def find_single_unique(alns, bam, debug=False):
"""Extracts single unique alignment for indel detection
If there is only one alignment reported by BWA-mem even when '-a' is turned on
Args:
alns: (list) Pysam AlignedRead objects of the same contig
bam: Pysam bam handle
Returns:
Alignment object or None
"""
primary_alns = [
aln for aln in alns if not aln.is_unmapped and not aln.is_secondary
]
if len(primary_alns) == 1:
if primary_alns[0].mapq > 0:
matched_and_insertion_len = sum(
[a[1] for a in primary_alns[0].cigar if a[0] <= 1])
if float(matched_and_insertion_len) / float(
primary_alns[0].rlen) < 0.95:
if debug:
sys.stdout.write(
'best alignment less than 0.95 mapped:%s %s\n' %
(alns[0].qname, alns[0].cigarstring))
return None
else:
edit_distance = effective_edit_distance(alns[0])
if edit_distance is not None and float(edit_distance) / float(
primary_alns[0].inferred_length) > 0.1:
if debug:
sys.stdout.write(
'filter out single uniq alignment %s: edit distance %s - > 0.1 of contig len %d (%.01f)\n'
% (alns[0].qname, edit_distance,
primary_alns[0].inferred_length,
float(edit_distance) /
float(primary_alns[0].inferred_length)))
return None
else:
if debug:
sys.stdout.write(
'filter out single uniq alignment %s: mapq = 0\n' %
primary_alns[0].qname)
return None
#ambiguous_NM = 5
#for aln in alns:
#if aln.is_secondary and \
#not re.search('[HS]', aln.cigarstring) and\
#re.match('\d+M', aln.cigarstring) and re.search('\d+M$', aln.cigarstring) and\
#int(aln.opt('NM')) - int(primary_alns[0].opt('NM')) <= ambiguous_NM:
#if debug:
#sys.stdout.write('secondary alignments too similar %s\n' % primary_alns[0].qname)
#return None
return Alignment.from_alignedRead(primary_alns[0], bam)
else:
return None
def find_chimera(alns, bam, debug=False, check_haplotype=True):
"""Determine if given alignments are chimeric
Args:
alns: (List) List of Pysam AlignedRead objects
bam: (AlignmentFile) Pysam handle to BAM file - for getting reference info
debug: (Boolean) debug mode - will output debugging statements
check_haplotype: (Boolean) whether to screen out alignments to references
containing '_'
"""
primary_alns = []
secondary_alns = []
for aln in alns:
if re.search('[HS]', aln.cigarstring) and not aln.is_secondary:
primary_alns.append(aln)
else:
secondary_alns.append(aln)
if check_haplotype and len(primary_alns) > 1:
replace_haplotype(primary_alns, secondary_alns, bam)
if len(primary_alns) > 1:
aligns = [Alignment.from_alignedRead(aln, bam) for aln in primary_alns]
bad_aligns = [align for align in aligns if not align.is_valid()]
if bad_aligns:
if debug:
for align in bad_aligns:
sys.stdout.write('bad alignment %s %s %s %s %s %s' %
(align.query, align.qstart, align.qend,
align.target, align.tstart, align.tend))
else:
valid_secondary_aligns = []
if secondary_alns:
secondary_aligns = [
Alignment.from_alignedRead(aln, bam)
for aln in secondary_alns
]
valid_secondary_aligns = [
align for align in secondary_aligns if align.is_valid()
]
return aligns, valid_secondary_aligns
return None, None
def find_chimera(alns, bam, debug=False, check_haplotype=True):
"""Determine if given alignments are chimeric
Args:
alns: (List) List of Pysam AlignedRead objects
bam: (AlignmentFile) Pysam handle to BAM file - for getting reference info
debug: (Boolean) debug mode - will output debugging statements
check_haplotype: (Boolean) whether to screen out alignments to references
containing '_'
"""
primary_alns = []
secondary_alns = []
for aln in alns:
if re.search('[HS]', aln.cigarstring) and not aln.is_secondary:
primary_alns.append(aln)
else:
secondary_alns.append(aln)
if check_haplotype and len(primary_alns) > 1:
replace_haplotype(primary_alns, secondary_alns, bam)
if len(primary_alns) > 1:
aligns = [Alignment.from_alignedRead(aln, bam) for aln in primary_alns]
bad_aligns = [align for align in aligns if not align.is_valid()]
if bad_aligns:
if debug:
for align in bad_aligns:
sys.stdout.write('bad alignment %s %s %s %s %s %s' % (align.query,
align.qstart,
align.qend,
align.target,
align.tstart,
align.tend))
else:
valid_secondary_aligns = []
if secondary_alns:
secondary_aligns = [Alignment.from_alignedRead(aln, bam) for aln in secondary_alns]
