/
tandem-genotypes-join
executable file
·241 lines (214 loc) · 8.6 KB
/
tandem-genotypes-join
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
#! /usr/bin/env python
# Copyright 2018 Martin C. Frith
from __future__ import division, print_function
import gzip
import itertools
import operator
import optparse
import os
import signal
import sys
def myOpen(fileName):
if fileName == "-":
return sys.stdin
if fileName.endswith(".gz"):
return gzip.open(fileName, "rt") # xxx dubious for Python2
return open(fileName)
def genePartScoresFromLines(lines):
for line in lines:
fields = line.split()
if fields:
yield fields[0], float(fields[1])
def complement(x):
return "TGCA"["ACGT".index(x)]
def reverseComplement(seq):
return "".join(map(complement, reversed(seq)))
def isRepeatedCodons(seq, codons):
r = range(0, len(seq), 3)
return all(seq[i:i+3] in codons for i in r)
def isRepeatedCodonsInAnyFrame3(seq, codons):
return any(isRepeatedCodons(seq[i:] + seq[:i], codons) for i in range(3))
def isRepeatedCodonsInAnyFrame6(seq, codons):
if isRepeatedCodonsInAnyFrame3(seq, codons):
return True
codons = [reverseComplement(i) for i in codons]
return isRepeatedCodonsInAnyFrame3(seq, codons)
def isBadCodons(seq):
glnCodons = "CAA", "CAG"
alaCodons = "GCA", "GCC", "GCG", "GCT"
badCodonSets = glnCodons, alaCodons
return any(isRepeatedCodonsInAnyFrame6(seq, i) for i in badCodonSets)
def allDatasetFields(numOfDatasets, hasAllAlleles, group):
numOfFields = 4 if hasAllAlleles else 2
fieldsEnd = len(group[0]) - 1
fieldsRange = range(fieldsEnd - numOfFields, fieldsEnd)
i = 0
for j in range(numOfDatasets):
if i < len(group) and group[i][0] == j:
for k in fieldsRange:
yield group[i][k]
i += 1
if i < len(group) and group[i][0] == j:
raise RuntimeError("duplicate repeat in dataset " + str(j))
else:
for k in fieldsRange:
yield "."
def coverageFromText(text):
return 0 if text == "." else text.count(",") + 1
def numOfOutliersToIgnore(dataset):
total = count = 0
for i in dataset:
coverage = coverageFromText(i[-2]) + coverageFromText(i[-1])
total += coverage
count += (coverage > 0)
return 1 if total >= 3 * count else 0
def copyNumberChangesFromText(record, fieldNum):
text = record[fieldNum]
if text != ".":
for i in text.split(","):
if ":" in i:
i = i[:i.index(":")]
yield int(i), fieldNum
def representativeCopyNumberChange(record, numOfOutliers):
c = sorted(itertools.chain(copyNumberChangesFromText(record, -2),
copyNumberChangesFromText(record, -1)))
for i in range(numOfOutliers):
if c and c[-1][0] > 0:
c.pop() # discard the most extreme expansion
if c and c[0][0] < 0:
c.pop(0) # discard the most extreme contraction
if c:
return c[-1] if c[-1][0] >= -c[0][0] else c[0]
return None, None
def average(x):
y = list(x)
return 1.0 * sum(y) / max(len(y), 1)
def cbrt(x): # xxx a tiny bit inaccurate
if x >= 0:
return x ** (1.0 / 3)
return -((-x) ** (1.0 / 3))
def cubicMean(x):
y = [i for i in x if i is not None]
if len(y) == 1: # avoid tiny but confusing inaccuracy of cbrt
return 1.0 * y[0]
return cbrt(average(i ** 3 for i in y))
def ordinaryMean(x):
return average(i for i in x if i is not None)
def scoredJoinedLines(opts, args):
repeatLengthBoost = 30 # xxx ???
