WORK", alive.value, data['id'] | |
522 | |
523 # Define result object for iteration and get data records | |
524 records = data.data | |
525 result = DbResult(data.id, records) | |
526 | |
527 # Create row of distance matrix and check for error | |
528 dist_row = clone_func(records, **clone_args) if data else None | |
529 if dist_row is not None: | |
530 result.results = dist_row | |
531 result.valid = True | |
532 | |
533 # Feed results to result queue | |
534 result_queue.put(result) | |
535 else: | |
536 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
537 % os.getpid()) | |
538 return None | |
539 except: | |
540 #sys.stderr.write('Exception in worker\n') | |
541 alive.value = False | |
542 raise | |
543 | |
544 return None | |
545 | |
546 | |
547 def collectQueue(alive, result_queue, collect_queue, db_file, out_args, cluster_func=None, cluster_args={}): | |
548 """ | |
549 Assembles results from a queue of individual sequence results and manages log/file I/O | |
550 | |
551 Arguments: | |
552 alive = a multiprocessing.Value boolean controlling whether processing continues | |
553 if False exit process | |
554 result_queue = a multiprocessing.Queue holding processQueue results | |
555 collect_queue = a multiprocessing.Queue to store collector return values | |
556 db_file = the input database file name | |
557 out_args = common output argument dictionary from parseCommonArgs | |
558 cluster_func = the function to call for carrying out clustering on distance matrix | |
559 cluster_args = a dictionary of arguments to pass to cluster_func | |
560 | |
561 Returns: | |
562 None | |
563 (adds 'log' and 'out_files' to collect_dict) | |
564 """ | |
565 # Open output files | |
566 try: | |
567 # Count records and define output format | |
568 out_type = getFileType(db_file) if out_args['out_type'] is None \ | |
569 else out_args['out_type'] | |
570 result_count = countDbFile(db_file) | |
571 | |
572 # Defined successful output handle | |
573 pass_handle = getOutputHandle(db_file, | |
574 out_label='clone-pass', | |
575 out_dir=out_args['out_dir'], | |
576 out_name=out_args['out_name'], | |
577 out_type=out_type) | |
578 pass_writer = getDbWriter(pass_handle, db_file, add_fields='CLONE') | |
579 | |
580 # Defined failed alignment output handle | |
581 if out_args['failed']: | |
582 fail_handle = getOutputHandle(db_file, | |
583 out_label='clone-fail', | |
584 out_dir=out_args['out_dir'], | |
585 out_name=out_args['out_name'], | |
586 out_type=out_type) | |
587 fail_writer = getDbWriter(fail_handle, db_file) | |
588 else: | |
589 fail_handle = None | |
590 fail_writer = None | |
591 | |
592 # Define log handle | |
593 if out_args['log_file'] is None: | |
594 log_handle = None | |
595 else: | |
596 log_handle = open(out_args['log_file'], 'w') | |
597 except: | |
598 #sys.stderr.write('Exception in collector file opening step\n') | |
599 alive.value = False | |
600 raise | |
601 | |
602 # Get results from queue and write to files | |
603 try: | |
604 #print 'START COLLECT', alive.value | |
605 # Iterator over results queue until sentinel object reached | |
606 start_time = time() | |
607 rec_count = clone_count = pass_count = fail_count = 0 | |
608 while alive.value: | |
609 # Get result from queue | |
610 if result_queue.empty(): continue | |
611 else: result = result_queue.get() | |
612 # Exit upon reaching sentinel | |
613 if result is None: break | |
614 #print "COLLECT", alive.value, result['id'] | |
615 | |
616 # Print progress for previous iteration and update record count | |
617 if rec_count == 0: | |
618 print('PROGRESS> Assigning clones') | |
619 printProgress(rec_count, result_count, 0.05, start_time) | |
620 rec_count += len(result.data) | |
621 | |
622 # Write passed and failed records | |
623 if result: | |
624 for clone in result.results.values(): | |
625 clone_count += 1 | |
626 for i, rec in enumerate(clone): | |
627 rec.annotations['CLONE'] = clone_count | |
628 pass_writer.writerow(rec.toDict()) | |
629 pass_count += 1 | |
630 result.log['CLONE%i-%i' % (clone_count, i + 1)] = str(rec.junction) | |
631 | |
632 else: | |
633 for i, rec in enumerate(result.data): | |
