def resolve_complex_sv_v2(resolve_CPX,
resolve_INV,
resolve_CNV,
cytobands,
disc_pairs,
mei_bed,
variant_prefix='CPX_',
min_rescan_support=4,
pe_blacklist=None,
quiet=False,
SR_only_cutoff=1000):
#resolve_CPX = [i for i in out_rec if i.info['SVTYPE']=='CPX']
#resolve_INV = [i for i in out_rec if i.info['SVTYPE']=='INV']
independent_INV = remove_CPX_from_INV(resolve_CPX, resolve_INV)
linked_INV = cluster_INV(independent_INV)
clusters_v2 = link_cpx_V2(linked_INV, resolve_CNV, cpx_dist=2000)
clusters_v2 = cluster_cleanup(clusters_v2)
#Print number of candidate clusters identified
if not quiet:
now = datetime.datetime.now()
print('svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'identified ' + str(len(clusters_v2)) +
' candidate complex clusters ' + 'during second pass')
cpx_records_v2 = deque()
cpx_record_ids_v2 = set()
for cluster in clusters_v2:
#Print status for each cluster
if not quiet:
now = datetime.datetime.now()
print(
'svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'resolving candidate cluster containing the following records: '
+ ', '.join([e.id for e in cluster]))
# Try finding opposite strand support for single ender inversions
if len(cluster) == 1 and cluster[0].info['SVTYPE'] == 'INV':
rec, opp = rescan_single_ender(cluster[0],
disc_pairs,
min_rescan_support,
pe_blacklist=pe_blacklist)
if opp is not None:
cluster = deque([rec, opp])
# if cxsv overlap pulled in unrelated insertions, keep them separate
if all([r.info['SVTYPE'] == 'INS' for r in cluster]):
for record in cluster:
cpx = ComplexSV([record], cytobands, mei_bed, SR_only_cutoff)
cpx_record_ids = cpx_record_ids.union(cpx.record_ids)
# Assign random string as resolved ID to handle sharding
cpx.vcf_record.id = variant_prefix + '_' + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
cpx_records.append(cpx.vcf_record)
# resolved_idx += 1
outcome = 'treated as separate unrelated insertions'
else:
cpx = ComplexSV(cluster, cytobands, mei_bed, SR_only_cutoff)
cpx_record_ids_v2 = cpx_record_ids_v2.union(cpx.record_ids)
if cpx.svtype == 'UNR':
# Assign random string as unresolved ID to handle sharding
unresolved_vid = 'UNRESOLVED_' + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
for i, record in enumerate(cpx.records):
record.info['EVENT'] = unresolved_vid
record.info['UNRESOLVED'] = True
cpx_records_v2.append(record)
# unresolved_idx += 1
outcome = 'is unresolved'
else:
cpx.vcf_record.id = variant_prefix + '_' + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
cpx_records_v2.append(cpx.vcf_record)
if 'CPX_TYPE' in cpx.vcf_record.info.keys():
outcome = 'resolved as ' + str(
cpx.vcf_record.info['CPX_TYPE'])
else:
outcome = 'resolved as ' + str(
cpx.vcf_record.info['SVTYPE'])
#Report outcome per cluster
if not quiet:
now = datetime.datetime.now()
print('svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'candidate cluster ' + outcome)
for i in cpx_records_v2:
if i.info['SVTYPE'] == 'CPX':
for info in 'UNRESOLVED EVENT UNRESOLVED_TYPE STRANDS'.split():
if info in i.info.keys():
i.info.pop(info)
return cpx_records_v2
def resolve_complex_sv(vcf,
cytobands,
disc_pairs,
mei_bed,
variant_prefix='CPX_',
min_rescan_support=4,
pe_blacklist=None,
quiet=False,
SR_only_cutoff=1000):
"""
Resolve complex SV from CNV intervals and BCA breakpoints.
Yields all resolved events, simple or complex, in sorted order.
Parameters
----------
vcf : pysam.VariantFile
cytobands : pysam.TabixFile
disc_pairs : pysam.TabixFile
mei_bed : pybedtools.BedTool
variant_prefix : str
Prefix to assign to resolved variants
min_rescan_support : int
Number of pairs required to count a sample as
supported during PE rescan
pe_blacklist : pysam.TabixFile, optional
Blacklisted genomic regions. Anomalous pairs in these regions will be
removed prior to clustering.
