def checkingSons(peaks):
"""
bring back both, actually works, may be used only for the display
and not inside the data structure
"""
toRemove = MSPeakList()
def recursiveMerging(root):
if not root in toRemove:
for frag in root.fragCluster:
if frag.fragCluster:
toRemove.addPeak(frag)
for son in frag.fragCluster:
if son != root:
root.fragCluster.append(son)
#toRemove.addPeak(son)#dont know if it is really necessary
#else:
# print "must do other not yet implemented"
root.fragCluster = MSPeakList(list(set(root.fragCluster)))
recursiveMerging(frag) #before was just one step ahead
return
return
for peak in peaks:
recursiveMerging(peak)
lastPeakList = MSPeakList()
for p in peaks:
if p not in toRemove:
lastPeakList.append(p)
return sorted(lastPeakList, key=lambda x: x.mass())
def checkingSons(peaks):
"""
bring back both, actually works, may be used only for the display
and not inside the data structure
"""
toRemove=MSPeakList()
def recursiveMerging(root):
if not root in toRemove:
for frag in root.fragCluster:
if frag.fragCluster:
toRemove.addPeak(frag)
for son in frag.fragCluster:
if son != root:
root.fragCluster.append(son)
#toRemove.addPeak(son)#dont know if it is really necessary
#else:
# print "must do other not yet implemented"
root.fragCluster=MSPeakList(list(set(root.fragCluster)))
recursiveMerging(frag)#before was just one step ahead
return
return
for peak in peaks:
recursiveMerging(peak)
lastPeakList=MSPeakList()
for p in peaks:
if p not in toRemove:
lastPeakList.append(p)
return sorted(lastPeakList, key=lambda x:x.mass())
def clusteringWrapper(sample, **kw):
"""
convenience function wrapping isotopic finders
fragment adducts finder, put the data in MetHelperFunctions
"""
badPeak = kw.get('badPeak', False)
kw['mode'] = sample.kind
mode = kw['mode']
adducts = kw.get('frags') #contains the union of frags and adducts values
clusterLength = kw.get('clusterLength', 3)
isomasses = []
if mode == 'HighRes': #tuple(mass, prob, nmol) nmol for calculating 2M+Na+ for example
m = [
(1.0033548378, 0.0107),
(2.003241988, 0.), #C Element
(1.006276746, 0.000115),
(2.008224235, 0.), #H Element
(0.997034893, 0.00368), #N Element
(1.004216878, 0.00038),
(2.004245778, 0.00205), #O Element, P got no isotopes
(0.99938781, 0.0076),
(1.99579614, 0.0429),
(3.99501019, 0.0002)
] #S Element
for i in range(1, clusterLength + 1): #setting isotopic stuffs
isomasses += ([(x[0] * i, x[1]**i) for x in m
if x[0] * i < clusterLength])
else:
isomasses = [(float(i), 0.) for i in xrange(1, clusterLength + 1)
] #map(float, range(1, clusterLength+1))
# adducts={(18., 1):'H20',
# (44.,1):'CO2',
# (79.,1):'PO3',
# (90.,1):'' ,
# (70.,1):'',
# (68.,1):'',
# (60.,1):'',
# (59.,1):'',
# (46.,1):'',
# (45.,1):'',
# (35.,1):'',
# (32.,1):'',
# (28.,1):'',
# (17.,1):'',
# (16.,1):'',
# (14.,1):''}
#change the adducts to be a dictionnary
#isomasses.sort(key=lambda x:x[0])
p = sample.rawPeaks if badPeak else MSPeakList(
[peak for peak in sample.rawPeaks if peak.isGood])
sample.mappedPeaks, sample.isotopicPeaks = isotopicPeakListFinder(
p, isomasses, **kw) #first remove isotopicPeaks in peaklist
sample.mappedPeaks, sample.adductPeaks = clusteringCAMERA(
sample.mappedPeaks, adducts,
**kw) #second remove fragment/adducts off the peaklist
print "raw vs mapped:", str(len(sample.rawPeaks)), str(
len(sample.mappedPeaks))
def covMatrix(lspl, **k):
"""
idea use the leastsq method to calculate the covariance matrix
"""
#depacking the arguments
rtError = k.get('rtError')
ppm = k.get('ppm')
matrix = []
for sple in lspl:
treatedPeaks = MSPeakList()
for peak in sple.peaks:
if peak not in treatedPeaks:
peaks = MSPeakList()
rValues = []
rValues.append((peak.mass(), peak.rt, sple.xmlfile))
peaks.append(peak)
for s in lspl:
if s != spl:
common = s.peaks.peakInMZRTRange(
peak.mass(), peak.rt, ppm, rtError)
if len(common) > 1:
#do something eliminate some peaks to get the closest in rt and mz
treatedPeaks.extend(common)
peaks.extend(common)
if not common:
pass
#treat peak
for i in range(len(peaks)):
corr = 0.
