def test_to_cigar_unit(self, length, opstr, expected):
# Check we can convert a tuple and list of length and opstr.
self.assertEqual(cigar.to_cigar_unit((length, opstr)), expected)
self.assertEqual(cigar.to_cigar_unit([length, opstr]), expected)
# Check that we can convert a string version len+opstr.
self.assertEqual(cigar.to_cigar_unit(str(length) + opstr), expected)
# Check that we can pass a CigarUnit through without modification.
self.assertEqual(cigar.to_cigar_unit(expected), expected)
# Check that we can pass length as a string.
self.assertEqual(cigar.to_cigar_unit((str(length), opstr)), expected)
def test_to_cigar_unit(self, length, opstr, expected):
# Check we can convert a tuple and list of length and opstr.
self.assertEqual(cigar.to_cigar_unit((length, opstr)), expected)
self.assertEqual(cigar.to_cigar_unit([length, opstr]), expected)
# Check that we can convert a string version len+opstr.
self.assertEqual(cigar.to_cigar_unit(str(length) + opstr), expected)
# Check that we can pass a CigarUnit through without modification.
self.assertEqual(cigar.to_cigar_unit(expected), expected)
# Check that we can pass length as a string.
self.assertEqual(cigar.to_cigar_unit((str(length), opstr)), expected)
def test_no_bad_soft_clipping(self):
self.skipTest('Enable when b/63143285 global alignment is fixed')
common = 'CTA'
read_seq = common + 'GA'
ref_seq = 'N' + common + 'CA' + 'N'
alt_seq = 'A' + ref_seq
targets = [ref_seq, alt_seq]
region = ranges.make_range('ref', 0, len(ref_seq))
align_reads = self.make_test_aligner(ref_seq, region)
read = test_utils.make_read(
read_seq,
chrom='ref',
start=0,
cigar=[(len(read_seq), 'M')],
quals=[35] * len(read_seq),
name='read')
realigned = align_reads.align_reads(targets, [read])[0]
# redacted
# 5M as we'd expect for this read:
# read_seq: -CTAGA-
# ref_seq : NCGTCAN
# But the current algorithm produces a local alignment of the read against
# the haplotypes, and the G <=> C mismatch causes the local aligner to
# simply skip those bases instead of incurring the mismatch penalty for it,
# resulting in a 3M2S read (GA clipped off) instead of the better 5M result.
self.assertEqual([_cigar.to_cigar_unit(len(read_seq), 'M')],
list(realigned.alignment.cigar))
def test_no_bad_soft_clipping(self):
self.skipTest('Enable when b/63143285 global alignment is fixed')
common = 'CTA'
read_seq = common + 'GA'
ref_seq = 'N' + common + 'CA' + 'N'
alt_seq = 'A' + ref_seq
targets = [ref_seq, alt_seq]
region = ranges.make_range('ref', 0, len(ref_seq))
align_reads = self.make_test_aligner(ref_seq, region)
read = test_utils.make_read(
read_seq,
chrom='ref',
start=0,
cigar=[(len(read_seq), 'M')],
quals=[35] * len(read_seq),
name='read')
realigned = align_reads.align_reads(targets, [read])[0]
# redacted
# 5M as we'd expect for this read:
# read_seq: -CTAGA-
# ref_seq : NCGTCAN
# But the current algorithm produces a local alignment of the read against
# the haplotypes, and the G <=> C mismatch causes the local aligner to
# simply skip those bases instead of incurring the mismatch penalty for it,
# resulting in a 3M2S read (GA clipped off) instead of the better 5M result.
self.assertEqual([_cigar.to_cigar_unit(len(read_seq), 'M')],
list(realigned.alignment.cigar))
def test_to_cigar_unit_detects_bad_args(self, bad):
with self.assertRaises(ValueError):
cigar.to_cigar_unit(bad)
def test_to_cigar_unit_detects_bad_args(self, bad):
with self.assertRaises(ValueError):
cigar.to_cigar_unit(bad)
