def test_cmp_vrt_iter_vrt2(self):
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf'))
vs2 = VariantSetFromFile(
pkg_file('genomvar.test', 'data/example_gnomad_1.vcf.gz'))
vrt = list(vs1.iter_vrt())
self.assertEqual(len(list(vs1.diff_vrt(vs2).iter_vrt())), len(vrt))
def test_cmp_vrt_iter_same(self):
vs = VariantSetFromFile(
pkg_file('genomvar.test', 'data/example2.vcf.gz'))
tot = list(vs.find_vrt())
# print(tot)
comm = list(vs.comm_vrt(vs).iter_vrt())
self.assertEqual(len(comm), len(tot))
def test_wrong_chrom_name_in_ref(self):
ref = Reference(pkg_file(__name__, 'data/chr25.fasta'))
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
reference=ref,
index=True)
self.assertEqual(len(list(vset.find_vrt(rgn='chr24:1200-1210'))), 2)
ref.close()
def test_diff_callback(self):
s1 = VariantSetFromFile(pkg_file('genomvar.test', 'data/example3.vcf'))
s2 = VariantSetFromFile(pkg_file('genomvar.test', 'data/example3.vcf'))
cb = lambda m: [v.attrib['vcf_notation']['row'] for v in m]
for N, vrt in enumerate(s1.comm_vrt(s2).iter_vrt(callback=cb)):
self.assertEqual(vrt.attrib['vcf_notation']['row'],
vrt.attrib['cmp'][0])
self.assertEqual(N, 7)
def test_cmp_stream(self):
s1 = VariantSetFromFile(pkg_file('genomvar.test', 'data/example1.vcf'))
s2 = VariantSetFromFile(
pkg_file('genomvar.test', 'data/example2.vcf.gz'))
nofv = 0
for vrt in s1.diff_vrt(s2).iter_vrt():
nofv += vrt.nof_unit_vrt()
self.assertEqual(nofv, 14)
def test_index_errors(self):
file = pkg_file('genomvar.test', 'data/example1.vcf')
with self.assertRaises(NoIndexFoundError):
vset = VariantSetFromFile(file, reference=self.chr24, index=True)
vset = VariantSetFromFile(file, reference=self.chr24)
with self.assertRaises(ValueError) as cm:
list(vset.find_vrt('chr1', 1, 100))
error = cm.exception
self.assertIn('index is required', error.args[0].lower())
def test_find_vrt(self):
ivfs = VariantSetFromFile(pkg_file('genomvar.test',
'data/example2.vcf.gz'),
index=True)
vs = VariantSet.from_vcf(
pkg_file('genomvar.test', 'data/example2.vcf.gz'))
self.assertEqual(
sum([v.nof_unit_vrt() for v in ivfs.find_vrt('chr24')]),
sum([v.nof_unit_vrt() for v in vs.find_vrt('chr24')]))
self.assertEqual(sum([v.nof_unit_vrt() for v in ivfs.iter_vrt()]),
sum([v.nof_unit_vrt() for v in vs.iter_vrt()]))
def test_class(self):
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
parse_info=True,
reference=self.chr24,
parse_samples='SAMP1')
# Test find_vrt and returned INFO
vrt = list(vset.find_vrt('chr24', 1200, 1210))
self.assertEqual(len(vrt), 2)
v1, v2 = vrt
self.assertEqual(v1.attrib['info']['NSV'], 1)
self.assertEqual(v2.attrib['info']['RECN'], 19)
# Test multiallelic
vrt = list(vset.find_vrt('chr24', 20, 30))
self.assertEqual(len(vrt), 2)
v1, v2 = vrt
self.assertEqual(v1.attrib['info']['AF'], 0.5)
self.assertEqual(v2.attrib['info']['AF'], 0.5)
# Test find_vrt precision
vrt = list(vset.find_vrt('chr24', 2095, 2096))
self.assertEqual(len(vrt), 1)
vrt = list(vset.find_vrt('chr24', 2098, 2100))
self.assertEqual(len(vrt), 1)
# Test find all variants
self.assertEqual(len(list(vset.find_vrt())), 16)
# Test finding all variants
self.assertEqual(len(list(vset.find_vrt())), 16)
def test_complex_info_example(self):
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example_gnomad_1.vcf.gz'),
parse_info=True,
index=True)
checked = False
for vrt in vset.find_vrt(rgn='chr1:69090-69091'):
if vrt.alt != 'C':
continue
if not vrt.attrib['info']['MutPred_Top5features'] is None:
