def classify_2_cnv(FF, RR, cnvs, min_frac=0.5):
"""
Classify the cxSV class of a pair of inv bkpts and two associated CNVs.
Matches each CNV to a 5' or 3' location, as constrained by the breakpoint
coordinates.
Parameters
----------
FF : pysam.VariantRecord
RR : pysam.VariantRecord
cnvs : [pysam.VariantRecord, pysam.VariantRecord]
min_frac : float, optional
Minimum reciprocal overlap of each cnv with a candidate CNV interval
defined by the breakpoint coordinates.
Returns
-------
svtype : str
"""
# Assign CNVs to 5' or 3' based on ordering
cnv5, cnv3 = sorted(cnvs, key=lambda r: r.pos)
# Check if 5' CNV matches breakpoints
if cnv5.info['SVTYPE'] == 'DEL':
interval5 = (FF.pos, RR.pos)
else:
interval5 = (RR.pos, FF.pos)
frac5 = svu.reciprocal_overlap(cnv5.pos, cnv5.stop, *interval5)
# Check if 3' CNV matches breakpoints
if cnv3.info['SVTYPE'] == 'DEL':
interval3 = (FF.stop, RR.stop)
else:
interval3 = (RR.stop, FF.stop)
frac3 = svu.reciprocal_overlap(cnv3.pos, cnv3.stop, *interval3)
# Report cxSV class based on whether CNVs matched intervals
if frac5 >= min_frac and frac3 >= min_frac:
svtype = (cnv5.info['SVTYPE'].lower() + 'INV' +
cnv3.info['SVTYPE'].lower())
elif frac5 >= min_frac and frac3 < min_frac:
return classify_1_cnv(FF, RR, cnv5)
elif frac5 < min_frac and frac3 >= min_frac:
return classify_1_cnv(FF, RR, cnv3)
else:
svtype = 'CNV_2_FAIL'
return svtype, cnvs
def _test_overlap(cnv):
svtype = cnv.info['SVTYPE']
if svtype == 'DEL':
frac5 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *del5)
frac3 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *del3)
else:
frac5 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *dup5)
frac3 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *dup3)
if frac5 >= min_frac and frac3 >= min_frac:
return svtype + '_53'
elif frac5 >= min_frac:
return svtype + '_5'
elif frac3 >= min_frac:
return svtype + '_3'
else:
return 'no_hit'
def classify_1_cnv(FF,
RR,
cnv,
min_frac=0.5,
min_bkpt_cnv_size=500,
max_bkpt_cnv_size=4000):
"""
Classify the cxSV class of a pair of inv bkpts and one associated CNV.
Matches each CNV to a 5' or 3' location, as constrained by the breakpoint
coordinates. After matching CNV, check if distance between breakpoints
at other end is sufficient to call a second flanking CNV.
Parameters
----------
FF : pysam.VariantRecord
RR : pysam.VariantRecord
cnvs : [pysam.VariantRecord, pysam.VariantRecord]
min_frac : float, optional
Minimum reciprocal overlap of each cnv with a candidate CNV interval
defined by the breakpoint coordinates.
min_bkpt_cnv_size : int, optional
Minimum distance between breakpoints to call flanking CNV.
max_bkpt_cnv_size : int, optional
Maximum distance between breakpoints to call flanking CNV.
Returns
-------
svtype : str
"""
# Make CNV class lowercase (for later concatenation with INV)
cnv_type = cnv.info['SVTYPE'].lower()
# Determine eligible 5'/3' CNV intervals defined by the breakpoints
if cnv_type == 'del':
interval5 = (FF.pos, RR.pos)
interval3 = (FF.stop, RR.stop)
else:
interval5 = (RR.pos, FF.pos)
interval3 = (RR.stop, FF.stop)
# Check overlap of CNV against full inversion length
start = min(FF.pos, RR.pos)
end = max(FF.stop, RR.stop)
total_frac = svu.reciprocal_overlap(cnv.pos, cnv.stop, start, end)
frac5 = svu.overlap_frac(*interval5, cnv.pos, cnv.stop)
frac3 = svu.overlap_frac(*interval3, cnv.pos, cnv.stop)
# If one CNV spans the entire event, it likely represents two CNV merged
# during preprocessing or clustering
if total_frac > 0.9 and frac5 > 0.95 and frac3 > 0.95:
svtype = cnv_type + 'INV' + cnv_type # + '_merged'
return svtype, [cnv]
# Otherwise, check whether it's 5' or 3'
frac5 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *interval5)
frac3 = svu.reciprocal_overlap(cnv.pos, cnv.stop, *interval3)
# 5' CNV; check 3' breakpoints for small flanking CNV
if frac5 >= min_frac and frac3 < min_frac:
svtype = cnv_type + 'INV'
dist3 = RR.stop - FF.stop
if min_bkpt_cnv_size <= dist3 < max_bkpt_cnv_size:
svtype = svtype + 'del'
elif min_bkpt_cnv_size <= -dist3 < max_bkpt_cnv_size:
svtype = svtype + 'dup'
# 3' CNV; check 5' breakpoints for small flanking CNV
elif frac5 < min_frac and frac3 >= min_frac:
svtype = 'INV' + cnv_type
dist5 = RR.pos - FF.pos
if min_bkpt_cnv_size <= dist5 < max_bkpt_cnv_size:
svtype = 'del' + svtype
elif min_bkpt_cnv_size <= -dist5 < max_bkpt_cnv_size:
svtype = 'dup' + svtype
# Couldn't match the CNV
else:
return classify_0_cnv(FF, RR)
return svtype, [cnv]