isosA, fragsA = peaks[i].isoAreaList(
includeM0=True), peaks[i].fragAreaList()
for j in range(i + 1, len(peaks)):
#mean of the frags and the iso
#idea ponderate with the length more length more weighted
isosB, fragsB = peaks[j].isoAreaList(
includeM0=True), peaks[j].fragAreaList()
corr += r_coef(isosA, isosB)
corr += r_coef(fragsA, fragsB)
rValues.append(corr / 2.)
matrix.append(rValues)
return matrix
def clusterComparison(list_):#receive a list of peak with clusters identified
"""
return the best peak
WARNING: p_ydata and p_.y_data are None
TODO:
"""
sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
longest=len(sortedList[-1].fragCluster)
sameSizePeaks=MSPeakList()
for p in sortedList:
if len(p.fragCluster) == longest:
sameSizePeaks.append(p)
if len(sameSizePeaks) == 1:
return sameSizePeaks[0]
corr=np.array([0.] * len(sameSizePeaks))
#for i, p in enumerate(sameSizePeaks):
# for p_ in p.fragCluster:
# corr[i] += r_coef(p_.y_data, p.y_data)
m=max_f(corr)
return sameSizePeaks[np.where(corr == m)[0][0]]
def recursiveMerging(root):
if not root in toRemove:
for frag in root.fragCluster:
if frag.fragCluster:
toRemove.addPeak(frag)
for son in frag.fragCluster:
if son != root:
root.fragCluster.append(son)
#toRemove.addPeak(son)#dont know if it is really necessary
#else:
# print "must do other not yet implemented"
root.fragCluster = MSPeakList(list(set(root.fragCluster)))
recursiveMerging(frag) #before was just one step ahead
return
return
def clusterComparison(
list_): #receive a list of peak with clusters identified
"""
return the best peak
WARNING: p_ydata and p_.y_data are None
TODO:
"""
sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
longest = len(sortedList[-1].fragCluster)
sameSizePeaks = MSPeakList()
for p in sortedList:
if len(p.fragCluster) == longest:
sameSizePeaks.append(p)
if len(sameSizePeaks) == 1:
return sameSizePeaks[0]
corr = np.array([0.] * len(sameSizePeaks))
#for i, p in enumerate(sameSizePeaks):
# for p_ in p.fragCluster:
# corr[i] += r_coef(p_.y_data, p.y_data)
m = max_f(corr)
return sameSizePeaks[np.where(corr == m)[0][0]]
def covMatrix(lspl, **k):
"""
idea use the leastsq method to calculate the covariance matrix
"""
#depacking the arguments
rtError = k.get('rtError')
ppm = k.get('ppm')
matrix = []
for sple in lspl:
treatedPeaks=MSPeakList()
for peak in sple.peaks:
if peak not in treatedPeaks:
peaks=MSPeakList()
rValues =[]
rValues.append((peak.mass(), peak.rt, sple.xmlfile))
peaks.append(peak)
for s in lspl:
if s != spl:
common = s.peaks.peakInMZRTRange(peak.mass(), peak.rt, ppm, rtError)
if len(common) > 1:
#do something eliminate some peaks to get the closest in rt and mz
treatedPeaks.extend(common)
peaks.extend(common)
if not common:
pass
#treat peak
for i in range(len(peaks)):
corr =0.
isosA, fragsA = peaks[i].isoAreaList(includeM0=True), peaks[i].fragAreaList()
for j in range(i+1, len(peaks)):
#mean of the frags and the iso
#idea ponderate with the length more length more weighted
isosB, fragsB = peaks[j].isoAreaList(includeM0=True), peaks[j].fragAreaList()
corr+=r_coef(isosA, isosB)
corr+=r_coef(fragsA, fragsB)
rValues.append(corr/2.)
