-
Notifications
You must be signed in to change notification settings - Fork 0
/
util.py
313 lines (277 loc) · 10.2 KB
/
util.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
from __future__ import division, print_function
import os
import os.path as osp
import numpy as np
import pandas as pd
import pysam
from cylocobs import LocObs
from cyregcov import RegCov
from cyrowmaker import CyCovariateRowMaker
def positive_int(val):
v = int(val)
if v <= 0:
raise ValueError('invalid positive integer {}'.format(val))
return v
def nonneg_int(val):
v = int(val)
if v < 0:
raise ValueError('invalid positive integer {}'.format(val))
return v
def probability(val):
v = float(val)
if v < 0 or v > 1:
raise ValueError('invalid probability {}'.format(val))
return v
REVCOMPBASES='ACGTN'
REVCOMPRCBASES='TGCAN'
def rev_comp(seq):
try:
return ''.join([REVCOMPRCBASES[REVCOMPBASES.index(b)] for b in seq])
except ValueError:
raise ValueError('invalid base in {}'.format(seq))
raise ValueError('problem in revcomp')
def get_counts(aln_file, *args, **kwargs):
counts = aln_file.count_coverage(*args, **kwargs)
counts = np.array([np.array(arr) for arr in counts])
counts = np.transpose(counts)
return counts
def get_consensus(counts, min_cov = 1):
consensus = np.array(list('ACGT'))[counts.argmax(1)]
consensus[counts.sum(1) < min_cov] = 'N'
consensus = ''.join(list(consensus))
return consensus
def get_bams(bams_fn):
'''
bams_fn file with a list of bams
Returns
tuple of basenames of bam fns (list), a prefix, and bams (dict, key is
basename)
'''
bam_fns = []
bams = {}
prefix = None
with open(bams_fn) as fin:
for line in fin:
bam_fp = line.strip()
head, bam_fn = osp.split(bam_fp)
if prefix is not None:
if head != prefix:
raise ValueError('all bam files must be in the same directory')
else:
prefix = head
if not osp.isfile(bam_fp):
raise ValueError('could not find bam file {}'.format(bam_fn))
if bam_fn in bam_fns:
raise ValueError('all BAM filenames must be unique')
bam_fns.append(bam_fn)
bam = pysam.AlignmentFile(bam_fp)
bams[bam_fn] = bam
return prefix, bam_fns, bams
def get_ref_names(refs_input, bams):
if osp.isfile(refs_input):
ref_names = []
with open(refs_input) as fin:
for line in fin:
ref_names.append(line.strip())
else:
ref_names = [el.strip() for el in refs_input.split(',')]
for bam_fn, bam in bams.iteritems():
for ref in ref_names:
if ref not in bam.references:
err = 'could not find reference {} in bam file {}'.format(
ref, bam_fn)
raise ValueError(err)
return ref_names
# these will be memory hogs
def get_all_counts(bams, refs, min_bq, bam_fn_prefix):
'''
bams dict of bam_fn:AlignmentFiles
refs list of reference names
min_bq minimum base quality
bam_fn_prefix filename prefix
Returns
a dictionary of counts, where counts[ref][bam_fn] gives the array of
counts for that reference sequence and bam file
'''
counts = {}
for ref in refs:
counts[ref] = {}
for bam_fn, bam in bams.iteritems():
bamref = bam.header['SQ']
reflen_found = False
for d in bamref:
if d['SN'] == ref:
reflen = d['LN']
reflen_found = True
break
if not reflen_found:
raise ValueError('length of ref {} not found in {}'.format(
ref, bam_fn))
print('# getting counts for {}'.format(bam_fn))
count_fn = bam_fn_prefix + '/' + bam_fn + '.counts'
if osp.exists(count_fn):
print('# loading counts for {} from counts file'.format(bam_fn))
c_pd = pd.read_csv(count_fn, sep='\t', header=0, comment='#')
c_pd['A'] = c_pd['A'] + c_pd['a']
c_pd['C'] = c_pd['C'] + c_pd['c']
c_pd['G'] = c_pd['G'] + c_pd['g']
c_pd['T'] = c_pd['T'] + c_pd['t']
c = c_pd[list('ACGT')].values
else:
c = get_counts(bam, contig=ref, start = 0, end = reflen,
quality_threshold = min_bq)
counts[ref][bam_fn] = c
return counts
def get_freqs(counts):
freqs = {}
for ref, ref_counts in counts.iteritems():
freqs[ref] = {}
for bam_fn, c in ref_counts.iteritems():
# using jit or cython, this can be done with 1/2 the memory
f = c / np.maximum(c.sum(1),1)[:,None]
freqs[ref][bam_fn] = f
return freqs
def determine_candidates(freqs, min_candidate_freq):
candidates = {}
for ref, ref_freqs in freqs.iteritems():
candidates[ref] = {}
