def profileTargetsHistoAS(tccList, cName, name='boxplot'):
range = 50
histDict = {} # {coord: []}
histDictAS = {}
for tcc in tccList:
chrom, strand, start, end = cg.tccSplit(tcc)
#Get highest peak (sense)
tccStretch = cgPeaks.stretch(tcc, cName)
tccStretch.createPeaks(span=2)
highestCoord = tccStretch.getHighestPeak()
if highestCoord == None: continue
#AS
tccAS = cg.convertToAS(tcc)
tccStretch = cgPeaks.stretch(tccAS, cName)
tccStretch.createPeaks(span=2)
highestCoordAS = tccStretch.getHighestPeak()
if highestCoordAS == None: continue
#profile around point (Sense)
zPoint = cg.makeTcc(chrom, strand, highestCoord, end)
cProfile = svs.profileAroundPoint(zPoint, range, cName, ratio=True)
for coord in cProfile:
try:
histDict[coord].append(cProfile[coord])
except: #quicker way to initialize
histDict[coord] = [cProfile[coord]]
#profile around point (AS)
zPoint = cg.convertToAS(zPoint)
cProfile = svs.profileAroundPoint(zPoint,
range,
cName,
ratio=True,
ratioCoord=highestCoordAS)
for coord in cProfile:
try:
histDictAS[coord].append(cProfile[coord])
except: #quicker way to initialize
histDictAS[coord] = [cProfile[coord]]
plot.boxPlotHistoAS(histDict, histDictAS, name=name)
def extendPeakTest(tcc, pRange, minVal, maxAvgNoise, minPeakLength, maxPeakLength, cName):
chrom, strand, peakPosition, end = cg.tccSplit(tcc)
cProfile = stepVectorScan.profileAroundPoint(tcc, pRange, cName, ratio = True)
#extend this peak left and right
leftRange = range(1-pRange, 0)
rightRange = range(1, pRange)
leftRange.reverse() #going from the middle outward
#left
startFinal = leftRange[-1]
for i in leftRange:
if cProfile[i] > minVal:
print ' extending stretch'
else:
print ' end of stretch L'
startFinal = i + 1
break
#right
endFinal = rightRange[-1]
for i in rightRange:
if cProfile[i] > minVal:
print ' extending stretch'
else:
print ' end of stretch R'
endFinal = i - 1
break
peakLength = endFinal - startFinal + 1
#avg expression around peak check...
#get total expression before peak
low = startFinal
high = endFinal
noiseExpression = 0
lowRange = range(1 - pRange, low)
highRange = range(high + 1, pRange)
totalLength = len(lowRange) + len(highRange)
print totalLength, pRange, low, high, lowRange, highRange
for i in lowRange:
noiseExpression += cProfile[i]
for i in highRange:
noiseExpression += cProfile[i]
try:
avgNoise = noiseExpression/float(totalLength)
except:
return False
#filter out peaks that look a certain way.