matrix.append(rValues)
return matrix
def identify(self):
idx = self.acTree.selectedIndexes()[0]
if not idx.isValid():
print "Error, index is not valid"
return
p = idx.parent()
if p.data().toString() in ("fragments/adducts:", "isotopic cluster:"):
p = idx.parent().parent().parent()
sample = self.currentSample[
1 if self.acTree is self.view.treeView_2 else
2] #self.model.sample(p.data().toString(), fullNameEntry=False)
if sample is None:
print "Error, sample not found %s" % sample.shortName()
return
#data=None
if len(idx.data().toString().split('/')) > 2:
data = idx.data().toString().split('/')[:-1]
else:
data = idx.data().toString().split('/')
peak = sample.peakAt(*map(float, data))
if peak is None:
self.view.showErrorMessage(
"Error", "Unable to find the corresponding peak...")
spl = MSSampleList()
s = MSSample("None")
s.rawPeaks = MSPeakList([peak], sample=s)
spl.addSample(s)
dial = MSIdentificationDialog(parent=self.view)
dial.setAttribute(Qt.WA_DeleteOnClose)
controller = MSIdentificationController(spl,
dial,
False,
showDirectResult=True)
def clusteringBASIC(peaks, adds, **k):
if not peaks:
return
t=time.clock()
errorRt = k.get('rtError', 6)
#ppm = float(kwargs.get('ppm'))/10**6
ppm = k.get('ppm')
if ppm is None:
try:
ppm = peaks[0].sample.ppm/1e6
except AttributeError:
print "No value found for ppm setting to 10/1E6"
ppm = 10./1e6
#mode = k.get('mode', 'HighRes')
resolveConflicts=k.get('resolveConflicts', False)
addsToCheck=np.array(adds.keys())
adductsFound = MSPeakList()
for i, p in enumerate(peaks):
a = MSClusterList()
for v in addsToCheck:
m = p.mz+v[0]
match = peaks.peaksInMZRTRange(m, p.rt, errorRt, deltam= 2 * ppm * m)
if match is None or not match:
continue
#take the closest in mass
goodP = sorted(match, key=lambda x:abs(x.mz - (p.mz + v[0])))[0]
#if goodP in set(adductsFound):
# if resolveConflicts:
# pass
#else:
if goodP is p:
continue
a.append(goodP)
goodP.parentPeak=p
adductsFound.append(goodP)
p.fragCluster=MSPeakList(set(a))#prevent from duplicates
# def clusterComparison(list_):#receive a list of peak with clusters identified
# """
# return the best peak
# WARNING: p_ydata and p_.y_data are None
# TODO:
#
# """
# sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
# longest=len(sortedList[-1].fragCluster)
# sameSizePeaks=MSPeakList()
#
# for p in sortedList:
# if len(p.fragCluster) == longest:
# sameSizePeaks.append(p)
#
# if len(sameSizePeaks) == 1:
# return sameSizePeaks[0]
# corr=np.array([0.] * len(sameSizePeaks))
# #for i, p in enumerate(sameSizePeaks):
# # for p_ in p.fragCluster:
# # corr[i] += r_coef(p_.y_data, p.y_data)
# m=max_f(corr)
# return sameSizePeaks[np.where(corr == m)[0][0]]
#
# if resolveConflicts:
# for add in set(adductsFound):
# if len(add.parentPeak) <= 1:
# #print "%s belong to several fragCluster"%str(add)
# continue
# #print "%s belong to several fragCluster"%str(add)
# goodParent=clusterComparison(add.parentPeak)
# #if goodParent is not None:
# # add.parentPeak = [goodParent]
#
# for parent in add.parentPeak:
# if parent != goodParent:
# try:
# parent.fragCluster.remove(add)
# except ValueError:
# print "Error removing %s from fragCluster of %s"%(str(add), str(parent))
# add.parentPeak = [goodParent] #the same of constructing a list 'toRemove then remove
# #print "after removing len add.parentPeak", len(add.parentPeak)
print "TiemElapsed: %s"%str(time.clock()-t)
return peaks, adductsFound
def clusteringCAMERA(peaks, adducts, **kwargs):
"""
arguments needed:
error_rt:rt_ drift
ppm:precision
useCorrelation: if we calculate correlations
"""
t=time.clock()
#unpack parameters
error_rt = kwargs.get('rtError', 6)
#ppm = float(kwargs.get('ppm'))/10**6
ppm=peaks[0].sample.ppm/1e6
mode=kwargs.get('mode', 'HighRes')
resolveConflicts=kwargs.get('resolveConflicts', False)
peaks_with_iso=peaks
print "peaklist length",len(peaks)
adducts_to_check=np.array(adducts.keys())
#===========================================================================
#START CAMERA ALGORITHM
print ("RT Grouping ...")