for bam_fn, f in ref_freqs.iteritems():
is_candidate = (f > min_candidate_freq).sum(1) > 1
candidates[ref][bam_fn] = is_candidate
return candidates
def get_all_consensuses(counts, min_coverage):
all_con = {}
for ref, ref_counts in counts.iteritems():
all_con[ref] = {}
for bam_fn, c in ref_counts.iteritems():
consensus = get_consensus(c, min_cov=min_coverage)
all_con[ref][bam_fn] = consensus
return all_con
def get_row_makers(bam_fns, refs, context_len, dend_roundby, consensuses,
use_mapq, use_bam):
'''
bam_fns list of bam filenames
refs list of reference sequence names
context_len number of preceding bases to include as covariates
dend_roundby how much to round distance from beginning by
consensuses dict of dicts, consensuses[bam_fn][ref] gives the
consensus sequence for ref sequence ref in bam file
bam_fn
Returns tuple (d, rowlen)
d is dict of dict of dict of CovariateRowMakers:
rm[ref][bam_fn][base] gives the CovariateRowMaker for this bam, ref,
and base
rowlen is the length of a row in the matrix
'''
rm = {}
rowlen = None
for ref in refs:
rm[ref] = {}
for bam_fn in bam_fns:
cons = consensuses[ref][bam_fn]
other_cons = [
consensuses[ref][fn] for fn in bam_fns if fn != bam_fn]
thisrm = CyCovariateRowMaker(
context_len,
dend_roundby,
cons,
other_cons,
bam_fns,
use_mq = use_mapq,
use_bam = use_bam)
rm[ref][bam_fn] = thisrm
l = thisrm.rowlen
if rowlen is None:
rowlen = l
else:
assert rowlen == l, 'error: multiple row lengths'
rowlen = rm[ref][bam_fn].rowlen
return rm, rowlen
def get_all_majorminor(all_counts):
all_majorminor = {}
allbases = np.array(list('ACGT'))
for ref, refcounts in all_counts.iteritems():
all_majorminor[ref] = {}
for bamname, bamcounts in refcounts.iteritems():
all_majorminor[ref][bamname] = []
thismm = all_majorminor[ref][bamname]
# bamcounts is a 2-d np.ndarray, shape (seqlen, 4)
# (yes, it does provide zero-counts for position at which
# no reads aligned)
for i, order in enumerate(bamcounts.argsort(1)):
order = order[::-1]
counts = bamcounts[i,order]
bases = allbases[order]
major = bytes(bases[0]) if counts[0] > 0 else b'N'
# note that minor is arbitrarily decided if there's a tie
minor = bytes(bases[1]) if counts[1] > 0 else b'N'
thismm.append((major, minor))
return all_majorminor
def get_covariate_matrices(rowlen):
covmat = RegCov(rowlen)
return covmat
def get_locus_observations(all_majorminor):
locobs = {}
for ref, refmm in all_majorminor.iteritems():
locobs[ref] = {}
for bam, bammm in refmm.iteritems():
locobs[ref][bam] = []
for i, _ in enumerate(bammm):
# indexed [forward(0)/reverse(1)][read1[0]/read2[1]]
locobs[ref][bam].append(((LocObs(),LocObs()),(LocObs(),LocObs())))
return locobs
def collect_covariate_matrices(cov):
ret = cov.covariate_matrix()
return ret
def collect_indiv_loc_obs(obs):
c = lambda x: x.counts()
m = obs
col = ( (c(m[0][0]), c(m[0][1])), (c(m[1][0]), c(m[1][1])) )
return col
def collect_loc_obs(locobs):
ret = {}
for ref, refobs in locobs.iteritems():
ret[ref] = {}
for bam, bamobs in refobs.iteritems():
ret[ref][bam] = []
for lobs in bamobs:
ret[ref][bam].append(collect_indiv_loc_obs(lobs))
return ret
def sort_lo(lo):
lo_sorted = {}
for chrom, lochrom in lo.iteritems():
lo_sorted[chrom] = {}
for bam_fn, lobam in lochrom.iteritems():
lo_sorted[chrom][bam_fn] = []
for locidx, loclo in enumerate(lobam):
thislo = []
for fr in [0,1]:
p = []
for r in [0,1]:
a = loclo[fr][r]
p.append(a[np.argsort(a[:,0])].astype(np.int32).copy())
p = tuple(p)
thislo.append(p)
thislo = tuple(thislo)
lo_sorted[chrom][bam_fn].append(thislo)
return lo_sorted
def normalize_covariates(cm):
retcm = cm.copy()
min_maxes = [(1,1)]
# starting with second column because first is constant!
for j in range(1,cm.shape[1]):
m = retcm[:,j].min()
M = retcm[:,j].max()
if m != M:
retcm[:,j] = (retcm[:,j]-m)/(M-m)
else:
retcm[:,j] = 0
min_maxes.append((m,M))
return retcm, min_maxes
def get_reference_lengths(reference_names, bam_fns, all_consensuses):
ref_lens = {}
for ref_name in reference_names:
ref_lens[ref_name] = {}
for bam_fn in bam_fns:
ref_lens[ref_name][bam_fn] = len(all_consensuses[ref_name][bam_fn])
return ref_lens