quiet : boolean, optional
Do not print status updates
Yields
------
sv : pysam.VariantRecord
"""
clusters = link_cpx(vcf)
clusters = clusters_cleanup(clusters)
#Print number of candidate clusters identified
if not quiet:
now = datetime.datetime.now()
print('svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'identified ' + str(len(clusters)) +
' candidate complex clusters ' + 'during first pass')
# resolved_idx = unresolved_idx = 1
if not variant_prefix.endswith('_'):
variant_prefix += '_'
cpx_records = deque()
cpx_record_ids = set()
for cluster in clusters:
#Print status for each cluster
if not quiet:
now = datetime.datetime.now()
print(
'svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'resolving candidate cluster containing the following records: '
+ ', '.join([e.id for e in cluster]))
# Try finding opposite strand support for single ender inversions
if len(cluster) == 1 and cluster[0].info['SVTYPE'] == 'INV':
rec, opp = rescan_single_ender(cluster[0],
disc_pairs,
min_rescan_support,
pe_blacklist=pe_blacklist)
if opp is not None:
cluster = deque([rec, opp])
# if cxsv overlap pulled in unrelated insertions, keep them separate
if all([r.info['SVTYPE'] == 'INS' for r in cluster]):
for record in cluster:
cpx = ComplexSV([record], cytobands, mei_bed, SR_only_cutoff)
cpx_record_ids = cpx_record_ids.union(cpx.record_ids)
# Assign random string as resolved ID to handle sharding
cpx.vcf_record.id = variant_prefix + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
cpx_records.append(cpx.vcf_record)
# resolved_idx += 1
outcome = 'treated as separate unrelated insertions'
else:
cpx = ComplexSV(cluster, cytobands, mei_bed, SR_only_cutoff)
cpx_record_ids = cpx_record_ids.union(cpx.record_ids)
if cpx.svtype == 'UNR':
# Assign random string as unresolved ID to handle sharding
unresolved_vid = 'UNRESOLVED_' + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
for i, record in enumerate(cpx.records):
record.info['EVENT'] = unresolved_vid
record.info['UNRESOLVED'] = True
cpx_records.append(record)
# unresolved_idx += 1
outcome = 'is unresolved'
elif cpx.svtype == 'SPLIT':
# check all CNVs for depth support and report the first
# insertion record and all CNVs with depth support. CNVs without
# depth support will have their IDs added to the MEMBERS field of
# the INS record
cnv_ids_to_append = []
for cnv in cpx.cnvs:
if 'RD' in cnv.info['EVIDENCE']:
cnv.info['MEMBERS'] = cnv.id
cpx_records.append(cnv)
else:
cnv_ids_to_append.append(cnv.id)
ins_rec = cpx.insertions[0]
ins_rec.info['MEMBERS'] = (
ins_rec.id, ) + tuple(cnv_ids_to_append)
cpx_records.append(ins_rec)
outcome = 'split into INS and CNV variants. ' + \
'The following records were merged into the INS record: ' + \
', '.join(cnv_ids_to_append)
else:
cpx.vcf_record.id = variant_prefix + ''.join(
random.choice(string.ascii_uppercase + string.digits)
for _ in range(10))
cpx_records.append(cpx.vcf_record)
if 'CPX_TYPE' in cpx.vcf_record.info.keys():
outcome = 'resolved as ' + str(
cpx.vcf_record.info['CPX_TYPE'])
else:
outcome = 'resolved as ' + str(
cpx.vcf_record.info['SVTYPE'])
# resolved_idx += 1
#Report outcome per cluster
if not quiet:
now = datetime.datetime.now()
print('svtk resolve @ ' + now.strftime("%H:%M:%S") + ': ' +
'candidate cluster ' + outcome)
# Output all variants
vcf.reset()
for record in _merge_records(vcf, cpx_records, cpx_record_ids):
#Clean all BNDs to ensure they are set to unresolved
#Reason: some SR-only BNDs will escape being set as UNRESOLVED if they
# are part of a multi-breakpoint complex cluster that remains UNRESOLVED
# after excluding SR-only breakpoints
if record.info['SVTYPE'] == 'BND':
record.info['UNRESOLVED'] = True
if 'UNRESOLVED_TYPE' not in record.info.keys():
record.info['UNRESOLVED_TYPE'] = 'SINGLE_ENDER'
if 'CPX_TYPE' in record.info.keys():
if 'UNRESOLVED' in record.info.keys():
record.info['UNRESOLVED_TYPE'] = record.info['CPX_TYPE']
record.info.pop('CPX_TYPE')
else:
record.info.pop('STRANDS')
if 'CIPOS' in record.info.keys():
record.info.pop('CIPOS')
if 'CIEND' in record.info.keys():
record.info.pop('CIEND')
if 'RMSSTD' in record.info.keys():
record.info.pop('RMSSTD')
yield record