#RT_peak=peaks_with_iso.rtClustering(error_rt)
#3,find for each peak peaks which matches with retention time
rtPeak =[]
for i, peak in enumerate(peaks_with_iso.ipeaks()):
l=MSPeakList()
l.addPeak(peak)
for j, peak_ in enumerate(peaks_with_iso.ipeaks()):
if i!=j:
if abs(peak.rt - peak_.rt) < error_rt:
l.append(peak_)
isIncluded=False
index=[]
for k, rtClust in enumerate(rtPeak):
if set(l)<=(set(rtClust)):#inclusion test of l already in rt ? seen as 'equivalent to'
isIncluded=True
break
if set(rtClust) <= (set(l)):
index.append(k)
#break
#del rtPeak[index]
rtPeak= [rtPeak[i] for i in xrange(len(rtPeak)) if i not in index]
if not isIncluded:
rtPeak.append(MSPeakList(l))
#isIncluded=True
#else:
# if rtClust.__eq__(l):
# rtPeak[k]=l
# break
#isIncluded=True
#if not isIncluded:
#l.sort(key=lambda x:x.mass())
# with open('test1.txt', 'w') as f:
# for r in rtPeak:
# s=""
# for i, p in enumerate(r):
# s+=str(p)+';' if i<len(r)-1 else str(p)+'\n'
# f.write(s)
#EXPERIMENTAL CODE
# cl=[]
# for cluster in rtPeak:
# list_=[];datapoints={}
# for i, p in enumerate(cluster):
# correspondingPeaks=set()
# correspondingPeaks.add(p)
# for j in xrange(i+1, len(cluster)):
# #put caching on that to avoid recalculation each time of the datapoints
# try:
# r=r_coef(list(datapoints[p]), list(datapoints[cluster[j]]))
# except KeyError:
# y, y_= None, None
# try:
# y=datapoints[p]
# except KeyError:
# x, y= massExtractionBisectAlgo(p.sample,p.mass(), ppm)
# datapoints[p]=y
#
# try:
# y_=datapoints[cluster[j]]
# except KeyError:
# x, y_= massExtractionBisectAlgo(cluster[j].sample, cluster[j].mass(), ppm)
# datapoints[cluster[j]]=y_
# r=r_coef(y, y_)
# if r >= threshold:
# correspondingPeaks.add(cluster[j])
# list_.append(correspondingPeaks)
#
# for i, p in enumerate(list_):
# for j in xrange(i+1, len(list_)):
# if list_[j].issubset(p):
# continue
# else:
# cl.append(MSPeakList(list(p)))
#merging step again
# print "cluster length, same without replicates",len(cl), len(set(map(set, [x for x in cl])))
# with open('test2.txt', 'w') as f:
# for r in cl:
# s=""
# for i, p in enumerate(r):
# s+=str(p)+';' if i<len(r)-1 else str(p)+'\n'
# f.write(s)
#
#END EXPERIMENTAL CODE
print 'len RTpeak', len(rtPeak)
print ("Creating possible M0...")
#Cython code
finalList = massGenPerGroup(rtPeak, adducts_to_check, ppm)
print("Mapping of calculated mass on peaklist...")
#4,see if one matches with peak in the raw peaklist
goodPeak=[]#list will contain good peak per rtCluster
for i, dic in enumerate(finalList):
matchingMass=defaultdict(list)
for mass in dic.iterkeys():
p = rtPeak[i].peaksInMZRange(mass, deltam=mass * ppm if mode=='HighRes' else 1.)#rtPeak[i] not necessarily sorted warning
if not p:
continue
peak=sorted(p, key=lambda x:abs(mass - x.mass()))[0]
#if peak not in matchingMass.keys():#may avoid this to see if one peak appears several times !then do 'set'
# matchingMass[peak]=[]
matchingMass[peak] += dic[mass]
goodPeak.append(matchingMass)
#start new stuffs here
print ("Merging informations...")