634 if fail_writer is not None: fail_writer.writerow(rec.toDict()) | |
635 fail_count += 1 | |
636 result.log['CLONE0-%i' % (i + 1)] = str(rec.junction) | |
637 | |
638 # Write log | |
639 printLog(result.log, handle=log_handle) | |
640 else: | |
641 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
642 % os.getpid()) | |
643 return None | |
644 | |
645 # Print total counts | |
646 printProgress(rec_count, result_count, 0.05, start_time) | |
647 | |
648 # Close file handles | |
649 pass_handle.close() | |
650 if fail_handle is not None: fail_handle.close() | |
651 if log_handle is not None: log_handle.close() | |
652 | |
653 # Update return list | |
654 log = OrderedDict() | |
655 log['OUTPUT'] = os.path.basename(pass_handle.name) | |
656 log['CLONES'] = clone_count | |
657 log['RECORDS'] = rec_count | |
658 log['PASS'] = pass_count | |
659 log['FAIL'] = fail_count | |
660 collect_dict = {'log':log, 'out_files': [pass_handle.name]} | |
661 collect_queue.put(collect_dict) | |
662 except: | |
663 #sys.stderr.write('Exception in collector result processing step\n') | |
664 alive.value = False | |
665 raise | |
666 | |
667 return None | |
668 | |
669 | |
670 def collectQueueClust(alive, result_queue, collect_queue, db_file, out_args, cluster_func, cluster_args): | |
671 """ | |
672 Assembles results from a queue of individual sequence results and manages log/file I/O | |
673 | |
674 Arguments: | |
675 alive = a multiprocessing.Value boolean controlling whether processing continues | |
676 if False exit process | |
677 result_queue = a multiprocessing.Queue holding processQueue results | |
678 collect_queue = a multiprocessing.Queue to store collector return values | |
679 db_file = the input database file name | |
680 out_args = common output argument dictionary from parseCommonArgs | |
681 cluster_func = the function to call for carrying out clustering on distance matrix | |
682 cluster_args = a dictionary of arguments to pass to cluster_func | |
683 | |
684 Returns: | |
685 None | |
686 (adds 'log' and 'out_files' to collect_dict) | |
687 """ | |
688 # Open output files | |
689 try: | |
690 | |
691 # Iterate over Ig records to count and order by junction length | |
692 result_count = 0 | |
693 records = {} | |
694 # print 'Reading file...' | |
695 db_iter = readDbFile(db_file) | |
696 for rec in db_iter: | |
697 records[rec.id] = rec | |
698 result_count += 1 | |
699 records = OrderedDict(sorted(list(records.items()), key=lambda i: i[1].junction_length)) | |
700 | |
701 # Define empty matrix to store assembled results | |
702 dist_mat = np.zeros((result_count,result_count)) | |
703 | |
704 # Count records and define output format | |
705 out_type = getFileType(db_file) if out_args['out_type'] is None \ | |
706 else out_args['out_type'] | |
707 | |
708 # Defined successful output handle | |
709 pass_handle = getOutputHandle(db_file, | |
710 out_label='clone-pass', | |
711 out_dir=out_args['out_dir'], | |
712 out_name=out_args['out_name'], | |
713 out_type=out_type) | |
714 pass_writer = getDbWriter(pass_handle, db_file, add_fields='CLONE') | |
715 | |
716 # Defined failed cloning output handle | |
717 if out_args['failed']: | |
718 fail_handle = getOutputHandle(db_file, | |
719 out_label='clone-fail', | |
720 out_dir=out_args['out_dir'], | |
721 out_name=out_args['out_name'], | |
722 out_type=out_type) | |
723 fail_writer = getDbWriter(fail_handle, db_file) | |
724 else: | |
725 fail_handle = None | |
726 fail_writer = None | |
727 | |
728 # Open log file | |
729 if out_args['log_file'] is None: | |
730 log_handle = None | |
731 else: | |
732 log_handle = open(out_args['log_file'], 'w') | |
733 except: | |
734 alive.value = False | |
735 raise | |
736 | |
737 try: | |
738 # Iterator over results queue until sentinel object reached | |
739 start_time = time() | |
740 row_count = rec_count = 0 | |
741 while alive.value: | |
742 # Get result from queue | |
743 if result_queue.empty(): continue | |
744 else: result = result_queue.get() | |
745 # Exit upon reaching sentinel | |
746 if result is None: break | |
747 | |
748 # Print progress for previous iteration | |
749 if row_count == 0: | |
750 print('PROGRESS> Assigning clones') | |
