def test_k_exceeds_ref_length(self):
"""This is a regression test for b/64564513."""
# We don't allow a k >= ref length. This crashed prior to the bugfix.
ref_str = 'GATTACA'
dbg = debruijn_graph.build(ref_str, [], self.single_k_dbg_options(7))
self.assertIsNone(dbg)
dbg = debruijn_graph.build(ref_str, [], self.single_k_dbg_options(8))
self.assertIsNone(dbg)
def test_k_exceeds_ref_length(self): """This is a regression test for b/64564513.""" # We don't allow a k >= ref length. This crashed prior to the bugfix. ref_str = 'GATTACA' dbg = debruijn_graph.build(ref_str, [], self.single_k_dbg_options(7)) self.assertIsNone(dbg) dbg = debruijn_graph.build(ref_str, [], self.single_k_dbg_options(8)) self.assertIsNone(dbg)
def test_filtering_by_qual(self):
"""Test that we filter out edges containing low-quality basecalls."""
ref_str = 'GATTACA'
read_str = 'GATGTACA'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30, 30, 30, 1, 30, 30, 30, 30],
name='read')
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(2))
self.assertGraphEqual("""\
digraph G {
0[label=GA];
1[label=AT];
2[label=TT];
3[label=TA];
4[label=AC];
5[label=CA];
0->1 [label=3 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=3 color=red];
4->5 [label=3 color=red];
}
""", dbg)
def test_pruning_2(self):
"""Test that pruning removes edges not between source and sink."""
ref_str = 'GATTACA'
read_str = 'CCGATGACACC'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(3))
self.assertGraphEqual("""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
5[label=ATG];
6[label=TGA];
7[label=GAC];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
0->5 [label=2];
5->6 [label=2];
6->7 [label=2];
7->4 [label=2];
}
""", dbg)
def test_pruning_1(self):
"""Test that pruning removes a path traced by only one read."""
ref_str = 'GATTACA'
read_str = 'GATGACA'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
dbg = debruijn_graph.build(ref_str, [read], self.single_k_dbg_options(3))
self.assertGraphEqual("""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
}
""", dbg)
def call_debruijn_graph(self, windows, reads):
"""Helper function to call debruijn_graph module."""
windows_haplotypes = []
# Build and process de-Bruijn graph for each window.
sam_reader = sam.InMemorySamReader(reads)
for window in windows:
if window.end - window.start > self.config.ws_config.max_window_size:
continue
if not self.ref_reader.is_valid(window):
continue
ref = self.ref_reader.query(window)
window_reads = list(sam_reader.query(window))
with timer.Timer() as t:
graph = debruijn_graph.build(ref, window_reads, self.config.dbg_config)
graph_building_time = t.GetDuration()
if not graph:
candidate_haplotypes = [ref]
else:
candidate_haplotypes = graph.candidate_haplotypes()
if candidate_haplotypes and candidate_haplotypes != [ref]:
candidate_haplotypes_info = realigner_pb2.CandidateHaplotypes(
span=window, haplotypes=candidate_haplotypes)
windows_haplotypes.append(candidate_haplotypes_info)
self.diagnostic_logger.log_graph_metrics(
window, graph, candidate_haplotypes, graph_building_time)
return windows_haplotypes
def call_debruijn_graph(self, windows, reads):
"""Helper function to call debruijn_graph module."""
windows_haplotypes = []
# Build and process de-Bruijn graph for each window.
for window in windows:
if window.end - window.start > self.config.ws_config.max_window_size:
continue
if not self.ref_reader.is_valid_interval(window):
continue
ref = self.ref_reader.bases(window)
# redacted
dbg_reads = [
read for read in reads
if ranges.ranges_overlap(window, utils.read_range(read))
]
with timer.Timer() as t:
graph = debruijn_graph.build(ref, dbg_reads,
self.config.dbg_config)
graph_building_time = t.GetDuration()
if not graph:
candidate_haplotypes = [ref]
else:
candidate_haplotypes = graph.candidate_haplotypes()
if candidate_haplotypes and candidate_haplotypes != [ref]:
candidate_haplotypes_info = realigner_pb2.CandidateHaplotypes(
span=window, haplotypes=candidate_haplotypes)
windows_haplotypes.append(candidate_haplotypes_info)
self.diagnostic_logger.log_graph_metrics(window, graph,
candidate_haplotypes,
graph_building_time)
return windows_haplotypes
def test_pruning_1(self):
"""Test that pruning removes a path traced by only one read."""
ref_str = 'GATTACA'
read_str = 'GATGACA'
read = test_utils.make_read(read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
dbg = debruijn_graph.build(ref_str, [read],
self.single_k_dbg_options(3))
self.assertGraphEqual(
"""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
}
""", dbg)
def test_pruning_2(self):
"""Test that pruning removes edges not between source and sink."""