if (minPeakLength < peakLength < maxPeakLength) and (avgNoise < maxAvgNoise):
goodTcc = cg.makeTcc(chrom, strand, peakPosition + startFinal, peakPosition + endFinal)
print '*KEEPER'
return goodTcc
else:
return False
def profileTargetsHistoAS(tccList, cName, name = 'boxplot'):
range = 50
histDict = {} # {coord: []}
histDictAS = {}
for tcc in tccList:
chrom, strand, start, end = cg.tccSplit(tcc)
#Get highest peak (sense)
tccStretch = cgPeaks.stretch(tcc, cName)
tccStretch.createPeaks(span = 2)
highestCoord = tccStretch.getHighestPeak()
if highestCoord == None: continue
#AS
tccAS = cg.convertToAS(tcc)
tccStretch = cgPeaks.stretch(tccAS, cName)
tccStretch.createPeaks(span = 2)
highestCoordAS = tccStretch.getHighestPeak()
if highestCoordAS == None: continue
#profile around point (Sense)
zPoint = cg.makeTcc(chrom, strand, highestCoord, end)
cProfile = svs.profileAroundPoint(zPoint, range, cName, ratio = True)
for coord in cProfile:
try:
histDict[coord].append(cProfile[coord])
except: #quicker way to initialize
histDict[coord] = [cProfile[coord]]
#profile around point (AS)
zPoint = cg.convertToAS(zPoint)
cProfile = svs.profileAroundPoint(zPoint, range, cName, ratio = True, ratioCoord = highestCoordAS)
for coord in cProfile:
try:
histDictAS[coord].append(cProfile[coord])
except: #quicker way to initialize
histDictAS[coord] = [cProfile[coord]]
plot.boxPlotHistoAS(histDict, histDictAS, name = name)
def profileTargetsHisto(tccList, cName, name='boxplot'):
histDict = {} # {coord: []}
for tcc in tccList:
chrom, strand, start, end = cg.tccSplit(tcc)
#Get highest peak
tccStretch = cgPeaks.stretch(tcc, cName)
tccStretch.createPeaks(span=2)
highestCoord = tccStretch.getHighestPeak()
if highestCoord == None: continue
#profile around point
zPoint = cg.makeTcc(chrom, strand, highestCoord, end)
cProfile = svs.profileAroundPoint(zPoint, 200, cName, ratio=True)
for coord in cProfile:
try:
histDict[coord].append(cProfile[coord])
except: #quicker way to initialize
histDict[coord] = [cProfile[coord]]
plot.boxPlotHisto(histDict, name=name)
def profileTargetsHisto(tccList, cName, name = 'boxplot'):
histDict = {} # {coord: []}
for tcc in tccList:
chrom, strand, start, end = cg.tccSplit(tcc)
#Get highest peak
tccStretch = cgPeaks.stretch(tcc, cName)
tccStretch.createPeaks(span = 2)
highestCoord = tccStretch.getHighestPeak()
if highestCoord == None: continue
#profile around point
zPoint = cg.makeTcc(chrom, strand, highestCoord, end)
cProfile = svs.profileAroundPoint(zPoint, 200, cName, ratio = True)
for coord in cProfile:
try:
histDict[coord].append(cProfile[coord])
except: #quicker way to initialize
histDict[coord] = [cProfile[coord]]
plot.boxPlotHisto(histDict, name = name)
def parallelMakePeaks(tcc, cName, minExpression):
conf = c.getConfig(cName)
f = open('out/peakData.%s.%s.%s' % (tcc, minExpression, conf.conf['assembly']), 'w')
print 'scanning range', tcc
chrom, strand, start, end = cg.tccSplit(tcc)
peaks = cgPeaks.stretch(tcc, cName)
#print 'getting peaks'
peaks.createPeaks(span = 1, minVal = int(minExpression))
print 'len peaks', len(peaks.peaks)
endCheck = 0
for x in peaks.peaks:
print x, endCheck
'''
if x < endCheck:
print 'endChecked'
continue
'''
#scan a 30 bp range around this point and find the best roof...
pRange = 40
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc, pRange, cName, ratio = True)
#now get highest stretch length and the rNext coord.
minVal = .70
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
print ' ', x + i, cProfile[i]
if cProfile[i] > minVal:
print ' extending stretch'
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
print 'end of stretch'
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- extend value...
val = [1.0, 1.0]
extend = 1
if (startFinal) and (endFinal):
low = startFinal - extend
high = endFinal + extend
if low > (1 - pRange) and high < pRange:
val[0] = float(cProfile[startFinal - extend])
val[1] = float(cProfile[endFinal + extend])
else:
print 'out of range'
continue
else:
print 'no start and end of peak'
continue
print low, high, x, endFinal
endCheck = x + endFinal
#avg expression around peak check...
