def filter_a_chrom(chrom, oriPeaks, subPeaks, chromWig, args):
'''
Filters a chromsome
oriPeaks and subPeaks are lists of peaks from the same chromosome.
Both peak lists are sorted by coordinates on the chromosome.
'''
sub_peak_starts = [k[1] for k in subPeaks]
result = []
for op in oriPeaks:
startIdx = bisect.bisect_left( chromWig[:,0], op[1] )
endIdx = bisect.bisect_right( chromWig[:,0], op[2] )
subStartIdx = bisect.bisect_left( sub_peak_starts, op[1] )
subEndIdx = bisect.bisect_right( sub_peak_starts, op[2] )
if endIdx <= startIdx or subEndIdx <= subStartIdx:
continue
currWig = wig.expandWig( chromWig[startIdx:endIdx,:],0, 1, smooth=False )
startPos = chromWig[ startIdx, 0 ]
endPos = chromWig[ endIdx - 1, 0 ]
maxV = np.max( currWig ) #The maximum value
for subI in range( subStartIdx, subEndIdx ):
relStart = max( 0, subPeaks[ subI ][1] - startPos )
relEnd = min( currWig.shape[0] - 1, subPeaks[ subI ][2] - startPos )
subMaxV = np.max( currWig[ relStart: relEnd ] )
if subMaxV > args.threshfrac * maxV and subMaxV > args.cutoff:
result.append( subPeaks[ subI ] )
result[-1][4] = subMaxV
result[-1][5] = args.strand
return result
def add_chrom_data(taskQ, outQ, processID, args, strand ):
if args.method == 'g':
kargs = (args.bw, args.nbw)
else:
kargs = (args.r, args.mean)
for chrom, chromWig in iter( taskQ.get, 'STOP' ):
print "add chrom data Process ", processID," is processing ", chrom
lines = []
lines.append("variableStep chrom=%s\n"%(chrom,))
startp = chromWig[0,0]
expanded = wig.expandWig( chromWig, offset, 1, strand = strand, method=args.method, kargs=kargs)
for i in range( expanded.shape[0] ):
if expanded[i] > 0.8:
lines.append( "%d\t%f\n"%(int(i + startp - offset), expanded[i], ) )
#for i in range( chromWig.shape[0] ):
# lines.append('%d\t%f\n'%( chromWig[ i, 0], chromWig[i,1]))
expanded.resize(100000, refcheck=False)
expanded.resize(0, refcheck=False)
chromWig.resize(100000, refcheck=False)
chromWig.resize(0, refcheck=False)
outQ.put(lines)
gc.collect()
def pair( fpeaks, rpeaks, fwig, rwig, ulimit, dlimit, prefix):
'''
Assuming that the peaks on one strand is mutually exclusive.
They do not overlap with each other.
In this case, the ordering of the starts of the peaks and the
ends of the peaks are the same. And that when the starts are
sorted, the ends are also sorted.
'''
print "fwig: ", fwig
print "rwig: ", rwig
offset = 5
expandCol = 1
out1 = open(prefix + "_singletons_dev.bed",'w')
out2 = open(prefix + "_pairs_dev.gff", "w")
print fpeaks.keys()
for chrom in fpeaks:
if chrom not in rpeaks:
continue
print chrom
fp = fpeaks[chrom]
rp = rpeaks[chrom]
pairF = [] #Store the pairing information, if unpaired, it will be negative.