ref_str = 'GATTACA'
read_str = 'CCGATGACACC'
read = test_utils.make_read(read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(3))
self.assertGraphEqual(
"""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
5[label=ATG];
6[label=TGA];
7[label=GAC];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
0->5 [label=2];
5->6 [label=2];
6->7 [label=2];
7->4 [label=2];
}
""", dbg)
def test_filtering_by_qual(self):
"""Test that we filter out edges containing low-quality basecalls."""
ref_str = 'GATTACA'
read_str = 'GATGTACA'
read = test_utils.make_read(read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30, 30, 30, 1, 30, 30, 30, 30],
name='read')
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(2))
self.assertGraphEqual(
"""\
digraph G {
0[label=GA];
1[label=AT];
2[label=TT];
3[label=TA];
4[label=AC];
5[label=CA];
0->1 [label=3 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=3 color=red];
4->5 [label=3 color=red];
}
""", dbg)
def call_debruijn_graph(self, windows, reads):
"""Helper function to call debruijn_graph module."""
windows_haplotypes = []
# Build and process de-Bruijn graph for each window.
for window in windows:
if window.end - window.start > self.config.ws_config.max_window_size:
continue
if not self.ref_reader.is_valid(window):
continue
ref = self.ref_reader.query(window)
# redacted
dbg_reads = [
read for read in reads
if ranges.ranges_overlap(window, utils.read_range(read))
]
with timer.Timer() as t:
graph = debruijn_graph.build(ref, dbg_reads, self.config.dbg_config)
graph_building_time = t.GetDuration()
if not graph:
candidate_haplotypes = [ref]
else:
candidate_haplotypes = graph.candidate_haplotypes()
if candidate_haplotypes and candidate_haplotypes != [ref]:
candidate_haplotypes_info = realigner_pb2.CandidateHaplotypes(
span=window, haplotypes=candidate_haplotypes)
windows_haplotypes.append(candidate_haplotypes_info)
self.diagnostic_logger.log_graph_metrics(
window, graph, candidate_haplotypes, graph_building_time)
return windows_haplotypes
def test_adding_edges_with_bad_positions(self, bad_position,
dropped_edges):
"""Test that we filter out edges containing low-quality basecalls."""
ref_str = 'GATTACA'
read_str = 'GATTACA'
kmer_indices = {
'GA': 0,
'AT': 1,
'TT': 2,
'TA': 3,
'AC': 4,
'CA': 5,
}
def kmer_to_index_edge(kmer_edge):
k1, k2 = kmer_edge.split('->')
return '{}->{}'.format(kmer_indices[k1], kmer_indices[k2])
dropped_edges = {kmer_to_index_edge(edge) for edge in dropped_edges}
for bad_type in ['qual', 'base']:
bases = list(read_str)
quals = [30] * len(bases)
cigar = [(len(bases), 'M')]
if bad_position is not None:
if bad_type == 'qual':
quals[bad_position] = 1
elif bad_type == 'base':
bases[bad_position] = 'N'
else:
raise ValueError('Unexpected base type')
read = test_utils.make_read(''.join(bases),
start=0,
cigar=cigar,
quals=quals)
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(2))
expected_edges = '\n'.join(
'{} [label={} color=red];'.format(
edge, 1 if edge in dropped_edges else 3)
for edge in ['0->1', '1->2', '2->3', '3->4', '4->5'])
self.assertGraphEqual(
"""\
digraph G {
0[label=GA];
1[label=AT];
2[label=TT];
3[label=TA];
4[label=AC];
5[label=CA];
%s
}
""" % expected_edges, dbg)
def test_straightforward_region(self):
ref_reader = fasta.RefFastaReader(testdata.CHR20_FASTA)
bam_reader = sam.SamReader(testdata.CHR20_BAM)
region = ranges.parse_literal('chr20:10,000,000-10,000,100')
ref_seq = ref_reader.query(region)
all_reads = list(bam_reader.query(region))
dbg30 = debruijn_graph.build(ref_seq, all_reads,
self.single_k_dbg_options(30))
self.assertIsNotNone(dbg30)
self.assertEqual([ref_seq], dbg30.candidate_haplotypes())
def test_straightforward_region(self):
ref_reader = fasta.IndexedFastaReader(testdata.CHR20_FASTA)
bam_reader = sam.SamReader(testdata.CHR20_BAM)
region = ranges.parse_literal('chr20:10,000,000-10,000,100')
ref_seq = ref_reader.query(region)
all_reads = list(bam_reader.query(region))
dbg30 = debruijn_graph.build(ref_seq, all_reads,
self.single_k_dbg_options(30))
self.assertIsNotNone(dbg30)
self.assertEqual([ref_seq], dbg30.candidate_haplotypes())
def test_ref_cycle_detector(self, ref, smallest_good_k):
min_k = max(smallest_good_k - 5, 1)
max_k = min(smallest_good_k + 5, len(ref))
for k in range(min_k, max_k):