partScores = {"coding": 50, "5'UTR": 20, "3'UTR": 20, "exon": 15,
"promoter": 15, "intron": 5}
if opts.scores:
partScores.update(genePartScoresFromLines(myOpen(opts.scores)))
hasAllAlleles = True
records = []
datasetCount = 0
numOfPositiveDatasets = -1
for arg in args:
if arg == ":":
numOfPositiveDatasets = datasetCount
continue
datasetsPerChunk = 0
hasAlleles = False
for line in myOpen(arg):
fields = line.split()
if line[0] == "#":
datasetCount += datasetsPerChunk
datasetsPerChunk = 0
hasAlleles = len(fields) > 1 and fields[1][-1] == "2"
elif fields:
if not hasAlleles: hasAllAlleles = False
fieldsPerDataset = 4 if hasAlleles else 2
dpc, mod = divmod(len(fields) - 6, fieldsPerDataset)
if dpc < 0 or mod > 0 or datasetsPerChunk not in (0, dpc):
raise RuntimeError("bad data in file: " + arg)
datasetsPerChunk = dpc
for i in range(dpc):
j = 6 + i * fieldsPerDataset
k = j + fieldsPerDataset
r = [datasetCount + i] + fields[:6] + fields[j:k]
records.append(r)
datasetCount += datasetsPerChunk
if numOfPositiveDatasets == -1:
numOfPositiveDatasets = datasetCount
records.sort()
outliers = [numOfOutliersToIgnore(v)
for k, v in itertools.groupby(records, operator.itemgetter(0))]
for r in records:
rep, fieldNum = representativeCopyNumberChange(r, outliers[r[0]])
if opts.shrink:
r[-2] = r[-1] = "."
if rep is not None:
r[fieldNum] = str(rep)
if opts.change != 0 and rep is not None:
if len(r) < 11:
raise RuntimeError("input lacks alleles")
alleles = r[7:9]
if "." in alleles:
if opts.change == 2:
rep = None
else:
func = max if opts.change == 1 else min
chosenAllele = func(map(int, alleles), key=abs)
rep = min(rep, chosenAllele, key=abs)
r.append(rep)
prog = "tandem-genotypes-join" + "2" * hasAllAlleles
print("#", prog, *sys.argv[1:])
records.sort(key=operator.itemgetter(1, 2, 3, 4)) # stable
for k, v in itertools.groupby(records, operator.itemgetter(1, 2, 3, 4)):
chrom, beg, end, unit = k
group = list(v)
if opts.shrink and all(i[-1] is None for i in group):
continue
geneName = genePart = "."
if opts.shrink < 2:
geneName, genePart = group[0][5:7]
dataFields = list(allDatasetFields(datasetCount, hasAllAlleles, group))
posMean = cubicMean if opts.mean == 3 else ordinaryMean
pos = posMean(i[-1] for i in group if i[0] < numOfPositiveDatasets)
neg = cubicMean(i[-1] for i in group if i[0] >= numOfPositiveDatasets)
if pos >= 0:
diff = max(pos - max(neg, 0), 0)
else:
diff = max(-pos - max(-neg, 0), 0)
geneScore = partScores.get(genePart, 1)
if genePart == "coding" and isBadCodons(unit):
geneScore *= 2
mul = geneScore * len(unit)
jointScore = mul * diff / (int(end) - int(beg) + repeatLengthBoost)
fields = [chrom, beg, end, unit, geneName, genePart] + dataFields
yield -jointScore, fields, (geneScore, pos, neg)
def tandemGenotypesJoin(opts, args):
for jointScore, fields, info in sorted(scoredJoinedLines(opts, args)):
if opts.verbose:
t = "{0:.3g}\t{1}\t{2:.9g}\t{3:.9g}".format(-jointScore, *info)
print(t, *fields, sep="\t")
else:
print(*fields, sep="\t")
if __name__ == "__main__":
signal.signal(signal.SIGPIPE, signal.SIG_DFL) # avoid silly error message
usage = "%prog positive-file(s) [: negative-file(s)]"
description = "Join and re-rank outputs of tandem-genotypes."
op = optparse.OptionParser(usage=usage, description=description)
op.add_option("-c", "--change", type="int", default=0, metavar="N", help=
"get importance from: most-changed non-outlier read (0), "
"most-changed allele (1), least-changed allele (2) "
"(default=%default)")
op.add_option("-s", "--shrink", action="count", default=0,
help="shrink the output")
op.add_option("-m", "--mean", type="int", default=1, metavar="NUM",
help="type of mean for positive/patient/case files: "
"1=ordinary, 3=cubic (default=%default)")
op.add_option("--scores", metavar="FILE",
help="importance scores for gene parts")
op.add_option("-v", "--verbose", action="count", default=0,
help="show more details")
opts, args = op.parse_args()
if args:
try:
tandemGenotypesJoin(opts, args)
except RuntimeError as e:
prog = os.path.basename(sys.argv[0])
sys.exit(prog + ": error: " + str(e))
else:
op.print_help()