def annotate_softclips(cv, read, ci_file):
sc_idxs = [idx for idx, clip in enumerate(read.cigar) if clip[0] == constants.CIGAR['soft-clip'] \
and clip[1] >= MIN_CLIP]
for sc_idx in sc_idxs:
cigar = read.cigar[sc_idx]
cv.cvtype, cv.vsize, cv.cvsize = 'UN', int(cigar[1]), int(cigar[1])
sc_left = sc_idx == 0
block_idx = 0 if sc_idx == 0 else np.max(
np.where(np.array([b[0] for b in cv.blocks]) < sc_idx)[0])
block = cv.blocks[block_idx][1]
cv.pos = int(block[0]) + 1 if sc_left else int(block[1])
rcigar = read.cigar[::-1] if cv.strand == '-' else read.cigar
cv.cpos = sum(
[v for c, v in rcigar[:sc_idx] if c in constants.AFFECT_CONTIG])
varseq = read.query_sequence[cv.cpos:(cv.cpos + cv.cvsize)]
refseq = read.get_reference_sequence()
cv.ref = refseq[0] if sc_left else refseq[-1]
cv.alt = '%s]%s:%d]%s' % (varseq, cv.chrom, cv.pos-1, cv.ref) if sc_left else \
'%s[%s:%d[%s' % (cv.ref, cv.chrom, cv.pos+1, varseq)
cv.vid = get_next_id(read.query_name)
CrypticVariant.write_contig_info(ci_file, cv)
print(cv.vcf_output())
def main():
args = parse_args()
set_globals(args)
init_logging(args.log)
print(VCF.get_header(args.sample))
CrypticVariant.write_contig_header(args.contig_info_output)
annotate_contigs(args)
def annotate_gaps(cv, read, ci_file):
'''
Annotate deletions and insertions
'''
gap_idxs = [idx for idx, gap in enumerate(read.cigar) if gap[0] in constants.GAPS and gap[1] >= MIN_GAP]
for gap_idx in gap_idxs:
cigar = read.cigar[gap_idx]
cv.vsize = int(cigar[1])
block_idx = 0 if gap_idx == 0 else np.max(np.where(np.array([b[0] for b in cv.blocks]) < gap_idx)[0])
block = cv.blocks[block_idx][1]
cv.pos = int(block[1])
# position of variant on contig
cv.cpos = sum([v for c,v in read.cigar[:gap_idx] if c in constants.AFFECT_CONTIG])
cv.cvsize = cv.vsize if read.cigar[gap_idx][0] == constants.CIGAR['insertion'] else 0
# ^ only insertions affect contig pos
if cigar[0] == constants.CIGAR['insertion']:
cv.cvtype = 'INS'
seq_pos1 = sum([v for c,v in read.cigar[:gap_idx] if c in constants.AFFECT_CONTIG \
and c != constants.CIGAR['hard-clip']])
seq_pos2 = seq_pos1 + cv.cvsize
seq = read.query_sequence[(seq_pos1-1):seq_pos2]
cv.ref, cv.alt = seq[:1], seq
else:
cv.cvtype = 'DEL'
seq_pos1 = sum([v for c,v in read.cigar[:gap_idx] if c in constants.AFFECT_REF])
seq_pos2 = seq_pos1 + cv.vsize
seq = read.get_reference_sequence()[(seq_pos1-1):seq_pos2]
cv.ref, cv.alt = seq, seq[:1]
cv.vid = get_next_id(read.query_name)
CrypticVariant.write_contig_info(ci_file, cv)
print(cv.vcf_output())
def annotate_single_read(args, read, juncs, ex_ref, ref_trees, outbam=None, genes=''):
'''
Annotate insertions, deletions and soft-clips on a single read
'''
ci_file = args.contig_info_output
genes = get_overlapping_genes(read, ref_trees) if genes == '' else genes
fusion = any([op == constants.CIGAR['hard-clip'] and val >= MIN_CLIP for op, val in read.cigar])
if genes == '' and not fusion:
logging.info('No gene(s) intersecting read %s; skipping' % read.query_name)
return
# check for contig gaps or soft-clips
has_gaps = any([op in constants.GAPS and val >= MIN_GAP for op, val in read.cigar])
has_scs = any([op == constants.CIGAR['soft-clip'] and val >= MIN_CLIP for op, val in read.cigar])
is_spliced = any([op == constants.CIGAR['skipped'] for op, val in read.cigar])
# check junctions
tx_juncs = get_tx_juncs(read)
unknown_juncs = ['%s:%s-%s' % (c, s, e) not in juncs[0] for c, s, e in tx_juncs]
has_novel_juncs = any(unknown_juncs)
# check for novel blocks
chr_ref = get_chr_ref(read, ex_ref)
has_novel_blocks = any([bh.is_novel_block(block, chr_ref, MIN_CLIP) for block in read.get_blocks()])
