def _smooth(self, snp_idx_to_smooth, snp_gts, samples):
snps, gt_for_snps_in_win = zip(*snp_gts)
# we need the recomb rates
# TODO, this is already calculated in ab coding, we have to cache
snp1 = snps[snp_idx_to_smooth]
snp1_calls = [snp1.genotype(sample) for sample in samples]
weights = []
for snp2 in snps:
snp2_calls = [snp2.genotype(sample) for sample in samples]
haplos = _count_biallelic_haplotypes(snp1_calls, snp2_calls,
return_alleles=True)
if haplos is None:
weight = 0
else:
haplo_cnt, alleles = haplos
recomb_rate = (haplo_cnt.aB + haplo_cnt.Ab) / sum(haplo_cnt)
weight = 2 * (0.5 - recomb_rate) if recomb_rate < 0.5 else 0
weights.append(weight)
# we have to transpose, we want the genotype for each indi not for
# each snp
indis_gts = {indi: [] for indi in samples}
for indi in samples:
for snp_gt in gt_for_snps_in_win:
snp_indi_gt = snp_gt.get(indi, (indi, None))
snp_indi_gt = tuple(sorted(snp_indi_gt))
indis_gts[indi].append(snp_indi_gt)
# Now we can do the smoothing
recomb_thres = self.recomb_threshold
smooth_threhsold = self.smooth_threhsold
smoothed_genos = []
for indi in samples:
indi_gt = indis_gts[indi]
n_recombs = self._count_recombinations(indi_gt)
counts = Counter()
for weight, geno in zip(weights, indi_gt):
counts[geno] += weight
smoothed_geno, vote = counts.most_common(1)[0]
index = vote / sum(counts.values())
self._recombs.append(n_recombs)
self._smoothes.append(index)
if recomb_thres is None:
if index > smooth_threhsold:
geno = smoothed_geno
else:
geno = None
else:
if n_recombs > recomb_thres:
# We're assuming diploid here
geno = ('A', 'B')
else:
if index > smooth_threhsold:
geno = smoothed_geno
else:
geno = None
smoothed_genos.append(geno)
return smoothed_genos
def test_empy_snv(self):
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t./.\t./.\t./.\t./.\t./.\t./.
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts is None
def test_empy_snv(self):
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t./.\t./.\t./.\t./.\t./.\t./.
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts is None
def _deduce_coding(self, snp_and_coding, snp1_calls, snp2_idxs):
votes = Counter()
offspring = self.offspring
for snp2_idx in snp2_idxs:
snp2, coding2 = snp_and_coding[snp2_idx]
if coding2 is None:
continue
snp2_calls = [snp2.genotype(sample) for sample in offspring]
haplos = _count_biallelic_haplotypes(snp1_calls,
snp2_calls,
return_alleles=True)
if haplos is None:
continue
else:
haplo_cnt, alleles = haplos
alleles_in_major_haplo = {
alleles.b: alleles.a,
alleles.B: alleles.A
}
if haplo_cnt is None:
continue
if (coding2.A not in alleles_in_major_haplo
or coding2.B not in alleles_in_major_haplo):
# The offspring alleles in snp2 do not match the alleles
# in the parents
continue
# This is allele A in snp 1
allele1A = alleles_in_major_haplo[coding2.A]
# This is allele B in snp 1
allele1B = alleles_in_major_haplo[coding2.B]
voted_coding1 = AlleleCoding(allele1A, allele1B)
recomb_rate = (haplo_cnt.aB + haplo_cnt.Ab) / sum(haplo_cnt)
weight = 2 * (0.5 - recomb_rate) if recomb_rate < 0.5 else 0
votes[voted_coding1] += weight
if not votes or sum(votes.values()) == 0:
deduced_coding1 = None
self.log[NO_INFO] += 1
elif len(votes) > 2:
deduced_coding1 = None
self.log[MORE_THAN_2_CODINGS] += 1
else:
deduced_coding1 = votes.most_common(1)[0][0]
index = votes[deduced_coding1] / sum(votes.values())
self.indexes.append(index)
if index < self.parent_index_threshold:
deduced_coding1 = None
self.log[NOT_ENOUGH_SUPPORT] += 1
else:
self.log[ENOUGH_SUPPORT] += 1
if deduced_coding1 is None:
return None
return {deduced_coding1.A: 'A', deduced_coding1.B: 'B'}
def _deduce_coding(self, snp_and_coding, snp1_calls, snp2_idxs):
votes = Counter()
offspring = self.offspring
for snp2_idx in snp2_idxs:
snp2, coding2 = snp_and_coding[snp2_idx]
if coding2 is None:
continue
snp2_calls = [snp2.genotype(sample) for sample in offspring]
haplos = _count_biallelic_haplotypes(snp1_calls, snp2_calls,
return_alleles=True)
if haplos is None:
continue
else:
haplo_cnt, alleles = haplos
alleles_in_major_haplo = {alleles.b: alleles.a,
alleles.B: alleles.A}
if haplo_cnt is None:
continue
if (coding2.A not in alleles_in_major_haplo or
coding2.B not in alleles_in_major_haplo):
# The offspring alleles in snp2 do not match the alleles
# in the parents
continue
# This is allele A in snp 1
allele1A = alleles_in_major_haplo[coding2.A]
# This is allele B in snp 1
allele1B = alleles_in_major_haplo[coding2.B]
voted_coding1 = AlleleCoding(allele1A, allele1B)
recomb_rate = (haplo_cnt.aB + haplo_cnt.Ab) / sum(haplo_cnt)
weight = 2 * (0.5 - recomb_rate) if recomb_rate < 0.5 else 0
votes[voted_coding1] += weight
if not votes or sum(votes.values()) == 0:
deduced_coding1 = None
self.log[NO_INFO] += 1
elif len(votes) > 2:
deduced_coding1 = None
self.log[MORE_THAN_2_CODINGS] += 1
else:
deduced_coding1 = votes.most_common(1)[0][0]
index = votes[deduced_coding1] / sum(votes.values())
self.indexes.append(index)
if index < self.parent_index_threshold:
deduced_coding1 = None
self.log[NOT_ENOUGH_SUPPORT] += 1
else:
self.log[ENOUGH_SUPPORT] += 1
if deduced_coding1 is None:
return None
return {deduced_coding1.A: 'A', deduced_coding1.B: 'B'}
