def filterOverlaps(self, overlapPercCutoff=70):
"""Filtering out amplicons that substantially overlap.
The amplicon with the highest PPC with be kept.
The MFEprimerRes attribute must be set.
in-place edit of MFEprimerRes object (table object filtered of overlaps)
Parmeters
---------
overlapPercCutoff : float
percent of overlap to consider 'substantially' overlapping
"""
if self.MFEprimerRes is None:
msg = 'genome object does not have MFEprimerRes attribute.' + \
' Run MFEprimer() first'
raise AttributeError, msg
# making interval tree
tree = IntervalTree()
# loading intervals
for count, row in self.MFEprimerRes.iterrows():
# sanity check for + strand
if row['BindingStart'] > row['BindingStop']:
raise TypeError('MFEprimer binding start-stop is not + strand')
tree.addi(row['BindingStart'], row['BindingStop'],
[count, row['PPC'], row['Size']])
# finding all that substantially overlap; keeping one with > PPC
tree2 = tree.copy()
for iv1 in tree.iter():
# skipping if already removed from tree2
if not iv1 in tree2:
continue
overlaps = tree.search(iv1.begin, iv1.end)
# skipping those that poorly overlap
lowOverlap = set()
for iv2 in overlaps:
if iv1.overlaps(iv2):
percOverlaps = self._calcPercOverlap(iv1, iv2)
if percOverlaps[0] < overlapPercCutoff:
lowOverlap.add(iv2)
overlaps = overlaps - lowOverlap # just list of substantially overlapping
# skipping those that have been already removed
prevRm = set([x for x in overlaps if x not in tree2])
overlaps = overlaps - prevRm
# removing all substantially overlapping intervals with lower PPC
if len(overlaps) > 1:
overlaps = sorted(overlaps,
key=lambda x: x.data[1],
reverse=True)
for o in overlaps[1:]:
if o in tree2:
tree2.remove(o)
else:
pass
# selecting columns
iv_idx = [x.data[0] for x in tree2.iter()]
self.MFEprimerRes = self.MFEprimerRes.iloc[iv_idx]