if has_gaps or has_scs or has_novel_juncs or has_novel_blocks:
cv = CrypticVariant().from_read(read)
cv.genes = genes
if has_gaps:
annotate_gaps(cv, read, ci_file)
if has_scs:
annotate_softclips(cv, read, ci_file)
if has_novel_juncs:
novel_juncs = [list(x) for x in np.array(tx_juncs)[unknown_juncs]]
annotate_juncs(cv, read, juncs[1], novel_juncs, ci_file)
if has_novel_blocks:
annotate_blocks(cv, read, chr_ref, ci_file)
if outbam:
outbam.write(read)
else:
logging.info('Nothing to annotate for read %s (read matches reference)' % read.query_name)
def annotate_juncs(cv, read, locs, novel_juncs, ci_file):
for junc in novel_juncs:
pos1, pos2 = int(junc[1]), int(junc[2])
junc_idx = [idx for idx, block in cv.blocks if block[1] == pos1][0]
junc_type = read.cigar[junc_idx + 1][0]
if junc_type in constants.GAPS or junc_type == constants.CIGAR[
'soft-clip']:
continue
cp = CrypticVariant().from_read(read) # partner variant
cp.genes = cv.genes
varseq, refseq = '', read.get_reference_sequence()
cpos = sum([
v for c, v in read.cigar[:(junc_idx + 1)]
if c in constants.AFFECT_CONTIG
])
rpos = sum([
v for c, v in read.cigar[:(junc_idx + 1)]
if c in constants.AFFECT_REF
])
cv.cpos, cp.cpos = cpos, cpos
cv.pos, cp.pos = pos1, pos2 + 1
cv.ref, cp.ref = refseq[rpos - 1], refseq[rpos]
cv.alt = '%s[%s:%d[%s' % (cv.ref, cv.chrom, cv.pos, varseq)
cp.alt = '%s]%s:%d]%s' % (varseq, cp.chrom, cp.pos, cp.ref)
cv.vid = get_next_id(read.query_name)
cp.vid = get_next_id(read.query_name)
cv.parid, cp.parid = cp.vid, cv.vid
loc_left = '%s:%d' % (cv.chrom, pos1)
loc_right = '%s:%d' % (cv.chrom, pos2)
if not (loc_left in locs) and not (loc_right in locs):
# neither end annotated, novel exon junction
cv.cvtype, cp.cvtype = 'NEJ', 'NEJ'
elif not (loc_left in locs and loc_right in locs):
# one end unannotated, partial novel junction
cv.cvtype, cp.cvtype = 'PNJ', 'PNJ'
else:
# both ends annotated, alternative splice site
cv.cvtype, cp.cvtype = 'AS', 'AS'
print(cv.vcf_output())
print(cp.vcf_output())
CrypticVariant.write_contig_info(ci_file, cv, cp)
def annotate_fusion(args, read, juncs, bam_idx, ex_ref, ref_trees, outbam):
try:
r1, r2 = bam_idx.find(read.query_name)
except ValueError:
logging.info(
'WARNING: found >2 reads matching hard-clipped read %s; cannot process'
% read.query_name)
return
ci_file = args.contig_info_output
cv1 = CrypticVariant().from_read(r1)
cv2 = CrypticVariant().from_read(r2)
cv1.cvtype, cv2.cvtype = 'FUS', 'FUS'
cv1.genes = get_overlapping_genes(r1, ref_trees)
cv2.genes = get_overlapping_genes(r2, ref_trees)
if cv1.genes == cv2.genes:
# intra-genic rearrangement
cv1.cvtype, cv2.cvtype = 'IGR', 'IGR'
if cv1.genes == '' and cv2.genes == '':
# no intersecting gene, this is not an interesting fusion
logging.info(
'No gene(s) intersecting candidate fusion contig %s; skipping' %
read.query_name)
record[read.query_name] = []
return
hc_idx1 = [
idx for idx, clip in enumerate(r1.cigar)
if clip[0] == constants.CIGAR['hard-clip']
][0]
hc_left1 = hc_idx1 == 0
block_idx1 = 0 if hc_left1 else np.max(
np.where(np.array([b[0] for b in cv1.blocks]) < hc_idx1)[0])
block1 = cv1.blocks[block_idx1][1]
cv1.pos = int(block1[0]) if hc_left1 else int(block1[1])
hc_idx2 = [
idx for idx, clip in enumerate(r2.cigar)
if clip[0] == constants.CIGAR['hard-clip']
][0]
hc_left2 = hc_idx2 == 0
block_idx2 = 0 if hc_left2 else np.max(
np.where(np.array([b[0] for b in cv2.blocks]) < hc_idx2)[0])
block2 = cv2.blocks[block_idx2][1]
cv2.pos = int(block2[0]) if hc_left2 else int(block2[1])
#TODO: handle inserted sequence btw fusion
varseq1, varseq2 = '', ''
refseq1 = r1.get_reference_sequence()
refseq2 = r2.get_reference_sequence()