checked = True
self.assertTrue(vrt.attrib['info']['MutPred_Top5features']\
.startswith('Loss of sheet (P = 0.0817)| L'))
self.assertTrue(checked)
def test_cmp_vrt_region_multisample2(self):
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example_1000genomes_1.vcf.gz'),
parse_samples=True,
index=True)
vs2 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example_1000genomes_2.vcf.gz'),
parse_samples=True,
index=True)
comm = []
for vrt in vs2.comm_vrt(vs1).region(rgn='7:152134922-152436005'):
comm.append(vrt)
self.assertTrue(hasattr(vrt, 'attrib'), msg='False for' + str(vrt))
comm = list(vs2.comm_vrt(vs1).region(rgn='7:152134922-152436005'))
self.assertGreater(len(comm), 0)
def test_cmp_vrt_iter_vrt(self):
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
parse_samples=True)
vs2 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example2.vcf.gz'),
parse_samples=True)
comm = list()
for vrt in vs1.comm_vrt(vs2).iter_vrt():
comm.append(vrt)
self.assertTrue(vrt.attrib['samples'],
msg='Vrt {} has no samples'.format(vrt))
self.assertEqual(len(comm), 4)
diff = vs1.diff_vrt(vs2).iter_vrt()
self.assertEqual(len(list(diff)), 12)
def test_match(self):
# REF TGG TT
# 2093 2099
# vs1 CCC GG
# vrt CG
# r1 r2,r3
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
index=True)
vrt = factory.from_edit('chr24', 2098, 'TT', 'CG')
self.assertEqual(len(vs1.match(vrt)), 1)
# Test insertion
vrt = factory.from_edit('chr24', 22, 'AG', 'AGG')
match = vs1.match(vrt)
self.assertEqual(len(match), 1)
def test_cmp_vrt_region(self):
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
parse_samples=True,
parse_info=True,
index=True)
vs2 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example2.vcf.gz'),
parse_samples='SAMP1',
parse_info=True,
index=True)
comm = list(vs1.comm_vrt(vs2).region(rgn='chr24:10040-10050'))
self.assertEqual(len(comm), 2)
v1, v2 = comm
self.assertEqual(v1.attrib['info']['AF'], 1.0)
self.assertEqual(v1.attrib['samples']['SAMP1']['GT'], (0, 1))
def test_from_variants(self):
vfset = VariantSetFromFile(
pkg_file('genomvar.test', 'data/example1.vcf'))
vset = VariantSet.from_variants(list(vfset.iter_vrt()))
vrt = list(vset.find_vrt('chr24', 1200, 1210))
self.assertEqual(len(vrt), 2)
# Test error on out of reference bounds
with self.assertRaises(ValueError):
VariantSet.from_variants(list(vfset.iter_vrt()) +
[variant.SNP('chr24', 10000000, 'C')],
reference=self.chr24)
# Test error on chromosome not in reference
with self.assertRaises(ValueError):
vs = VariantSet.from_variants(list(vfset.iter_vrt()) +
[variant.SNP('chr2', 10, 'C')],
reference=self.chr24)
def test_unsorted_VCF_input(self):
header = []
lines = []
with open(pkg_file('genomvar.test', 'data/example1.vcf'), 'rt') as fh:
for line in fh:
if line.startswith('#'):
header.append(line)
else:
lines.append(line)
tf = tempfile.NamedTemporaryFile(suffix='.vcf')
with open(tf.name, 'wt') as fh:
fh.writelines(header)
fh.writelines(reversed(lines))
vs1 = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf'))
vs2 = VariantSetFromFile(tf.name)
with self.assertRaises(UnsortedVariantFileError):
list(vs1.diff_vrt(vs2).iter_vrt())
def test_differently_sorted_chroms(self):
s1 = VariantSetFromFile(pkg_file('genomvar.test', 'data/example3.vcf'))
header = []
variants = {}
with open(pkg_file('genomvar.test', 'data/example3.vcf')) as fh:
for line in fh:
if line.startswith('#'):
header.append(line)
else:
