def testCounts(counts, snp):
winningParent='?'
snpchr, snppos, snprec = snp
mat_genotype, pat_genotype, child_genotype, mat_allele, pat_allele, typ, ref, hetSNP = snprec
# first, make sure that the expected alleles are the bulk of the counts
total = counts['a']+counts['c']+counts['g']+counts['t']
a1,a2=convert(child_genotype)
if a1==a2:
allelecnts = counts[a1]
else:
allelecnts = counts[a1]+counts[a2]
both=counts[a1]+counts[a2]
sortedCounts=sorted([(counts['a'], 'a'), (counts['c'],'c'), (counts['g'], 'g'), (counts['t'], 't')], reverse=True)
majorAllele=sortedCounts[0][1]
smaller=min(counts[a1], counts[a2])
#pval=binomialDist.cdf(smaller, both, 0.5)*2 # This had problems for large sample sizes. Switched to using scipy
pval = binom.binomtest(smaller, both, 0.5) # scipy.binom_test was unstable for large counts
if float(allelecnts)/total < THRESH1:
print >>LOGFP, "WARNING %s:%d failed thresh 1 %d %d" % (snpchr, snppos, allelecnts, total)
return (WEIRD, pval, a1, a2, counts, winningParent)
# if the snp was phased
if mat_allele and pat_allele:
if mat_allele.lower()==majorAllele.lower():
winningParent='M'
elif pat_allele.lower()==majorAllele.lower():
winningParent='P'
else:
winningParent='?'
if a1!=a2:
# we expect roughly 50/50.
if pval < THRESH2:
print >>LOGFP, "NOTE %s:%d Looks interesting: failed thresh 2 %d %d %f" % (snpchr, snppos, both, smaller, pval)
print >>LOGFP, "SNPS %s/%s, COUNTS a:%d c:%d g:%d t:%d" % (a1, a2, counts['a'], counts['c'], counts['g'], counts['t'])
print >>LOGFP, "Phasing P:%s M:%s D:%s" % (pat_allele, mat_allele, snprec)
print >>LOGFP, "\n"
return (ASYMMETRIC, pval, a1, a2, counts, winningParent)
else:
return (SYMMETRIC, pval, a1, a2, counts, winningParent)
else:
return (HOMOZYGOUS, pval, a1, a2, counts, winningParent)
outwrite(
line,
weaker + ':' + str(mm_basecnts[weaker]) + ';within_thresh')
# or adjust counts in the most conservative way: add the mm counts to the weaker allele:
# all of them or until balanced with the stonger
# to make sure the count imbalance is not caused by the multimapping reads
else:
adj = min((um_basecnts[weaker] + mm_basecnts[weaker]),
max(um_basecnts[hap1_allele], um_basecnts[hap2_allele]))
diff = adj - um_basecnts[weaker]
um_basecnts[weaker] = adj
new_tot = int(um_basecnts[hap1_allele] + um_basecnts[hap2_allele])
new_ratio = float(um_basecnts[ref_allele]) / float(new_tot)
new_p_binom = binom.binomtest(um_basecnts[ref_allele], new_tot,
0.5)
sys.stdout.write('\t'.join([
chrm, ref_coord, hap1_coord, hap2_coord, ref_allele,
hap1_allele, hap2_allele,
str(int(um_basecnts['A'])),
str(int(um_basecnts['C'])),
str(int(um_basecnts['G'])),
str(int(um_basecnts['T'])),
str(int(um_basecnts['N'])),
str(new_ratio),
str(new_tot),
str(new_p_binom), cnv[:-1], weaker + ':+' + str(int(diff))
]) + '\n')
logwrite(
def __init__(self, n):
self.n = n
self.cache = [[binom.binomtest(j, i, 0.5) for j in range(i + 1)]
for i in range(n)]
def binomtest(self, a, cnt):
if cnt < self.n:
return self.cache[cnt][a]
else:
return binom.binomtest(a, cnt, 0.5)
def binomtest(self, a, cnt):
if cnt<self.n:
return self.cache[cnt][a]
else:
return binom.binomtest(a, cnt, 0.5)
def __init__(self, n):
self.n=n
self.cache=[[binom.binomtest(j, i, 0.5) for j in range(i+1)] for i in range(n)]
+ k + '\n')
rmvdhets_file.write(
'\t'.join(k.split('_') + ['vcf_vs_mpileup_alleles']) +
'\n')
continue
ref_cnt = basecnts[hetSNV_dict[k]['r_a']]
if ref_cnt > tot_cnt: # maybe a miscalled multi-allelic variant?