bracket_dir1 = '[' if r1.is_reverse == r2.is_reverse else ']'
bracket_dir2 = ']' if r1.is_reverse == r2.is_reverse else ']'
cv1.ref = refseq1[0] if hc_left1 else refseq1[-1]
cv2.ref = refseq2[0] if hc_left2 else refseq1[-1]
if r1.is_reverse == r2.is_reverse:
cv1.alt = '%s]%s:%d]%s' % (varseq1, cv2.chrom, cv2.pos-1, cv1.ref) if hc_left1 else \
'%s[%s:%d[%s' % (cv1.ref, cv2.chrom, cv2.pos+1, varseq1)
cv2.alt = '%s]%s:%d]%s' % (varseq2, cv1.chrom, cv1.pos-1, cv2.ref) if hc_left2 else \
'%s[%s:%d[%s' % (cv2.ref, cv1.chrom, cv1.pos+1, varseq2)
else:
# contigs align on opposite strands
cv1.alt = '%s[%s:%d[%s' % (varseq1, cv2.chrom, cv2.pos-1, cv1.ref) if hc_left1 else \
'%s]%s:%d]%s' % (cv1.ref, cv2.chrom, cv2.pos+1, varseq1)
cv2.alt = '%s[%s:%d[%s' % (varseq2, cv1.chrom, cv1.pos-1, cv2.ref) if hc_left2 else \
'%s]%s:%d]%s' % (cv2.ref, cv1.chrom, cv1.pos+1, varseq2)
cv1.vid = get_next_id(r1.query_name)
cv2.vid = get_next_id(r2.query_name)
cv1.parid, cv2.parid = cv2.vid, cv1.vid
print(cv1.vcf_output())
print(cv2.vcf_output())
outbam.write(r1)
outbam.write(r2)
CrypticVariant.write_contig_info(ci_file, cv1, cv2)
annotate_single_read(args, r1, juncs, ex_ref, ref_trees, genes=cv1.genes)
annotate_single_read(args, r2, juncs, ex_ref, ref_trees, genes=cv2.genes)
def annotate_blocks(cv, read, chr_ref, ci_file):
'''
Annotate any sequence that is outside of exonic regions
'''
cv.parid = '.' # blocks don't have pairs
novel_blocks = [(idx, block) for idx, block in cv.blocks
if bh.is_novel_block(block, chr_ref, MIN_CLIP)]
for block_idx, block in novel_blocks:
cpos1 = sum([
v for c, v in read.cigar[:block_idx]
if c in constants.AFFECT_CONTIG
])
cpos2 = sum([
v for c, v in read.cigar[:block_idx + 1]
if c in constants.AFFECT_CONTIG
])
# whether sequence block is overlapping, or on left or right side of contig block
olapping = chr_ref[np.logical_and(block[0] < chr_ref.start,
block[1] > chr_ref.end)]
left = chr_ref[np.logical_and(block[1] > chr_ref.start,
block[1] <= chr_ref.end)]
right = chr_ref[np.logical_and(block[0] >= chr_ref.start,
block[0] < chr_ref.end)]
if len(left) > 0 and len(right) > 0:
# retained intron
cv.cvtype = 'RI'
qseq, rseq = bh.get_block_sequence(read, block_idx)
seq_right_pos = block[1] - min(left.start)
seq_left_pos = max(right.end) - block[0]
cv.pos = block[0] + seq_left_pos + 1
cv.ref = rseq[seq_left_pos:(-seq_right_pos)]
cv.alt = ']' + qseq[seq_left_pos:(-seq_right_pos)] + '['
cv.cpos = cpos1 + seq_left_pos
cv.vsize, cv.cvsize = abs(len(cv.alt) - 2 -
len(cv.ref)), len(cv.alt) - 2
cv.vid = get_next_id(read.query_name)
elif len(olapping) > 0:
# annotate left side
cv.cvtype = 'EE'
cv = annotate_block_left(cv, read, cpos2, olapping, block,
block_idx)
cv.vid = get_next_id(read.query_name)
print(cv.vcf_output())
CrypticVariant.write_contig_info(ci_file, cv)
# annotate right side
cv = annotate_block_right(cv, read, cpos1, olapping, block,
block_idx)
cv.vid = get_next_id(read.query_name)
elif len(left) > 0:
# annotate left side
cv.cvtype = 'EE'
cv = annotate_block_left(cv, read, cpos2, left, block, block_idx)
cv.vid = get_next_id(read.query_name)
elif len(right) > 0:
# annotate right side
cv.cvtype = 'EE'
cv = annotate_block_right(cv, read, cpos1, right, block, block_idx)
cv.vid = get_next_id(read.query_name)
else:
# block does not cross any annotation
qseq, rseq = bh.get_block_sequence(read, block_idx)
cv.ref, cv.alt = rseq, '[' + qseq + ']'
cv.pos, cv.cvtype = block[0] + 1, 'NE'
cv.cpos = cpos1
cv.vid = get_next_id(read.query_name)
cv.vsize, cv.cvsize = abs(len(cv.alt) - 2 -
len(cv.ref)), len(cv.alt) - 2
print(cv.vcf_output())
CrypticVariant.write_contig_info(ci_file, cv)