751 printProgress(row_count, result_count, 0.05, start_time) | |
752 | |
753 # Update counts for iteration | |
754 row_count += 1 | |
755 rec_count += len(result) | |
756 | |
757 # Add result row to distance matrix | |
758 if result: | |
759 dist_mat[list(range(result_count-len(result),result_count)),result_count-len(result)] = result.results | |
760 | |
761 else: | |
762 sys.stderr.write('PID %s: Error in sibling process detected. Cleaning up.\n' \ | |
763 % os.getpid()) | |
764 return None | |
765 | |
766 # Calculate linkage and carry out clustering | |
767 # print dist_mat | |
768 clusters = cluster_func(dist_mat, **cluster_args) if dist_mat is not None else None | |
769 clones = {} | |
770 # print clusters | |
771 for i, c in enumerate(clusters): | |
772 clones.setdefault(c, []).append(records[list(records.keys())[i]]) | |
773 | |
774 # Write passed and failed records | |
775 clone_count = pass_count = fail_count = 0 | |
776 if clones: | |
777 for clone in clones.values(): | |
778 clone_count += 1 | |
779 for i, rec in enumerate(clone): | |
780 rec.annotations['CLONE'] = clone_count | |
781 pass_writer.writerow(rec.toDict()) | |
782 pass_count += 1 | |
783 #result.log['CLONE%i-%i' % (clone_count, i + 1)] = str(rec.junction) | |
784 | |
785 else: | |
786 for i, rec in enumerate(result.data): | |
787 fail_writer.writerow(rec.toDict()) | |
788 fail_count += 1 | |
789 #result.log['CLONE0-%i' % (i + 1)] = str(rec.junction) | |
790 | |
791 # Print final progress | |
792 printProgress(row_count, result_count, 0.05, start_time) | |
793 | |
794 # Close file handles | |
795 pass_handle.close() | |
796 if fail_handle is not None: fail_handle.close() | |
797 if log_handle is not None: log_handle.close() | |
798 | |
799 # Update return list | |
800 log = OrderedDict() | |
801 log['OUTPUT'] = os.path.basename(pass_handle.name) | |
802 log['CLONES'] = clone_count | |
803 log['RECORDS'] = rec_count | |
804 log['PASS'] = pass_count | |
805 log['FAIL'] = fail_count | |
806 collect_dict = {'log':log, 'out_files': [pass_handle.name]} | |
807 collect_queue.put(collect_dict) | |
808 except: | |
809 alive.value = False | |
810 raise | |
811 | |
812 return None | |
813 | |
814 | |
815 def defineClones(db_file, feed_func, work_func, collect_func, clone_func, cluster_func=None, | |
816 group_func=None, group_args={}, clone_args={}, cluster_args={}, | |
817 out_args=default_out_args, nproc=None, queue_size=None): | |
818 """ | |
819 Define clonally related sequences | |
820 | |
821 Arguments: | |
822 db_file = filename of input database | |
823 feed_func = the function that feeds the queue | |
824 work_func = the worker function that will run on each CPU | |
825 collect_func = the function that collects results from the workers | |
826 group_func = the function to use for assigning preclones | |
827 clone_func = the function to use for determining clones within preclonal groups | |
828 group_args = a dictionary of arguments to pass to group_func | |
829 clone_args = a dictionary of arguments to pass to clone_func | |
830 out_args = common output argument dictionary from parseCommonArgs | |
831 nproc = the number of processQueue processes; | |
832 if None defaults to the number of CPUs | |
833 queue_size = maximum size of the argument queue; | |
834 if None defaults to 2*nproc | |
835 | |
836 Returns: | |
837 a list of successful output file names | |
838 """ | |
839 # Print parameter info | |
840 log = OrderedDict() | |
841 log['START'] = 'DefineClones' | |
842 log['DB_FILE'] = os.path.basename(db_file) | |
843 if group_func is not None: | |
844 log['GROUP_FUNC'] = group_func.__name__ | |
845 log['GROUP_ARGS'] = group_args | |
846 log['CLONE_FUNC'] = clone_func.__name__ | |
847 | |
848 # TODO: this is yucky, but can be fixed by using a model class | |
849 clone_log = clone_args.copy() | |
850 if 'dist_mat' in clone_log: del clone_log['dist_mat'] | |
851 log['CLONE_ARGS'] = clone_log | |
852 | |
853 if cluster_func is not None: | |
854 log['CLUSTER_FUNC'] = cluster_func.__name__ | |
855 log['CLUSTER_ARGS'] = cluster_args | |
856 log['NPROC'] = nproc | |
857 printLog(log) | |
858 | |
859 # Define feeder function and arguments | |
