def test_002_check_split(self):
self.split_fps.extend(split_variants(self.vcf_merged))
# 1-based positions on either position to exclude
# these are where two variants have been merged and cannot be easily
# separated without alignment, or where an indel has been rolled
# forwards due to the merge and splitting apart process.
expt_excluded = [
{675, 677, 1582, 1734, 1775}, # hap 1
{370, 1194}, # hap 2
]
for expt_vcf, got_vcf, excluded in zip([self.vcf1, self.vcf2],
self.split_fps, expt_excluded):
expt_vcfr = VCFReader(expt_vcf)
got_vcfr = VCFReader(got_vcf)
for expt in expt_vcfr.fetch():
if expt.pos + 1 in excluded:
continue
got = list(
got_vcfr.fetch(expt.chrom, expt.pos,
expt.pos + len(expt.ref) + 1))
self.assertEqual(
len(got), 1,
'Could not find split variant for {}:{}.'.format(
expt.chrom, expt.pos + 1))
got = got[0]
for key in ('chrom', 'pos', 'ref', 'alt'):
expected = getattr(expt, key)
result = getattr(got, key)
self.assertEqual(
expected, result,
'Splitting failed for {}:{} {}.'.format(
expt.chrom, expt.pos + 1, key))
def merge_haploid_vcfs(vcf1, vcf2, vcf_out):
"Merge SNPs from two haploid VCFs into an unphased diploid vcf."
loci_by_chrom = defaultdict(set)
vcf1 = VCFReader(vcf1)
vcf2 = VCFReader(vcf2)
for v in chain(vcf1.fetch(), vcf2.fetch()):
loci_by_chrom[v.chrom].add(v.pos)
with VCFWriter(vcf_out, 'w', version='4.1') as vcf_writer:
for chrom, loci in loci_by_chrom.items():
for pos in sorted(loci):
v1 = list(vcf1.fetch(ref_name=chrom, start=pos, end=pos+1))
v2 = list(vcf2.fetch(ref_name=chrom, start=pos, end=pos+1))
# the QC is -10*log10(1-p(label)) where p(label) is the medaka consensus
# probability. To combine these, we probably want to multiply the
# (1-p(label)) values, i.e. add the QC scores. However, in the case of a
# herterozygous SNPs where one of the haplotypes is the reference, we
# won't have the QC value of the reference haplotype (no variant was
# called).
# Hence if we want a common scale we need to assume we can apprimate the missing
# QC score for the reference haplotypes as being equal to the non-reference
# haplotype so we can set the overall score to double the latter.
def get_gq(v1, v2):
if len(v1) == 1 and len(v2) == 1:
gq = float(v1[0].sample_dict['GQ']) + float(v2[0].sample_dict['GQ'])
else:
v = v1[0] if len(v1) == 1 else v2[0]
gq = 2 * float(v.sample_dict['GQ'])
return gq
def get_ref(v1, v2):
return v1[0].ref if len(v1) == 1 else v2[0].ref
# Note we output unphased GTs as we might have multiple phased
# regions and the phase can switch between regions
# heterozygous on v1:
if len(v1) == 1 and (len(v2) == 0 or v2[0].alt == ['.']):
alt = v1[0].alt
gt = '0/1' # not 1/0 by convention since this is unphased
# heterozygous on v2
elif (len(v1) == 0 or v1[0].alt == ['.']) and len(v2) == 1:
alt = v2[0].alt
gt = '0/1'
else:
assert len(v1) == 1 and len(v2) == 1
if v1[0].alt == v2[0].alt: #homozygous snp
alt = v1[0].alt
gt = '1/1'
else: #heterozygous snp
alt = v1[0].alt + v2[0].alt
gt = '1/2'
gq = get_gq(v1, v2)
v = Variant(chrom, pos, get_ref(v1, v2), alt=alt, qual=gq, sample_dict={'GT':gt, 'GQ':gq})
vcf_writer.write_variant(v)
def test_vcf_annotate(self):
variants_annotated = [
Variant('MN908947.3', 29748, 'ACGATCGAGTG', alt=['A'],
ident='.', qual=243.965, filt='PASS',
info='AR=0,0;DP=200;DPS=100,100;DPSP=199;SC=19484,20327,22036,23215;SR=1,2,98,98',
genotype_data=OrderedDict([('GT','1'), ('GQ', '244')])),
Variant('MN908947.3', 29764, 'TGAACAATGCT',
alt=['A'], ident='.', qual=243.965, filt='PASS',
info='AR=0,0;DP=200;DPS=100,100;DPSP=199;SC=19970,21140,15773,16751;SR=99,100,0,0',
genotype_data=OrderedDict([('GT','1'), ('GQ', '244')])),
Variant('MN908947.3', 29788, 'TATATGGAAGA',
alt=['A'], ident='.', qual=243.965, filt='PASS',
info='AR=0,0;DP=199;DPS=99,100;DPSP=197;SC=26174,28129,19085,20315;SR=96,100,1,0',
genotype_data=OrderedDict([('GT', '1'), ('GQ','244')]))]
variants_annotated = variants_annotated + deepcopy(variants_annotated)
for i in range(3, 6):
variants_annotated[i].chrom = "Duplicate"
with tempfile.NamedTemporaryFile() as vcfout:
# Annotate vcf
args = Namespace(RG=self.rg, vcf=self.vcf,ref_fasta=self.ref_fasta,
bam=self.bam, vcfout=vcfout.name,
chunk_size=100000, pad=25, dpsp=True)
annotate_vcf_n_reads(args)
# Read in output variants and compare with expected annotated variants
vcf_reader = VCFReader(vcfout.name)
for i, v in enumerate(vcf_reader.fetch()):
self.assertEqual(v, variants_annotated[i],
'Annotation failed for variant {}: {} {}.'.format(i, v.chrom, v.pos))