#conflicts=False
adds=MSPeakList()#object sor storing adducts found
newGoodPeaks=defaultdict(list)#{}
for peaksInOneRtGroup in goodPeak:
for peak in peaksInOneRtGroup.iterkeys():
newGoodPeaks[peak] += peaksInOneRtGroup[peak]
for p in newGoodPeaks.iterkeys():
p.fragCluster=MSClusterList(list(set(newGoodPeaks[p])))
for f in p.fragCluster:
f.parentPeak.append(p)
adds += p.fragCluster
finalPeaks=MSPeakList(newGoodPeaks.keys())
print ("Resolving conflicts if any...")
#removing peak that appears many times that is to say in different clusters
def clusterComparison(list_):#receive a list of peak with clusters identified
"""
return the best peak
WARNING: p_ydata and p_.y_data are None
TODO:
"""
sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
longest=len(sortedList[-1].fragCluster)
sameSizePeaks=MSPeakList()
for p in sortedList:
if len(p.fragCluster) == longest:
sameSizePeaks.append(p)
if len(sameSizePeaks) == 1:
return sameSizePeaks[0]
corr=np.array([0.] * len(sameSizePeaks))
#for i, p in enumerate(sameSizePeaks):
# for p_ in p.fragCluster:
# corr[i] += r_coef(p_.y_data, p.y_data)
m=max_f(corr)
return sameSizePeaks[np.where(corr == m)[0][0]]
if resolveConflicts:
for add in set(adds):
if len(add.parentPeak) <= 1:
#print "%s belong to several fragCluster"%str(add)
continue
#print "%s belong to several fragCluster"%str(add)
goodParent=clusterComparison(add.parentPeak)
#if goodParent is not None:
# add.parentPeak = [goodParent]
for parent in add.parentPeak:
if parent != goodParent:
try:
parent.fragCluster.remove(add)
except ValueError:
print "Error removing %s from fragCluster of %s"%(str(add), str(parent))
add.parentPeak = [goodParent] #the same of constructing a list 'toRemove then remove
#print "after removing len add.parentPeak", len(add.parentPeak)
#make the annotation
for peak in finalPeaks.ipeaks():
for f in peak.fragCluster:
#results = makeAnnotations(adducts_to_check, adducts, f.mass(), ppm)
for annot in adducts.iterkeys():
p = f.mass() / annot[1] + annot[0]
diff = peak.mass()*ppm if mode =='HighRes' else 1
if peak.mass() > p-diff and peak.mass() < p+diff:
f.annotation[annot]=adducts[annot]
break
finalPeaks=checkingSons(finalPeaks)
#5,second filter, correlation on the isotopic cluster between samples
# if useCorrelation:
# print "Calculating correlation between samples..."
# interSamplesCorr(spl, **kwargs)
# print "Calculating correlation intra sample..."
# intraSampleCorr(spl)
# #6 merging
print "Merging interesting peaks"
for peak in peaks_with_iso.ipeaks():#wring merging must take out those which allow to construct this peak
if peak not in finalPeaks and peak not in adds:#matching_peaks:
finalPeaks.append(peak) #matching_peaks to
if not finalPeaks:
print ("no cluster found, please increase the ppm, or rt drift parameters")
print ("finished, time elapsed:",time.clock()-t)
return MSPeakList(sorted(finalPeaks, key=lambda x:x.mass)), adds#checkingSons(finalPeaks), adds
def isotopicPeakListFinder(peaks, isomasses, **kwargs):
"""
assign an isotopic cluster for each peak, and try to find an idms
we may use a system like the CAMERA algorithm to see...
input:
list of peak must an obj.MSPeakList object
clusterLength = 6 never go to six in LOW_RES
size expected of an isotopic cluster
rtError: maximum drift of the retention time
decreaseOrder: allow or not allow that the successive peak of the isotopic cluster
intensity are going down, can be confusing for finding idms
output:
two MSPeakList, the first one corresponding to the peaks with an isotopic cluster
and the other one all peaks belonging to an isotopic cluster
"""
#unpacking parameters
print "Isotopic cluster calculation..."