#get total expression before peak
noiseExpression = 0
lowRange = range(1 - pRange, low)
highRange = range(high + 1, pRange)
totalLength = len(lowRange) + len(highRange)
for i in lowRange:
noiseExpression += cProfile[i]
for i in highRange:
noiseExpression += cProfile[i]
avgNoise = noiseExpression/float(totalLength)
#filter out peaks that look a certain way.
print highest, val[0], val[1], avgNoise
if 0 < highest < 5: #rooflength 14/26
if val[0] < 0.20 and val[1] < .20: #drop values
if avgNoise < .3:
goodTcc = cg.makeTcc(chrom, strand, x + low, x + high)
print '*KEEPER'
f.write('%s\n' % goodTcc)
f.close()
print 'DONE', tcc
def makePeakInput(cName, minExpression=2000):
mConf = c.getConfig('Main.conf')
conf = c.getConfig(cName)
assembly = conf.conf['assembly']
tccList = []
chromLens = cg.returnChromLengthDict(assembly)
f = open('peakData.%s' % minExpression, 'w')
for chrom in chromLens:
if chrom not in cg.acceptableChroms: continue
for strand in ['1', '-1']:
print 'Getting Peaks for ', chrom, strand
prevI = 0
endCheck = 0
for i in rangePoints(1, chromLens[chrom], 1000):
if i == 1:
prevI = i
continue
start = prevI
end = i
prevI = i
tcc = cg.makeTcc(chrom, strand, start, end)
#print 'scanning range', tcc
peaks = cgPeaks.stretch(tcc, cName)
peaks.createPeaks(span=3, minVal=minExpression)
for x in peaks.peaks:
if x < endCheck:
continue
#scan a 30 bp range around this point and find the best roof...
pRange = 30
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc,
pRange,
cName,
ratio=True)
#now get highest stretch length and the rNext coord.
minVal = .80
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
if cProfile[i] > minVal:
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- 4 value...
val = [1.0, 1.0]
if (startFinal) and (endFinal):
low = startFinal - 4
high = endFinal + 4
if low > (1 - pRange) and high < pRange:
val[0] = float(cProfile[startFinal - 4])
val[1] = float(cProfile[endFinal + 4])
else:
continue
else:
continue
endCheck = x + high
#filter out peaks that look a certain way.
if 14 < highest < 26: #rooflength
if val[0] < 0.2 and val[1] < .2: #drop values
goodTcc = cg.makeTcc(chrom, strand, x + low,
x + high)
#print goodTcc
f.write('%s\n' % goodTcc)
f.close()
def findPeaks(pType, cName = None):
#init
mConf = c.cgConfig('Main.conf')
conf = c.getConfig(cName)
if pType == 'E':
predName = conf.conf['resultsExonsSorted']
else:
predName = conf.conf['resultsIntronsSorted']
print predName
#make CID:hairpin:peak dictionary
cHairs = getHairpins.getHairpins(predName)
peakDict = {}
for CID in cHairs:
peakDict[CID] = [cHairs[CID],'None']
timer = cg.cgTimer()
timer.start()
#put peaks in memory
print 'Creating peak data'
peaks = {} # chr:peak:value
for CID in cHairs:
chrom, strand, start, end = cg.tccSplit(cHairs[CID])
tcc = cHairs[CID]
#init dictionary
if chrom not in peaks:
peaks[chrom] = {}
if strand not in peaks[chrom]:
peaks[chrom][strand] = {}
#create peaks for tcc and add to peak dictionary
stretch = cgPeaks.stretch(tcc, cName)
stretch.createPeaks()
for peakCoord in stretch.peaks:
peaks[chrom][strand][peakCoord] = 0
print timer.split()
print 'finding best combos'
bestCombos = []
aPass = 0
bPass = 0
cPass = 0
numT = 0
for CID in peakDict:
cgFlag = False
if CID == '538':cgFlag = True
tcc = peakDict[CID][0]
#print tcc
tccPeaks = []
chrom = cg.ss(tcc, ':')[0]
strand = cg.ss(tcc, ':')[1]
start = int(cg.ss(tcc, ':')[2])
end = int(cg.ss(tcc, ':')[3])
#get all peaks
for i in range(start, end + 1):
if i in peaks[chrom][strand]:
#print ' peak added', i
tccPeaks.append(i)
#Calculate parameters...
pairStrings = [] #used to check if pair already added
peakCombos = []
for x in tccPeaks:
#scan a 30 bp range around this point and find the best roof...