rmvdhets_file.write(
k.split('_')[0] + '\t' + k.split('_')[1] +
'\tref_allele:' + str(ref_cnt) + '__hap1:' +
str(basecnts[hetSNV_dict[k]['hap1_a']]) + '__hap2:' +
str(basecnts[hetSNV_dict[k]['hap2_a']]) + '\n')
continue
pbinom = binom.binomtest(ref_cnt, tot_cnt, 0.5)
sys.stdout.write('\t'.join([
k.split('_')[0],
k.split('_')[1], hetSNV_dict[k].get('hap1_pos', 'not_lifted?'),
hetSNV_dict[k].get('hap2_pos', 'not_lifted?'), hetSNV_dict[k]
['r_a'], basecnts_hap1['hap1_a'], basecnts_hap2['hap2_a'],
str(basecnts['A']),
str(basecnts['C']),
str(basecnts['G']),
str(basecnts['T']),
str(basecnts['N']),
str(float(ref_cnt) / float(tot_cnt)),
str(tot_cnt),
str(pbinom), basecnts_hap1['warning'], basecnts_hap2['warning']
]) + '\n')
with open(sys.argv[2], 'r') as in1:
for line in in1:
if line.split()[-1] in e_dict:
e_dict[line.split()[-1]]['snv_count'] += 1
if sys.argv[1] == 'uniq':
sys.stdout.write('\t'.join([
'#region', 'hap1_count', 'hap2_count', 'hap1_allele_ratio', 'p_binom',
'snv_count', 'snv_hap1_hap2_coords'
]) + '\n')
for e in e_list:
hap1_count = len(e_dict[e]['hap1'])
hap2_count = len(e_dict[e]['hap2'])
pbinom = binom.binomtest(hap1_count, hap1_count + hap2_count, 0.5)
hap1_allele_ratio = float(hap1_count) / (float(hap1_count) +
float(hap2_count))
sys.stdout.write('\t'.join([
e,
str(hap1_count),
str(hap2_count),
str(hap1_allele_ratio),
str(pbinom),
str(e_dict[e]['snv_count']), e_dict[e]['snvs']
]) + '\n')
elif sys.argv[1] == 'mmap':
sys.stdout.write('\t'.join(['#region', 'hap1_count', 'hap2_count']) + '\n')
region, weaker_hap + ':' + str(weaker_hap_mm_count) +
';within_thresh')
# or adjust counts in the most conservative way: add the mm counts to the weaker allele:
# all of them or until balanced with the stonger
# to make sure the count imbalance is not caused by the multimapping reads
else:
adj = min((weaker_hap_un_count + weaker_hap_mm_count),
max(int(hap1_count), int(hap2_count)))
diff = adj - weaker_hap_un_count
if weaker_hap == 'hap1': hap1_count = adj
else: hap2_count = adj
new_tot = int(hap1_count) + int(hap2_count)
new_hap1_allele_ratio = float(hap1_count) / float(new_tot)
new_p_binom = binom.binomtest(int(hap1_count), new_tot, 0.5)
sys.stdout.write('\t'.join([
region,
str(hap1_count),
str(hap2_count),
str(new_hap1_allele_ratio),
str(new_p_binom), snv_count, snv_hap1_hap2_coords, weaker_hap +
':+' + str(int(diff))
]) + '\n')
#logwrite(line, mm_counts_dict[region]['hap1_count'], mm_counts_dict[region]['hap2_count'], weaker_hap+':+'+str(int(diff)))
logwrite(region, mm_counts_dict[region]['hap1_count'],
mm_counts_dict[region]['hap2_count'],
weaker_hap + ':+' + str(int(diff)))