rtError = np.float(kwargs.get('rtError', 6))
ppm=np.float(peaks[0].sample.ppm/1e6)
MAX_GAP_ALLOWED = np.int(len(isomasses))
decreaseOrder = kwargs.get('decreaseOrder', True) #we use the less restrictive...
mode = kwargs.get('mode', 'Highres')
#sort isomasses
#isomasses = sorted(isomasses, key=lambda x:x[0])
peaks_with_iso =MSPeakList()
peaks_without_iso = MSPeakList()#peaks without isotopic cluster but which does not have a isotopic cluster
list_iso = set()#MSPeakList()
t = time.clock()
for peak in peaks.ipeaks():#iterating over peaks
if peak in list_iso:
continue#avoid to calculate for every peaks
isoCluster= MSClusterList()
gap = 0
#isos = resolutionAdjustment(isomasses, peak.mass()*ppm) if mode=='HighRes' else isomasses
for i, isomass in enumerate(sorted(isomasses, key=lambda x:x[0])):
#pic = _getMatchingPeaks(peaks, peak, isomass[0], ppm, rtError)
mass=isomass[0]
massToCheck=peak.mass()+mass
p = peaks.peaksInMZRange(massToCheck, deltam=ppm*massToCheck if mode=='HighRes' else 1.) #deltart
matchingRtPeaks = MSPeakList()#will contain all matching peak in rt
for pk in p.ipeaks():
if pk != peak:
if abs(peak.rt - pk.rt) <= rtError:
matchingRtPeaks.append(pk)
if matchingRtPeaks:
pic = sorted(matchingRtPeaks, key=lambda pics: abs(pics.mass()-peak.mass()))[0] #take the closest in mass
if pic is not None:
if decreaseOrder:#we want peak area inferior a peak
#if isoCluster:
areaToCompare=isoCluster[-1].area if isoCluster else peak.area
if areaToCompare < pic.area:#idms found ???
break
if pic not in list_iso:#pic not in isoCluster and
isoCluster.append(pic)
list_iso.add(pic)
else:
gap+=1
if gap >=MAX_GAP_ALLOWED:
break
# #set parent for all peaks found
if isoCluster:
for pics in isoCluster:
#pics.parentPeak=peak
pics.parentPeak.append(peak)
peak.isoCluster = isoCluster
peaks_with_iso.addPeak(peak)
else:
peaks_without_iso.addPeak(peak)
# for p in peaks.ipeaks():
# if p not in peaks_with_iso and p not in list_iso:
# peaks_without_iso.addPeak(p)
print time.clock()-t
print "peaks with isotopes: " ,len(peaks_with_iso)
print "list isotopes: " ,len(list_iso)
print "peaks without isotopes: " ,len(peaks_without_iso)
return peaks_with_iso+peaks_without_iso, list_iso
def clusteringBASIC(peaks, adds, **k):
if not peaks:
return
t = time.clock()
errorRt = k.get('rtError', 6)
#ppm = float(kwargs.get('ppm'))/10**6
ppm = k.get('ppm')
if ppm is None:
try:
ppm = peaks[0].sample.ppm / 1e6
except AttributeError:
print "No value found for ppm setting to 10/1E6"
ppm = 10. / 1e6
#mode = k.get('mode', 'HighRes')
resolveConflicts = k.get('resolveConflicts', False)
addsToCheck = np.array(adds.keys())
adductsFound = MSPeakList()
for i, p in enumerate(peaks):
a = MSClusterList()
for v in addsToCheck:
m = p.mz + v[0]
match = peaks.peaksInMZRTRange(m,
p.rt,
errorRt,
deltam=2 * ppm * m)
if match is None or not match:
continue
#take the closest in mass
goodP = sorted(match, key=lambda x: abs(x.mz - (p.mz + v[0])))[0]
#if goodP in set(adductsFound):
# if resolveConflicts:
# pass
#else:
if goodP is p:
continue
a.append(goodP)
goodP.parentPeak = p
adductsFound.append(goodP)
p.fragCluster = MSPeakList(set(a)) #prevent from duplicates
# def clusterComparison(list_):#receive a list of peak with clusters identified
# """
# return the best peak
# WARNING: p_ydata and p_.y_data are None
# TODO:
#
# """
# sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
# longest=len(sortedList[-1].fragCluster)
# sameSizePeaks=MSPeakList()
#
# for p in sortedList:
# if len(p.fragCluster) == longest:
# sameSizePeaks.append(p)
#
# if len(sameSizePeaks) == 1:
# return sameSizePeaks[0]
# corr=np.array([0.] * len(sameSizePeaks))
# #for i, p in enumerate(sameSizePeaks):
# # for p_ in p.fragCluster:
# # corr[i] += r_coef(p_.y_data, p.y_data)
# m=max_f(corr)
# return sameSizePeaks[np.where(corr == m)[0][0]]
#
# if resolveConflicts:
# for add in set(adductsFound):
# if len(add.parentPeak) <= 1:
# #print "%s belong to several fragCluster"%str(add)
# continue
# #print "%s belong to several fragCluster"%str(add)
# goodParent=clusterComparison(add.parentPeak)
# #if goodParent is not None:
# # add.parentPeak = [goodParent]
#
# for parent in add.parentPeak:
# if parent != goodParent:
# try:
# parent.fragCluster.remove(add)
# except ValueError:
# print "Error removing %s from fragCluster of %s"%(str(add), str(parent))
# add.parentPeak = [goodParent] #the same of constructing a list 'toRemove then remove
# #print "after removing len add.parentPeak", len(add.parentPeak)
print "TiemElapsed: %s" % str(time.clock() - t)
return peaks, adductsFound
def clusteringCAMERA(peaks, adducts, **kwargs):
"""
arguments needed:
error_rt:rt_ drift
ppm:precision
useCorrelation: if we calculate correlations
"""
t = time.clock()
#unpack parameters
error_rt = kwargs.get('rtError', 6)
#ppm = float(kwargs.get('ppm'))/10**6
ppm = peaks[0].sample.ppm / 1e6
mode = kwargs.get('mode', 'HighRes')
resolveConflicts = kwargs.get('resolveConflicts', False)
peaks_with_iso = peaks
print "peaklist length", len(peaks)
adducts_to_check = np.array(adducts.keys())
#===========================================================================
#START CAMERA ALGORITHM
print("RT Grouping ...")
#RT_peak=peaks_with_iso.rtClustering(error_rt)
#3,find for each peak peaks which matches with retention time
rtPeak = []
for i, peak in enumerate(peaks_with_iso.ipeaks()):
l = MSPeakList()
l.addPeak(peak)
for j, peak_ in enumerate(peaks_with_iso.ipeaks()):
if i != j:
if abs(peak.rt - peak_.rt) < error_rt:
l.append(peak_)
isIncluded = False
index = []
for k, rtClust in enumerate(rtPeak):
if set(l) <= (
set(rtClust)
): #inclusion test of l already in rt ? seen as 'equivalent to'
isIncluded = True
break
if set(rtClust) <= (set(l)):
index.append(k)
#break
#del rtPeak[index]
rtPeak = [rtPeak[i] for i in xrange(len(rtPeak)) if i not in index]
if not isIncluded:
rtPeak.append(MSPeakList(l))
#isIncluded=True
#else:
# if rtClust.__eq__(l):
# rtPeak[k]=l
# break
#isIncluded=True
#if not isIncluded:
#l.sort(key=lambda x:x.mass())
# with open('test1.txt', 'w') as f:
# for r in rtPeak:
# s=""
# for i, p in enumerate(r):
# s+=str(p)+';' if i<len(r)-1 else str(p)+'\n'
# f.write(s)
#EXPERIMENTAL CODE
# cl=[]
# for cluster in rtPeak:
# list_=[];datapoints={}
# for i, p in enumerate(cluster):
# correspondingPeaks=set()
# correspondingPeaks.add(p)
# for j in xrange(i+1, len(cluster)):
# #put caching on that to avoid recalculation each time of the datapoints
# try:
# r=r_coef(list(datapoints[p]), list(datapoints[cluster[j]]))
# except KeyError:
# y, y_= None, None
# try:
# y=datapoints[p]
# except KeyError:
# x, y= massExtractionBisectAlgo(p.sample,p.mass(), ppm)
# datapoints[p]=y
#
# try:
# y_=datapoints[cluster[j]]
# except KeyError:
# x, y_= massExtractionBisectAlgo(cluster[j].sample, cluster[j].mass(), ppm)
# datapoints[cluster[j]]=y_
# r=r_coef(y, y_)
# if r >= threshold:
# correspondingPeaks.add(cluster[j])
# list_.append(correspondingPeaks)
#
# for i, p in enumerate(list_):
# for j in xrange(i+1, len(list_)):
# if list_[j].issubset(p):
# continue
# else:
# cl.append(MSPeakList(list(p)))
#merging step again
# print "cluster length, same without replicates",len(cl), len(set(map(set, [x for x in cl])))
# with open('test2.txt', 'w') as f:
# for r in cl:
# s=""
# for i, p in enumerate(r):
# s+=str(p)+';' if i<len(r)-1 else str(p)+'\n'
# f.write(s)
#
#END EXPERIMENTAL CODE
print 'len RTpeak', len(rtPeak)
print("Creating possible M0...")
