def scan_alignment(bamfile):
inpsam = os.popen('samtools view "{}"'.format(bamfile))
totaltags = totalbases = 0
cnt_subst = cnt_del = cnt_ins = 0
for record in sam.parse_sam_simple(inpsam):
# <ParsedLine qname='6-3799' flag=0 rname='chr9' pos=120960330 mapq=40 cigar='36M' rnext='*' pnext='0' tlen=0 seq='AAAAACTTTATCGGGGATACATGCGGTAGGGTAAAT' qual='*'>
nreads = int(record.qname.split('-')[1])
totaltags += nreads
reflen, readlen = sam.calculate_cigar_length(record.cigar)
totalbases += nreads * reflen
nmtag = [int(tok.split(':')[2]) for tok in str(record).split('\t')[11:]
if tok.startswith('NM:i:')][0]
rec_inserts = rec_deletions = 0
for size, cigarcmd in sam.cigar_pattern.findall(record.cigar):
if cigarcmd == 'D':
rec_deletions += int(size)
elif cigarcmd == 'I':
rec_inserts += int(size)
rec_substs = nmtag - rec_inserts - rec_deletions
assert rec_substs >= 0
cnt_subst += rec_substs * nreads
cnt_del += rec_deletions * nreads
cnt_ins += rec_inserts * nreads
return {
'tags': totaltags,
'bases': totalbases,
'substitutions': cnt_subst,
'deletions': cnt_del,
'insertions': cnt_ins,
}
def process(inpfile, output, refseqfile, delpos='5'):
if delpos == '5':
is_leftbind = lambda line: not line.flag & F_REVERSE_STRAND
elif delpos == '3':
is_leftbind = lambda line: line.flag & F_REVERSE_STRAND
else:
raise ValueError('Unknown binding oriention %s' % repr(delpos))
for line in parse_sam_simple(inpfile):
if str(line)[0] == '@' or 'D' not in line.cigar:
output.write(str(line))
continue
leftpos = line.pos - 1 # to zero-based coord
reflen, seqlen = calculate_cigar_length(line.cigar)
refseq = refseqfile.get(line.rname, leftpos, leftpos + reflen).upper()
assert len(refseq) == reflen
cigar_tokens = map(list, cigar_pattern.findall(line.cigar))
leftclip = int(cigar_tokens[0][0]) if cigar_tokens[0][1] == 'S' else 0
rightclip = int(cigar_tokens[-1][0]) if cigar_tokens[-1][1] == 'S' else 0
refleft = 0
tainted = False
for toki, (width, cmd) in enumerate(cigar_tokens):
width = int(width)
if cmd != 'D':
if cmd in 'MPN':
refleft += width
continue
deleted = refseq[refleft:refleft+width]
if len(set(deleted)) > 1:
# do not break deletions with two or more heteronucleotide, e.g.:
# AGGGGAGTGGTGCAGTGTGATTCCCGCCGGT
# AGGG--GTGGTGCAGTGTGATTCCCGCCGGT
# ^^ 2D
refleft += width
continue
lefttok = cigar_tokens[toki-1]
righttok = cigar_tokens[toki+1]
if is_leftbind(line): # push to the left as possible
spanleft = refleft
if lefttok[1] == 'M':
lefttokwidth = int(lefttok[0])
for i in xrange(spanleft - 1, spanleft - lefttokwidth - 1, -1):
if refseq[i] == deleted[0]:
spanleft = i
else:
break
leftexpand = refleft - spanleft
if leftexpand > 0:
lefttok[0] = str(int(lefttok[0]) - leftexpand)
righttok[0] = str(int(righttok[0]) + leftexpand)
tainted = True
else: # push to the right as possible
spanright = refleft + width - 1
if righttok[1] == 'M':
righttokwidth = int(righttok[0])
for i in xrange(refleft + width, refleft + width + righttokwidth):
if refseq[i] == deleted[0]:
spanright = i
else:
break
rightexpand = spanright - refleft - width + 1
if rightexpand > 0:
lefttok[0] = str(int(lefttok[0]) + rightexpand)
righttok[0] = str(int(righttok[0]) - rightexpand)
tainted = True
refleft += width
if not tainted:
output.write(str(line))
else:
fields = str(line).split('\t')
fields[5] = ''.join(w+cmd for w, cmd in cigar_tokens)
output.write('\t'.join(fields))