# The build fails, returning None, with a k < smallest_good_k. If
# k >= smallest_good_k, then we expect a real non-None instance.
result = debruijn_graph.build(ref, [], self.single_k_dbg_options(k))
if k < smallest_good_k:
self.assertIsNone(result, 'Cycle not detected for k={}'.format(k))
else:
self.assertIsNotNone(result, 'False cycle detected for k={}'.format(k))
def test_complex_region(self):
# There is a heterozygous 9 bp deletion of tandem TGA repeat.
# "chr20:10,095,379-10,095,500"
ref_reader = fasta.IndexedFastaReader(testdata.CHR20_FASTA)
bam_reader = sam.SamReader(testdata.CHR20_BAM)
region = ranges.parse_literal('chr20:10,095,379-10,095,500')
ref_seq = ref_reader.query(region)
reads = list(bam_reader.query(region))
dbg = debruijn_graph.build(ref_seq, reads, self.dbg_options())
self.assertIsNotNone(dbg)
self.assertEqual(44, dbg.kmer_size)
self.assertEqual(2, len(dbg.candidate_haplotypes()))
self.assertIn(ref_seq, dbg.candidate_haplotypes())
def test_complex_region(self):
# There is a heterozygous 9 bp deletion of tandem TGA repeat.
# "chr20:10,095,379-10,095,500"
ref_reader = fasta.RefFastaReader(testdata.CHR20_FASTA)
bam_reader = sam.SamReader(testdata.CHR20_BAM)
region = ranges.parse_literal('chr20:10,095,379-10,095,500')
ref_seq = ref_reader.query(region)
reads = list(bam_reader.query(region))
dbg = debruijn_graph.build(ref_seq, reads, self.dbg_options())
self.assertIsNotNone(dbg)
self.assertEqual(44, dbg.kmer_size)
self.assertEqual(2, len(dbg.candidate_haplotypes()))
self.assertIn(ref_seq, dbg.candidate_haplotypes())
def test_k_exceeds_read_length(self):
"""This is a regression test for b/64564513."""
# If k > read length, no edges will go into the graph from this read.
# This crashed prior to the bugfix.
ref_str = 'GATTACATG'
read_str = 'GATGACA'
read = test_utils.make_read(read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(8))
self.assertIsNotNone(dbg)
def test_k_exceeds_read_length(self):
"""This is a regression test for b/64564513."""
# If k > read length, no edges will go into the graph from this read.
# This crashed prior to the bugfix.
ref_str = 'GATTACATG'
read_str = 'GATGACA'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(8))
self.assertIsNotNone(dbg)
def test_basics(self):
"""Basic example."""
ref_str = 'GATTACA'
read_str = 'GATGACA'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
self.assertEqual(self.single_k_dbg_options(3).min_k, 3)
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(3))
self.assertItemsEqual([ref_str, read_str], dbg.candidate_haplotypes())
self.assertGraphEqual(
"""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
5[label=ATG];
6[label=TGA];
7[label=GAC];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
0->5 [label=2];
5->6 [label=2];
6->7 [label=2];
7->4 [label=2];
}
""", dbg)
def test_basics(self):
"""Basic example."""
ref_str = 'GATTACA'
read_str = 'GATGACA'
read = test_utils.make_read(
read_str,
chrom='chr20',
start=1,
cigar=[(len(read_str), 'M')],
quals=[30] * len(read_str),
name='read')
self.assertEqual(self.single_k_dbg_options(3).min_k, 3)
# Use two reads so read path doesn't get pruned.
dbg = debruijn_graph.build(ref_str, [read, read],
self.single_k_dbg_options(3))
self.assertItemsEqual([ref_str, read_str], dbg.candidate_haplotypes())
self.assertGraphEqual("""\
digraph G {
0[label=GAT];
1[label=ATT];
2[label=TTA];
3[label=TAC];
4[label=ACA];
5[label=ATG];
6[label=TGA];
7[label=GAC];
0->1 [label=1 color=red];
1->2 [label=1 color=red];
2->3 [label=1 color=red];
3->4 [label=1 color=red];
0->5 [label=2];
5->6 [label=2];
6->7 [label=2];
7->4 [label=2];
}
""", dbg)