variants.setdefault(line.split(maxsplit=1)[0],[])\
.append(line)
tf = tempfile.NamedTemporaryFile(suffix='.vcf')
with open(tf.name, 'wt') as fh:
fh.writelines(header)
for chrom in ['chr1', 'chr10', 'chr2']:
fh.writelines(variants[chrom])
s2 = VariantSetFromFile(tf.name)
with self.assertRaises(DifferentlySortedChromsError):
list(s1.diff_vrt(s2).iter_vrt())
def test_find_vrt2(self):
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
reference=self.chr24,
index=True)
self.assertEqual(len(list(vset.find_vrt(rgn='chr24:1200-1210'))), 2)
v1, v2 = list(vset.find_vrt(rgn='chr24:1200-1210'))
self.assertEqual([v1.start, v1.end], [1206, 1207])
self.assertEqual([v2.start, v2.end], [1206, 1207])
self.assertEqual(len(list(vset.find_vrt(rgn='chr24:3200-3205'))), 1)
v1 = list(vset.find_vrt(rgn='chr24:3200-3205'))[0]
self.assertEqual([v1.start, v1.end], [3201, 3202])
self.assertEqual(len(list(vset.find_vrt(rgn='chr24:20-30'))), 2)
v1, v2 = list(vset.find_vrt(rgn='chr24:20-30'))
self.assertEqual([v1.start, v1.end, type(v1.base)],
[23, 24, variant.Del])
self.assertEqual([v2.start, v2.end, type(v2.base)],
[24, 25, variant.Ins])
def _cmp_vcf(f1, f2, out, match_partial=False, chunk=1000):
"""
Writes comparison of two VCF files to a specified file handle.
"""
info = [('VT', 1, 'String', 'Variant type'),
('whichVCF', 1, 'String',
'Which input VCF contains the variant; first, second or both'),
('ln', 1, 'Integer',
'Line number in input VCF variant originating from'),
('ln2', '.', 'Integer',
'If whichVCF is both indicates line numberin the second file')]
writer = VCFWriter(info_spec=info)
header = writer.get_header()
out.write(header)
if _isindexed(f1):
vs1 = VariantSetFromFile(f1)
else:
warnings.warn('{} not indexed; may impact performance.'.format(f1))
vs1 = VariantSetFromFile(f1)
if _isindexed(f2):
vs2 = VariantSetFromFile(f2)
else:
warnings.warn('{} not indexed; may impact performance.'.format(f2))
vs2 = VariantSetFromFile(f2)
_which = {0: 'first', 1: 'second', 2: 'both'}
nof_vrt = {i: 0 for i in _which}
cb = lambda m: [v.attrib['vcf_notation']['row'] for v in m]
for which,vrt in vs1._cmp_vrt(vs2,action='all')\
.iter_vrt(callback=cb):
nof_vrt[which] += vrt.nof_unit_vrt()
if which == 0:
lineno = vrt.attrib['vcf_notation'][
'row'] + vs1.vcfreader.header_len + 1
elif which == 1:
lineno = vrt.attrib['vcf_notation'][
'row'] + vs2.vcfreader.header_len + 1
if which == 2:
lineno = vrt.attrib['vcf_notation'][
'row'] + vs1.vcfreader.header_len + 1
lineno2 = [
vs2.vcfreader.header_len + n + 1 for n in vrt.attrib['cmp']
]
vrt.attrib['info'] = {'whichVCF':_which[which],
'ln':lineno,
'ln2':lineno2 if which==2\
else None}
try:
row = writer.get_row(vrt)
except ValueError as exc:
if vrt.is_variant_instance(variant.Haplotype) \
or vrt.is_variant_instance(variant.Asterisk):
continue
else:
raise exc
try:
out.write(str(row) + '\n')
except BrokenPipeError:
exit(1)
return nof_vrt
def test_no_index(self):
with self.assertRaises(NoIndexFoundError):
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example3.vcf'),
index=True)
def test_find_nonexistent_chrom(self):
vcf = pkg_file('genomvar.test', 'data/example_1000genomes_1.vcf.gz')
vset = VariantSetFromFile(vcf, index=True)
self.assertEqual(list(vset.find_vrt('chr24')), [])
def test_ctg_len_without_ref(self):
vset = VariantSetFromFile(pkg_file('genomvar.test',
'data/example1.vcf.gz'),
parse_samples='SAMP1',
index=True)
self.assertEqual(vset.chroms, {'chr24'})