pRange = 30
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#quickly get max value...kinda a long way to do it but whatever
cProfile = stepVectorScan.profileAroundPoint(rTcc, 1, cName, ratio = False)
xval = cProfile[0]
max = xval
highestValueCoord = x
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc, pRange, cName, ratio = True)
#now get highest stretch length and the rNext coord.
minVal = .80
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
if cProfile[i] > minVal:
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- 4 value...
val = [1.0, 1.0]
if (startFinal) and (endFinal):
low = startFinal - 4
high = endFinal + 4
if low > (1 - pRange):
if high < pRange:
val[0] = float(cProfile[startFinal - 4])
val[1] = float(cProfile[endFinal + 4])
#fill in other details...
y = 'S'
dist = 'S'
ratio = 'S'
peakCombos.append([tcc,x,y,dist,ratio,max,highest,val])
#print ' ', peakCombos[-1]
#find best combo...
topCombo = None
for combo in peakCombos:
roofLength = combo[6]
dropValue = combo[7][0]
if combo[7][1] > dropValue:
dropValue = combo[7][1]
#print roofLength, dropValue
if 14 < roofLength < 26:
if 0.0 < dropValue < 0.2:
#pick one with rooflength nearest 20:
if topCombo:
if (math.fabs(22 - roofLength)) < (math.fabs(22 - topCombo[6])):
topCombo = combo
else:
topCombo = combo
if topCombo:
peakDict[CID][1] = topCombo
bestCombos.append(topCombo)
print bestCombos[-1]
else:
#print 'None'
pass
print timer.split()
#now update predFile (SLOT 13)
predFile = open(predName, 'r')
newLines = []
for line in predFile:
CID = cg.ss(line)[7]
if peakDict[CID][1] == 'None':
peakInfo = 'None'
else:
peakInfo = '%s:%s:%s:%s:%s:%s' % (str(peakDict[CID][1][1])[-3:], 'S', str(peakDict[CID][1][4]).split('.')[0], peakDict[CID][1][5],peakDict[CID][1][6], peakDict[CID][1][7])
newLines.append(cg.appendToLine(line, peakInfo, 13))
predFile.close()
predFile = open(predName, 'w')
predFile.writelines(newLines)
predFile.close()
def roofPeakTest(tcc, pRange, minRoofVal, maxAvgNoise, maxDropVal, extend, minPeakLength, maxPeakLength, cName):
'''Note: extend does not extend the coordinates into the final tcc, just
used for declaring peak'''
coveredStartingPoints = set()
chrom, strand, peakPosition, end = cg.tccSplit(tcc)
cProfile = stepVectorScan.profileAroundPoint(tcc, pRange, cName, ratio = True)
#extend this peak left and right
leftRange = range(1-pRange, 0)
rightRange = range(1, pRange)
leftRange.reverse() #going from the middle outward
startFinalE = 0
#left
startFinal = leftRange[-1]
for i in leftRange:
if cProfile[i] > minRoofVal:
pass
else:
startFinal = i + 1
startFinalE = cProfile[i]
break
endFinalE = 0
#right
endFinal = rightRange[-1] #this only holds if it extends to the end of the range...
for i in rightRange:
if cProfile[i] > minRoofVal:
pass
else:
endFinal = i - 1
endFinalE = cProfile[i]
break
peakLength = endFinal - startFinal + 1
extend = int(extend)
low = startFinal - extend
high = endFinal + extend
if low > (1 - pRange) and high < pRange:
dropPassL = False
dropPassR = False
#find if any of the values in the extended range are below drop range
leftDrop = [float(cProfile[startFinal - x]) for x in range(1, extend + 1)]
rightDrop = [float(cProfile[endFinal + x]) for x in range(1, extend + 1)]
leftDropPass = [True if x < maxDropVal else False for x in leftDrop]
rightDropPass = [True if x < maxDropVal else False for x in rightDrop]
if True in leftDropPass:
dropPassL = True
if True in rightDropPass:
dropPassR = True
if (not dropPassL) or (not dropPassR):
print 'dropVal Fail', 'dropLeft', 'dropRight'
print startFinal, endFinal, leftDrop, rightDrop, leftDropPass, rightDropPass
return False
else:
print 'out of range'
return False
#avg expression around peak check...