valid_secondary_aligns = [align for align in secondary_aligns if align.is_valid()]
return aligns, valid_secondary_aligns
return None, None
def map_aligns(self,
bam,
query_fasta,
genome_fasta,
accessory_known_features=None,
find_events=True,
max_diff=1):
mappings = defaultdict(list)
junc_adjs = []
events = []
for query, group in groupby(bam.fetch(until_eof=True),
lambda aln: aln.query_name):
print 'processing', query
aligns = []
for aln in list(group):
if not aln.is_unmapped:
aligns.append(Alignment.from_alignedRead(aln, bam))
if not aligns:
continue
query_seq = query_fasta.fetch(query)
for align in aligns:
if not align.has_canonical_target() or align.blocks is None:
continue
block_matches = self.map_align(align)
if block_matches:
tid = self.pick_best_mapping(block_matches, align)
if tid is not None:
transcript = self.transcripts_dict[tid]
olap = self.overlap(align, transcript)
mappings[query].append(
(transcript.gene, transcript.id, olap))
junc_adjs.extend(
self.collect_junctions(align, transcript,
block_matches[tid]))
if find_events:
events.extend(
find_novel_junctions(block_matches[tid],
align,
transcript,
query_seq,
self.genome_fasta,
accessory_known_features=
accessory_known_features,
max_diff=max_diff))
return mappings, junc_adjs, events
def find_single_unique(alns, bam, debug=False):
"""Extracts single unique alignment for indel detection
If there is only one alignment reported by BWA-mem even when '-a' is turned on
Args:
alns: (list) Pysam AlignedRead objects of the same contig
bam: Pysam bam handle
Returns:
Alignment object or None
"""
primary_alns = [aln for aln in alns if not aln.is_unmapped and not aln.is_secondary]
if len(primary_alns) == 1:
if primary_alns[0].mapq > 0:
matched_and_insertion_len = sum([a[1] for a in primary_alns[0].cigar if a[0] <= 1])
if float(matched_and_insertion_len) / float(primary_alns[0].rlen) < 0.95:
if debug:
sys.stdout.write('best alignment less than 0.95 mapped:%s %s\n' % (alns[0].qname, alns[0].cigarstring))
return None
else:
edit_distance = effective_edit_distance(alns[0])
if edit_distance is not None and float(edit_distance)/float(primary_alns[0].inferred_length) > 0.1:
if debug:
sys.stdout.write('filter out single uniq alignment %s: edit distance %s - > 0.1 of contig len %d (%.01f)\n' % (alns[0].qname,
edit_distance,
primary_alns[0].inferred_length,
float(edit_distance)/float(primary_alns[0].inferred_length)
))
return None
else:
if debug:
sys.stdout.write('filter out single uniq alignment %s: mapq = 0\n' % primary_alns[0].qname)
return None
#ambiguous_NM = 5
#for aln in alns:
#if aln.is_secondary and \
#not re.search('[HS]', aln.cigarstring) and\
#re.match('\d+M', aln.cigarstring) and re.search('\d+M$', aln.cigarstring) and\
#int(aln.opt('NM')) - int(primary_alns[0].opt('NM')) <= ambiguous_NM:
#if debug:
#sys.stdout.write('secondary alignments too similar %s\n' % primary_alns[0].qname)
#return None
return Alignment.from_alignedRead(primary_alns[0], bam)
else:
return None
def map_aligns(self, bam, query_fasta, genome_fasta, accessory_known_features=None, find_events=True,
max_diff=1):
mappings = defaultdict(list)
junc_adjs = []
events = []
for query, group in groupby(bam.fetch(until_eof=True), lambda aln: aln.query_name):
print 'processing', query
aligns = []
for aln in list(group):
if not aln.is_unmapped:
aligns.append(Alignment.from_alignedRead(aln, bam))
if not aligns:
continue
query_seq = query_fasta.fetch(query)
for align in aligns:
if not align.has_canonical_target() or align.blocks is None:
continue
block_matches = self.map_align(align)
if block_matches:
tid = self.pick_best_mapping(block_matches, align)
if tid is not None:
transcript = self.transcripts_dict[tid]
olap = self.overlap(align, transcript)
mappings[query].append((transcript.gene, transcript.id, olap))
junc_adjs.extend(self.collect_junctions(align, transcript, block_matches[tid]))
if find_events:
events.extend(find_novel_junctions(block_matches[tid],
align,
transcript,
query_seq,
self.genome_fasta,
accessory_known_features=accessory_known_features,
max_diff=max_diff)
)
return mappings, junc_adjs, events