def test_count_homo_haplotypes(self):
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 3
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t2/2\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 3
r_sqr = calculate_r_sqr(snps[0], snps[1])
self.assertAlmostEqual(r_sqr, 1)
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t1/1\t1/1\t0/0\t1/1\t0/0\t1/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr - 1.0 < 0.0001
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t1/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr - 1.0 < 0.0001
# monomorphic
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
assert _count_biallelic_haplotypes(call1, call2) is None
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr is None
assert fisher_exact(snps[0], snps[1]) is None
# Ab and aB
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t0/0\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t0/0\t1/1\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 1
# different major allele names in snp1 (1, 2) and snp2 (2,3)
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t0/0\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t3/3\t3/3\t3/3\t2/2\t2/2\t3/3\t3/3\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 0
# missing data
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t./.\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t3/3\t3/3\t3/3\t2/2\t2/2\t3/3\t3/3\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 0
def test_count_homo_haplotypes(self):
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 3
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t2/2\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 3
r_sqr = calculate_r_sqr(snps[0], snps[1])
self.assertAlmostEqual(r_sqr, 1)
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t1/1\t1/1\t0/0\t1/1\t0/0\t1/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr - 1.0 < 0.0001
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t0/0
20\t14\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t1/1\t1/1\t0/0\t0/0\t0/1\t1/1\t1/1'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr - 1.0 < 0.0001
# monomorphic
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7 8
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0\t0/0'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
assert _count_biallelic_haplotypes(call1, call2) is None
r_sqr = calculate_r_sqr(snps[0], snps[1])
assert r_sqr is None
assert fisher_exact(snps[0], snps[1]) is None
# Ab and aB
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t0/0\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t0/0\t1/1\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 1
# different major allele names in snp1 (1, 2) and snp2 (2,3)
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t0/0\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t3/3\t3/3\t3/3\t2/2\t2/2\t3/3\t3/3\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 4
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 0
# missing data
vcf = '''#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT 1 2 3 4 5 6 7
20\t2\t.\tG\tA\t29\tPASS\tNS=3\tGT\t0/0\t0/0\t0/0\t1/1\t1/1\t1/1\t./.\t
20\t3\t.\tG\tA\t29\tPASS\tNS=3\tGT\t3/3\t3/3\t3/3\t2/2\t2/2\t3/3\t3/3\t'''
vcf = StringIO(VCF_HEADER + vcf)
snps = list(VCFReader(vcf).parse_snvs())
call1 = snps[0].record.samples
call2 = snps[1].record.samples
counts = _count_biallelic_haplotypes(call1, call2)
assert counts.AB == 3
assert counts.ab == 2
assert counts.aB == 1
assert counts.Ab == 0
def _smooth(self, snp_idx_to_smooth, snp_gts, samples):
snps, gt_for_snps_in_win = zip(*snp_gts)
# we need the recomb rates
# TODO, this is already calculated in ab coding, we have to cache
snp1 = snps[snp_idx_to_smooth]
snp1_calls = [snp1.genotype(sample) for sample in samples]
weights = []
for snp2 in snps:
snp2_calls = [snp2.genotype(sample) for sample in samples]
haplos = _count_biallelic_haplotypes(snp1_calls,
snp2_calls,
return_alleles=True)
if haplos is None:
weight = 0
else:
haplo_cnt, alleles = haplos
recomb_rate = (haplo_cnt.aB + haplo_cnt.Ab) / sum(haplo_cnt)
weight = 2 * (0.5 - recomb_rate) if recomb_rate < 0.5 else 0
weights.append(weight)
# we have to transpose, we want the genotype for each indi not for
# each snp
indis_gts = {indi: [] for indi in samples}
for indi in samples:
for snp_gt in gt_for_snps_in_win:
snp_indi_gt = snp_gt.get(indi, (indi, None))
snp_indi_gt = tuple(sorted(snp_indi_gt))
indis_gts[indi].append(snp_indi_gt)
# Now we can do the smoothing
recomb_thres = self.recomb_threshold
smooth_threhsold = self.smooth_threhsold
smoothed_genos = []
for indi in samples:
indi_gt = indis_gts[indi]
n_recombs = self._count_recombinations(indi_gt)
counts = Counter()
for weight, geno in zip(weights, indi_gt):
counts[geno] += weight
smoothed_geno, vote = counts.most_common(1)[0]
index = vote / sum(counts.values())
self._recombs.append(n_recombs)
self._smoothes.append(index)
if recomb_thres is None:
if index > smooth_threhsold:
geno = smoothed_geno
else:
geno = None
else:
if n_recombs > recomb_thres:
# We're assuming diploid here
geno = ('A', 'B')
else:
if index > smooth_threhsold:
geno = smoothed_geno
else:
geno = None
smoothed_genos.append(geno)
return smoothed_genos
