def realign_reads(self, reads, region):
"""Run realigner.
This is the main function that
- parses the input reads and reference sequence.
- select candidate windows for local assembly (WindowSelector (ws)
module).
- Windows larger than max_window_size are skipped.
- build pruned De-Bruijn graph for each candidate window (DeBruijnGraph
(dbg) module).
- Graphs with more than max_num_paths candidate haplotypes or
with reference sequence as the only candidate are skipped.
- Align reads based on candidate haplotypes (Aligner (aln) module).
- Output all input reads (whether they required realignment or not).
Args:
reads: [`third_party.nucleus.protos.Read` protos]. The list of input reads
to realign.
region: A `third_party.nucleus.protos.Range` proto. Specifies the region
on the genome we should process.
Returns:
[realigner_pb2.CandidateHaplotypes]. Information on the list of candidate
haplotypes.
[`third_party.nucleus.protos.Read` protos]. The realigned
reads for the region. NOTE THESE READS MAY NO LONGER BE IN THE SAME
ORDER AS BEFORE.
"""
# Compute the windows where we need to assemble in the region.
candidate_windows = window_selector.select_windows(
self.config.ws_config, self.ref_reader, reads, region)
# Assemble each of those regions.
candidate_haplotypes = self.call_debruijn_graph(
candidate_windows, reads)
# Create our simple container to store candidate / read mappings.
assembled_regions = [AssemblyRegion(ch) for ch in candidate_haplotypes]
# Our realigned_reads start off with all of the unassigned reads.
realigned_reads = assign_reads_to_assembled_regions(
assembled_regions, reads)
# Walk over each region and align the reads in that region, adding them to
# our realigned_reads.
for assembled_region in assembled_regions:
if flags.FLAGS.use_fast_pass_aligner:
realigned_reads_copy = self.call_fast_pass_aligner(
assembled_region)
else:
raise ValueError(
'--use_fast_pass_aligner is always true. '
'The older implementation is deprecated and removed.')
realigned_reads.extend(realigned_reads_copy)
self.diagnostic_logger.log_realigned_reads(region, realigned_reads,
self.shared_header)
return candidate_haplotypes, realigned_reads
def test_select_windows_returns_empty_list_when_no_reads(self):
self.assertEqual([],
window_selector.select_windows(
self.config,
ref_reader=fasta.InMemoryFastaReader([('chr1', 0,
'A' * 500)]),
reads=[],
region=ranges.make_range('chr1', 1, 100)))
def test_select_windows(self):
# Simple end-to-end test of the high-level select_windows function. We give
# it a few reads with a single candidate at 100 and we expect a window back
# centered at 100.
reads = [
test_utils.make_read('AGA', start=99, cigar='3M', quals=[64] * 3),
test_utils.make_read('AGA', start=99, cigar='3M', quals=[63] * 3),
test_utils.make_read('AGA', start=99, cigar='3M', quals=[62] * 3),
]
chrom = reads[0].alignment.position.reference_name
ref_reader = fasta.InMemoryFastaReader([(chrom, 0, 'A' * 300)])
region = ranges.make_range(chrom, 0, 200)
self.assertEqual(
window_selector.select_windows(self.config, ref_reader, reads, region),
[ranges.make_range(chrom, 96, 104)])