860 feed_args = {'db_file': db_file, | |
861 'group_func': group_func, | |
862 'group_args': group_args} | |
863 # Define worker function and arguments | |
864 work_args = {'clone_func': clone_func, | |
865 'clone_args': clone_args} | |
866 # Define collector function and arguments | |
867 collect_args = {'db_file': db_file, | |
868 'out_args': out_args, | |
869 'cluster_func': cluster_func, | |
870 'cluster_args': cluster_args} | |
871 | |
872 # Call process manager | |
873 result = manageProcesses(feed_func, work_func, collect_func, | |
874 feed_args, work_args, collect_args, | |
875 nproc, queue_size) | |
876 | |
877 # Print log | |
878 result['log']['END'] = 'DefineClones' | |
879 printLog(result['log']) | |
880 | |
881 return result['out_files'] | |
882 | |
883 | |
884 def getArgParser(): | |
885 """ | |
886 Defines the ArgumentParser | |
887 | |
888 Arguments: | |
889 None | |
890 | |
891 Returns: | |
892 an ArgumentParser object | |
893 """ | |
894 # Define input and output fields | |
895 fields = dedent( | |
896 ''' | |
897 output files: | |
898 clone-pass | |
899 database with assigned clonal group numbers. | |
900 clone-fail | |
901 database with records failing clonal grouping. | |
902 | |
903 required fields: | |
904 SEQUENCE_ID, V_CALL or V_CALL_GENOTYPED, D_CALL, J_CALL, JUNCTION_LENGTH | |
905 | |
906 <field> | |
907 sequence field specified by the --sf parameter | |
908 | |
909 output fields: | |
910 CLONE | |
911 ''') | |
912 | |
913 # Define ArgumentParser | |
914 parser = ArgumentParser(description=__doc__, epilog=fields, | |
915 formatter_class=CommonHelpFormatter) | |
916 parser.add_argument('--version', action='version', | |
917 version='%(prog)s:' + ' %s-%s' %(__version__, __date__)) | |
918 subparsers = parser.add_subparsers(title='subcommands', dest='command', metavar='', | |
919 help='Cloning method') | |
920 # TODO: This is a temporary fix for Python issue 9253 | |
921 subparsers.required = True | |
922 | |
923 # Parent parser | |
924 parser_parent = getCommonArgParser(seq_in=False, seq_out=False, db_in=True, | |
925 multiproc=True) | |
926 | |
927 # Distance cloning method | |
928 parser_bygroup = subparsers.add_parser('bygroup', parents=[parser_parent], | |
929 formatter_class=CommonHelpFormatter, | |
930 help='''Defines clones as having same V assignment, | |
931 J assignment, and junction length with | |
932 specified substitution distance model.''') | |
933 parser_bygroup.add_argument('-f', nargs='+', action='store', dest='fields', default=None, | |
934 help='Additional fields to use for grouping clones (non VDJ)') | |
935 parser_bygroup.add_argument('--mode', action='store', dest='mode', | |
936 choices=('allele', 'gene'), default='gene', | |
937 help='''Specifies whether to use the V(D)J allele or gene for | |
938 initial grouping.''') | |
939 parser_bygroup.add_argument('--act', action='store', dest='action', default='set', | |
940 choices=('first', 'set'), | |
941 help='''Specifies how to handle multiple V(D)J assignments | |
942 for initial grouping.''') | |
943 parser_bygroup.add_argument('--model', action='store', dest='model', | |
944 choices=('aa', 'ham', 'm1n', 'hs1f', 'hs5f'), | |
945 default=default_bygroup_model, | |
946 help='''Specifies which substitution model to use for | |
947 calculating distance between sequences. Where m1n is the | |
948 mouse single nucleotide transition/trasversion model | |
949 of Smith et al, 1996; hs1f is the human single | |
950 nucleotide model derived from Yaari et al, 2013; hs5f | |
951 is the human S5F model of Yaari et al, 2013; ham is | |
952 nucleotide Hamming distance; and aa is amino acid | |
953 Hamming distance. The hs5f data should be | |
954 considered experimental.''') | |
955 parser_bygroup.add_argument('--dist', action='store', dest='distance', type=float, | |
956 default=default_distance, | |
957 help='The distance threshold for clonal grouping') | |
958 parser_bygroup.add_argument('--norm', action='store', dest='norm', | |
959 choices=('len', 'mut', 'none'), default=default_norm, | |
960 help='''Specifies how to normalize distances. One of none | |
961 (do not normalize), len (normalize by length), | |