#get total expression before peak
noiseExpression = 0
lowRange = range(1 - pRange, low)
highRange = range(high + 1, pRange)
totalLength = len(lowRange) + len(highRange)
#for i in lowRange:
#noiseExpression += cProfile[i]
#for i in highRange:
#noiseExpression += cProfile[i]
#avgNoise = noiseExpression/float(totalLength)
avgNoise = 0.0
#filter out peaks that look a certain way.
print startFinal, startFinalE, endFinal, endFinalE, peakLength,
goodTcc = cg.makeTcc(chrom, strand, peakPosition + startFinal, peakPosition + endFinal)
if (minPeakLength <= peakLength <= maxPeakLength) and (avgNoise < maxAvgNoise):
#if (minPeakLength < peakLength < maxPeakLength):
print ' *KEEPER', goodTcc, peakLength, avgNoise, maxAvgNoise
return goodTcc
else:
#print 'bad peak', chrom, strand, peakPosition + startFinal, peakPosition + endFinal
print ' reason', goodTcc, peakLength, avgNoise, maxAvgNoise
return False
def parallelMakePeaks(tcc, cName, minExpression):
conf = c.getConfig(cName)
f = open(
'out/peakData.%s.%s.%s' % (tcc, minExpression, conf.conf['assembly']),
'w')
print 'scanning range', tcc
chrom, strand, start, end = cg.tccSplit(tcc)
peaks = cgPeaks.stretch(tcc, cName)
#print 'getting peaks'
peaks.createPeaks(span=1, minVal=int(minExpression))
print 'len peaks', len(peaks.peaks)
endCheck = 0
for x in peaks.peaks:
print x, endCheck
'''
if x < endCheck:
print 'endChecked'
continue
'''
#scan a 30 bp range around this point and find the best roof...
pRange = 40
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc,
pRange,
cName,
ratio=True)
#now get highest stretch length and the rNext coord.
minVal = .70
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
print ' ', x + i, cProfile[i]
if cProfile[i] > minVal:
print ' extending stretch'
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
print 'end of stretch'
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- extend value...
val = [1.0, 1.0]
extend = 1
if (startFinal) and (endFinal):
low = startFinal - extend
high = endFinal + extend
if low > (1 - pRange) and high < pRange:
val[0] = float(cProfile[startFinal - extend])
val[1] = float(cProfile[endFinal + extend])
else:
print 'out of range'
continue
else:
print 'no start and end of peak'
continue
print low, high, x, endFinal
endCheck = x + endFinal
#avg expression around peak check...
#get total expression before peak
noiseExpression = 0
lowRange = range(1 - pRange, low)
highRange = range(high + 1, pRange)
totalLength = len(lowRange) + len(highRange)
for i in lowRange:
noiseExpression += cProfile[i]
for i in highRange:
noiseExpression += cProfile[i]
avgNoise = noiseExpression / float(totalLength)
#filter out peaks that look a certain way.
print highest, val[0], val[1], avgNoise
if 0 < highest < 5: #rooflength 14/26
if val[0] < 0.20 and val[1] < .20: #drop values
if avgNoise < .3:
goodTcc = cg.makeTcc(chrom, strand, x + low, x + high)
print '*KEEPER'
f.write('%s\n' % goodTcc)
f.close()
print 'DONE', tcc
def makePeakInput(cName, minExpression = 2000):
mConf = c.getConfig('Main.conf')
conf = c.getConfig(cName)
assembly = conf.conf['assembly']
tccList = []
chromLens = cg.returnChromLengthDict(assembly)
f = open('peakData.%s' % minExpression, 'w')
for chrom in chromLens:
if chrom not in cg.acceptableChroms: continue
for strand in ['1', '-1']:
print 'Getting Peaks for ', chrom, strand
prevI = 0
endCheck = 0
for i in rangePoints(1, chromLens[chrom], 1000):
if i == 1:
prevI = i
continue
start = prevI
end = i
prevI = i
tcc = cg.makeTcc(chrom, strand, start, end)
#print 'scanning range', tcc
peaks = cgPeaks.stretch(tcc, cName)
peaks.createPeaks(span = 3, minVal = minExpression)
for x in peaks.peaks:
if x < endCheck:
continue
#scan a 30 bp range around this point and find the best roof...