pairR = []
fw = fwig[chrom]
expandedFw = WIG.expandWig( fw, offset, expandCol, smooth=False )
rw = rwig[chrom]
expandedRw = WIG.expandWig( rw, offset, expandCol, smooth=False )
rstarts = []
rends = []
fprefer = []
rprefer = []
unpairedF = []
for f in fp:
fprefer.append( SortedCollection( key=itemgetter(1) ) )
pairF.append( ( -1, 0, 0) ) #( index of the mate, score, distance )
for r in rp:
rprefer.append( SortedCollection( key=itemgetter(1) ) )
pairR.append( (-1, 0, 0) )
rstarts.append( r[1] )
rends.append( r[2] )
for i in range( len( fp ) ):
currfp = fp[ i ]
start = currfp[1]
end = currfp[2]
es = start - ulimit
ee = end + dlimit
currFw = expandedFw[ max( 0, start - fw[0,0] ) + offset : max( 0, end - fw[0,0] ) + offset + 1]
si = bisect.bisect_left( rends, es )
ei = bisect.bisect_right( rstarts, ee )
ftagCounts,_,_,_ = gs.getTagCount( fwig, chrom, start, end )
#print ei - si
maxScore = 0
bestDist = 0
bestIdx = 0
bestRpos = 0
bestHeight = 0
for idx in range( si, ei ):
currrp = rp[ idx ]
rstart = currrp[1]
rend = currrp[2]
currRw = expandedRw[ max( 0, rstart - rw[0, 0] ) + offset : max( 0, rend - rw[0,0] ) + offset + 1 ]
rtagCoungs,_,_,_ = gs.getTagCount( rwig, chrom, currrp[1], currrp[2] )
tempScore, tempDist, tempRpos, tempHeight = getScore( currFw, currRw, start, rstart )
fprefer[ i ].insert( (idx, tempScore, tempDist, tempRpos, tempHeight) )
rprefer[ idx ].insert( (i, tempScore, tempDist, tempRpos, tempHeight) )
if tempScore > maxScore:
maxScore = tempScore
bestDist = tempDist
bestIdx = idx
bestRpos = tempRpos
bestHeight = tempHeight
if maxScore > pairR[ bestIdx ][1]:
pairF[i] = ( bestIdx, maxScore, bestDist , bestRpos, bestHeight)
if pairR[ bestIdx ][0] >= 0:
pairF[ pairR[ bestIdx ][ 0 ] ] = (-1, 0, 0)
unpairedF.append( pairR[ bestIdx ][0] )
pairR[bestIdx] = ( i, maxScore, bestDist, bestRpos, bestHeight )
else:
unpairedF.append( i )
try:
fprefer[ i ].remove( ( bestIdx, maxScore, bestDist, bestRpos, bestHeight ) )
except ValueError:
#print "Value error: ", bestIdx, ' ',maxScore, ' ',bestDist,' ', si,' ', ei
pass
singletons = []
pairs = []
while len(unpairedF) > 0:
for u in unpairedF:
if len( fprefer[u] ) > 0:
ridx = fprefer[u][-1][0]
if pairR[ ridx ][1] < fprefer[u][-1][1]:
if pairR[ ridx ][0] > 0:
pairF[ pairR[ ridx ][0] ] = (-1, 0, 0)
unpairedF.append( pairR[ ridx ][0] )
pairR[ ridx ] = ( u, fprefer[u][-1][1], fprefer[u][-1][2], fprefer[u][-1][3], fprefer[u][-1][4] )
pairF[ u ] = ( ridx, fprefer[u][-1][1], fprefer[u][-1][2], fprefer[u][-1][3], fprefer[u][-1][4] )
fprefer[u].remove( (ridx, fprefer[u][-1][1], fprefer[u][-1][2], fprefer[u][-1][3], fprefer[u][-1][4] ) )
else:
unpairedF.remove( u )
for i,f in enumerate(pairF):
fp[i][1] -= 1
if f[0] == -1:
singletons.append( fp[i] )
else:
rp[f[0]][1] -= 1
#pairs.append( fp[i] )
#pairs.append( rp[f[0]] )
pairStart = (2*f[3] - f[2])/2
pairEnd = pairStart + 1
pairs.append( [fp[i][0],'.','.',pairStart-10, pairEnd+10,f[4],'.','.','cw_distance='+str(f[2]) ] )
for i,f in enumerate(pairR):
rp[i][1] -= 1
if f[0] == -1:
singletons.append( rp[i] )
singletons.sort(key=lambda k:( k[0], k[1], k[2] ))
pairs.sort(key = lambda k:( k[0], k[1], k[2]))
print "singletons: ", len(singletons)
print "pairs: ", len(pairs)
for s in singletons:
out1.write('\t'.join([str(i) for i in s]))
out1.write('\n')
for p in pairs:
out2.write('\t'.join([str(i) for i in p]))
out2.write('\n')
out1.close()
out2.close()
