class ReadMapper(object):
def __init__(self, alphabet, wordlen, db_path):
self.db = DB(db_path, alphabet)
self.kmer_index = KmerIndex(self.db, wordlen)
self.seed_index = SeedIndex(self.kmer_index)
self.bands_indexed = False
def log(self, *args, **kwargs):
self.db.log(*args, **kwargs)
def initialize(self, reads_fa, refs_fa=None, num_reads=-1):
self.db.initialize()
with open(reads_fa) as f:
self.db.load_fasta(f, num=num_reads, rc=True)
if refs_fa is not None:
with open(refs_fa) as f:
self.db.load_fasta(f, rc=False)
def index_bands(self, **kw):
self.kmer_index.score_kmers()
self.seed_index.score_diagonals(**kw)
self.bands_indexed = True
def load_reads(self):
recs_by_content_id = {r.content_id: r for r in list(self.db.find())}
reads = []
for record in recs_by_content_id.values():
if 'rc_of' in record.attrs:
pair = (recs_by_content_id[record.attrs['rc_of']], record)
reads.append(Read(self.seed_index, *pair))
return sorted(reads, key=lambda read: read.record.id)
def load_refs(self):
recs_by_content_id = {r.content_id: r for r in list(self.db.find())}
for record in recs_by_content_id.values():
if 'rc_of' in record.attrs:
recs_by_content_id.pop(record.attrs['rc_of'])
recs_by_content_id.pop(record.content_id)
return recs_by_content_id.values()
def map_all_to_all(self, min_band_score, **aligner_kw):
assert self.bands_indexed, 'Bands must be indexed first'
self.log('Mapping all reads against each other')
reads = self.load_reads() # NOTE comes in sorted order of id
indic = ProgressIndicator(num_total=len(reads))
indic.start()
for read in reads:
indic.progress()
# NOTE only compare to reads after us
others = (r.record for r in reads if r.record.id > read.record.id)
for other in others:
rec, target_rec, aln = read.map(other,
min_band_score=min_band_score,
**aligner_kw)
if rec is None:
continue
yield rec, target_rec, aln
indic.finish()
def map_all_to_refs(self, min_band_score, **aligner_kw):
# FIXME it would be nice to only calculate bands for read v. ref not
# all pairwise of reads too
assert self.bands_indexed, 'Bands must be indexed first'
self.log('Mapping all reads against reference sequences')
reads, refs = self.load_reads(), self.load_refs()
indic = ProgressIndicator(num_total=len(reads))
indic.start()
for read in reads:
indic.progress()
rec, target_rec, aln = read.map(refs,
min_band_score=min_band_score,
**aligner_kw)
if rec is not None:
yield rec, target_rec, aln
indic.finish()
def mappings_from_sam(self, sampath):
"""Loads mappings from a SAM mapping file and translates sequence
names to integer identifiers as stored by :class:`biseqt.database.DB`.
Args:
db (database.DB): The sequence database where ids are looked up.
sampath (str): The path to SAM mappings file.
Yields:
tuple:
A 3-tuple containing the read record, the reference name and
the ``pysam.calignedsegment.AlignedSegment`` mapping it to the
reference.
"""
self.log('Loading SAM mappings from %s.' % sampath)
reads_by_name = {r.record.attrs['name']: r for r in self.load_reads()}
samfile = pysam.AlignmentFile(sampath)
for mapping in samfile.fetch():
qname, rname = mapping.query_name, mapping.reference_name
# NOTE this is because BLASR does a weird thing with sequence names
qname = qname.rsplit('/', 1)[0]
if qname not in reads_by_name:
continue
yield reads_by_name[qname], rname, mapping
def overlaps_from_sam_mappings(self, sampath, min_overlap=-1):
"""Finds all pairs of overlapping sequences based on their mappings to
a reference.
Args:
sampath (str): The path to SAM mappings file.
min_overlap (int): The minimum required length for overlaps to be
reported; default is -1 in which case no overlap is excluded.
Yields:
tuple:
A tuple of sequence integer ids (in increasing order) that are
deemed as overlapping based on SAM mappings.
"""
self.log('Finding overlaps from SAM mappings.')
mappings = {read.record.id: (read, ref, mapping)
for read, ref, mapping in self.mappings_from_sam(sampath)}
seqids = sorted(mappings.keys())
for id0, id1 in combinations(seqids, 2):
(r0, ref0, map0), (r1, ref1, map1) = mappings[id0], mappings[id1]
if ref0 != ref1:
continue
# TODO ignoring query_alignment_start and query_alignment_end
overlap_len = min(map0.reference_end, map1.reference_end) - \
max(map0.reference_start, map1.reference_start)
if overlap_len <= 0 or overlap_len < min_overlap:
continue
# FIXME the second thing we yield is not reported by our own
# map_all_to_all.
if map0.is_reverse == map1.is_reverse:
yield r0.record, r1.record
yield r0.rc_record, r1.rc_record
else:
yield r0.record, r1.rc_record
yield r0.rc_record, r1.record