pRange = 30
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc, pRange, cName, ratio = True)
#now get highest stretch length and the rNext coord.
minVal = .80
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
if cProfile[i] > minVal:
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- 4 value...
val = [1.0, 1.0]
if (startFinal) and (endFinal):
low = startFinal - 4
high = endFinal + 4
if low > (1 - pRange) and high < pRange:
val[0] = float(cProfile[startFinal - 4])
val[1] = float(cProfile[endFinal + 4])
else:
continue
else:
continue
endCheck = x + high
#filter out peaks that look a certain way.
if 14 < highest < 26: #rooflength
if val[0] < 0.2 and val[1] < .2: #drop values
goodTcc = cg.makeTcc(chrom, strand, x + low, x + high)
#print goodTcc
f.write('%s\n' % goodTcc)
f.close()
def roofPeakTest(tcc, pRange, minRoofVal, maxAvgNoise, maxDropVal, extend, minPeakLength, maxPeakLength, cName):
'''Note: extend does not extend the coordinates into the final tcc, just
used for declaring peak'''
chrom, strand, peakPosition, end = cg.tccSplit(tcc)
cProfile = stepVectorScan.profileAroundPoint(tcc, pRange, cName, ratio = True)
#extend this peak left and right
leftRange = range(1-pRange, 0)
rightRange = range(1, pRange)
leftRange.reverse() #going from the middle outward
#left
startFinal = leftRange[-1]
for i in leftRange:
if cProfile[i] > minRoofVal:
print ' extending stretch'
else:
print ' end of stretch L'
startFinal = i + 1
break
#right
endFinal = rightRange[-1] #this only holds if it extends to the end of the range...
for i in rightRange:
if cProfile[i] > minRoofVal:
print ' extending stretch'
else:
print ' end of stretch R'
endFinal = i - 1
break
peakLength = endFinal - startFinal + 1
extend = int(extend)
val = [0.0, 0.0]
low = startFinal - extend
high = endFinal + extend
if low > (1 - pRange) and high < pRange:
val[0] = float(cProfile[startFinal - extend])
val[1] = float(cProfile[endFinal + extend])
if not (val[0] < maxDropVal and val[1] < maxDropVal):
return False
else:
print 'out of range'
return False
#avg expression around peak check...
#get total expression before peak
noiseExpression = 0
lowRange = range(1 - pRange, low)
highRange = range(high + 1, pRange)
totalLength = len(lowRange) + len(highRange)
for i in lowRange:
noiseExpression += cProfile[i]
for i in highRange:
noiseExpression += cProfile[i]
avgNoise = noiseExpression/float(totalLength)
#filter out peaks that look a certain way.