962 or mut (normalize by number of mutations between sequences).''') | |
963 parser_bygroup.add_argument('--sym', action='store', dest='sym', | |
964 choices=('avg', 'min'), default=default_sym, | |
965 help='''Specifies how to combine asymmetric distances. One of avg | |
966 (average of A->B and B->A) or min (minimum of A->B and B->A).''') | |
967 parser_bygroup.add_argument('--link', action='store', dest='linkage', | |
968 choices=('single', 'average', 'complete'), default=default_linkage, | |
969 help='''Type of linkage to use for hierarchical clustering.''') | |
970 parser_bygroup.add_argument('--sf', action='store', dest='seq_field', | |
971 default=default_seq_field, | |
972 help='''The name of the field to be used to calculate | |
973 distance between records''') | |
974 parser_bygroup.set_defaults(feed_func=feedQueue) | |
975 parser_bygroup.set_defaults(work_func=processQueue) | |
976 parser_bygroup.set_defaults(collect_func=collectQueue) | |
977 parser_bygroup.set_defaults(group_func=indexJunctions) | |
978 parser_bygroup.set_defaults(clone_func=distanceClones) | |
979 | |
980 | |
981 # Hierarchical clustering cloning method | |
982 parser_hclust = subparsers.add_parser('hclust', parents=[parser_parent], | |
983 formatter_class=CommonHelpFormatter, | |
984 help='Defines clones by specified distance metric on CDR3s and \ | |
985 cutting of hierarchical clustering tree') | |
986 # parser_hclust.add_argument('-f', nargs='+', action='store', dest='fields', default=None, | |
987 # help='Fields to use for grouping clones (non VDJ)') | |
988 parser_hclust.add_argument('--method', action='store', dest='method', | |
989 choices=('chen2010', 'ademokun2011'), default=default_hclust_model, | |
990 help='Specifies which cloning method to use for calculating distance \ | |
991 between CDR3s, computing linkage, and cutting clusters') | |
992 parser_hclust.set_defaults(feed_func=feedQueueClust) | |
993 parser_hclust.set_defaults(work_func=processQueueClust) | |
994 parser_hclust.set_defaults(collect_func=collectQueueClust) | |
995 parser_hclust.set_defaults(cluster_func=hierClust) | |
996 | |
997 return parser | |
998 | |
999 | |
1000 if __name__ == '__main__': | |
1001 """ | |
1002 Parses command line arguments and calls main function | |
1003 """ | |
1004 # Parse arguments | |
1005 parser = getArgParser() | |
1006 args = parser.parse_args() | |
1007 args_dict = parseCommonArgs(args) | |
1008 # Convert case of fields | |
1009 if 'seq_field' in args_dict: | |
1010 args_dict['seq_field'] = args_dict['seq_field'].upper() | |
1011 if 'fields' in args_dict and args_dict['fields'] is not None: | |
1012 args_dict['fields'] = [f.upper() for f in args_dict['fields']] | |
1013 | |
1014 # Define clone_args | |
1015 if args.command == 'bygroup': | |
1016 args_dict['group_args'] = {'fields': args_dict['fields'], | |
1017 'action': args_dict['action'], | |
1018 'mode':args_dict['mode']} | |
1019 args_dict['clone_args'] = {'model': args_dict['model'], | |
1020 'distance': args_dict['distance'], | |
1021 'norm': args_dict['norm'], | |
1022 'sym': args_dict['sym'], | |
1023 'linkage': args_dict['linkage'], | |
1024 'seq_field': args_dict['seq_field']} | |
1025 | |
1026 # TODO: can be cleaned up with abstract model class | |
1027 args_dict['clone_args']['dist_mat'] = getModelMatrix(args_dict['model']) | |
1028 | |
1029 del args_dict['fields'] | |
1030 del args_dict['action'] | |
1031 del args_dict['mode'] | |
1032 del args_dict['model'] | |
1033 del args_dict['distance'] | |
1034 del args_dict['norm'] | |
1035 del args_dict['sym'] | |
1036 del args_dict['linkage'] | |
1037 del args_dict['seq_field'] | |
1038 | |
1039 # Define clone_args | |
1040 if args.command == 'hclust': | |
1041 dist_funcs = {'chen2010':distChen2010, 'ademokun2011':distAdemokun2011} | |
1042 args_dict['clone_func'] = dist_funcs[args_dict['method']] | |
1043 args_dict['cluster_args'] = {'method': args_dict['method']} | |
1044 #del args_dict['fields'] | |
1045 del args_dict['method'] | |
1046 | |
1047 # Call defineClones | |
1048 del args_dict['command'] | |
1049 del args_dict['db_files'] | |
1050 for f in args.__dict__['db_files']: | |
1051 args_dict['db_file'] = f | |
1052 defineClones(**args_dict) |