if (minPeakLength < peakLength < maxPeakLength) and (avgNoise < maxAvgNoise):
goodTcc = cg.makeTcc(chrom, strand, peakPosition + startFinal, peakPosition + endFinal)
print '*KEEPER'
return goodTcc
else:
return False
def findPeaks(pType, cName=None):
#init
mConf = c.cgConfig('Main.conf')
conf = c.getConfig(cName)
if pType == 'E':
predName = conf.conf['resultsExonsSorted']
else:
predName = conf.conf['resultsIntronsSorted']
print predName
#make CID:hairpin:peak dictionary
cHairs = getHairpins.getHairpins(predName)
peakDict = {}
for CID in cHairs:
peakDict[CID] = [cHairs[CID], 'None']
timer = cg.cgTimer()
timer.start()
#put peaks in memory
print 'Creating peak data'
peaks = {} # chr:peak:value
for CID in cHairs:
chrom, strand, start, end = cg.tccSplit(cHairs[CID])
tcc = cHairs[CID]
#init dictionary
if chrom not in peaks:
peaks[chrom] = {}
if strand not in peaks[chrom]:
peaks[chrom][strand] = {}
#create peaks for tcc and add to peak dictionary
stretch = cgPeaks.stretch(tcc, cName)
stretch.createPeaks()
for peakCoord in stretch.peaks:
peaks[chrom][strand][peakCoord] = 0
print timer.split()
print 'finding best combos'
bestCombos = []
aPass = 0
bPass = 0
cPass = 0
numT = 0
for CID in peakDict:
cgFlag = False
if CID == '538': cgFlag = True
tcc = peakDict[CID][0]
#print tcc
tccPeaks = []
chrom = cg.ss(tcc, ':')[0]
strand = cg.ss(tcc, ':')[1]
start = int(cg.ss(tcc, ':')[2])
end = int(cg.ss(tcc, ':')[3])
#get all peaks
for i in range(start, end + 1):
if i in peaks[chrom][strand]:
#print ' peak added', i
tccPeaks.append(i)
#Calculate parameters...
pairStrings = [] #used to check if pair already added
peakCombos = []
for x in tccPeaks:
#scan a 30 bp range around this point and find the best roof...
pRange = 30
rTcc = cg.makeTcc(chrom, strand, x, x + 1)
#quickly get max value...kinda a long way to do it but whatever
cProfile = stepVectorScan.profileAroundPoint(rTcc,
1,
cName,
ratio=False)
xval = cProfile[0]
max = xval
highestValueCoord = x
#now make profile for roof...
cProfile = stepVectorScan.profileAroundPoint(rTcc,
pRange,
cName,
ratio=True)
#now get highest stretch length and the rNext coord.
minVal = .80
highest = 0
stretch = 0
startCurrent = None
startFinal = None
endFinal = None
for i in range(1 - pRange, pRange):
if cProfile[i] > minVal:
stretch += 1
if startCurrent == None:
startCurrent = i
else:
if stretch > 0:
if stretch > highest: #stretch ended and was higher than previous
highest = stretch
endFinal = i - 1
startFinal = startCurrent
startCurrent = None
else:
startCurrent = None
stretch = 0
#get +/- 4 value...
val = [1.0, 1.0]
if (startFinal) and (endFinal):
low = startFinal - 4
high = endFinal + 4
if low > (1 - pRange):
if high < pRange:
val[0] = float(cProfile[startFinal - 4])
val[1] = float(cProfile[endFinal + 4])
#fill in other details...
y = 'S'
dist = 'S'
ratio = 'S'
peakCombos.append([tcc, x, y, dist, ratio, max, highest, val])
#print ' ', peakCombos[-1]
#find best combo...
topCombo = None
for combo in peakCombos:
roofLength = combo[6]
dropValue = combo[7][0]
if combo[7][1] > dropValue:
dropValue = combo[7][1]
#print roofLength, dropValue
if 14 < roofLength < 26:
if 0.0 < dropValue < 0.2:
#pick one with rooflength nearest 20:
if topCombo:
if (math.fabs(22 - roofLength)) < (
math.fabs(22 - topCombo[6])):
topCombo = combo
else:
topCombo = combo
if topCombo:
peakDict[CID][1] = topCombo
bestCombos.append(topCombo)
print bestCombos[-1]
else:
#print 'None'
pass
print timer.split()
#now update predFile (SLOT 13)
predFile = open(predName, 'r')
newLines = []
for line in predFile:
CID = cg.ss(line)[7]
if peakDict[CID][1] == 'None':
peakInfo = 'None'
else:
peakInfo = '%s:%s:%s:%s:%s:%s' % (
str(peakDict[CID][1][1])[-3:], 'S', str(
peakDict[CID][1][4]).split('.')[0], peakDict[CID][1][5],
peakDict[CID][1][6], peakDict[CID][1][7])
newLines.append(cg.appendToLine(line, peakInfo, 13))
predFile.close()
predFile = open(predName, 'w')
predFile.writelines(newLines)
predFile.close()