#Cython code
finalList = massGenPerGroup(rtPeak, adducts_to_check, ppm)
print("Mapping of calculated mass on peaklist...")
#4,see if one matches with peak in the raw peaklist
goodPeak = [] #list will contain good peak per rtCluster
for i, dic in enumerate(finalList):
matchingMass = defaultdict(list)
for mass in dic.iterkeys():
p = rtPeak[i].peaksInMZRange(
mass, deltam=mass * ppm if mode == 'HighRes' else
1.) #rtPeak[i] not necessarily sorted warning
if not p:
continue
peak = sorted(p, key=lambda x: abs(mass - x.mass()))[0]
#if peak not in matchingMass.keys():#may avoid this to see if one peak appears several times !then do 'set'
# matchingMass[peak]=[]
matchingMass[peak] += dic[mass]
goodPeak.append(matchingMass)
#start new stuffs here
print("Merging informations...")
#conflicts=False
adds = MSPeakList() #object sor storing adducts found
newGoodPeaks = defaultdict(list) #{}
for peaksInOneRtGroup in goodPeak:
for peak in peaksInOneRtGroup.iterkeys():
newGoodPeaks[peak] += peaksInOneRtGroup[peak]
for p in newGoodPeaks.iterkeys():
p.fragCluster = MSClusterList(list(set(newGoodPeaks[p])))
for f in p.fragCluster:
f.parentPeak.append(p)
adds += p.fragCluster
finalPeaks = MSPeakList(newGoodPeaks.keys())
print("Resolving conflicts if any...")
#removing peak that appears many times that is to say in different clusters
def clusterComparison(
list_): #receive a list of peak with clusters identified
"""
return the best peak
WARNING: p_ydata and p_.y_data are None
TODO:
"""
sortedList = sorted(list_, key=lambda x: len(x.fragCluster))
longest = len(sortedList[-1].fragCluster)
sameSizePeaks = MSPeakList()
for p in sortedList:
if len(p.fragCluster) == longest:
sameSizePeaks.append(p)
if len(sameSizePeaks) == 1:
return sameSizePeaks[0]
corr = np.array([0.] * len(sameSizePeaks))
#for i, p in enumerate(sameSizePeaks):
# for p_ in p.fragCluster:
# corr[i] += r_coef(p_.y_data, p.y_data)
m = max_f(corr)
return sameSizePeaks[np.where(corr == m)[0][0]]
if resolveConflicts:
for add in set(adds):
if len(add.parentPeak) <= 1:
#print "%s belong to several fragCluster"%str(add)
continue
#print "%s belong to several fragCluster"%str(add)
goodParent = clusterComparison(add.parentPeak)
#if goodParent is not None:
# add.parentPeak = [goodParent]
for parent in add.parentPeak:
if parent != goodParent:
try:
parent.fragCluster.remove(add)
except ValueError:
print "Error removing %s from fragCluster of %s" % (
str(add), str(parent))
add.parentPeak = [
goodParent
] #the same of constructing a list 'toRemove then remove
#print "after removing len add.parentPeak", len(add.parentPeak)
#make the annotation
for peak in finalPeaks.ipeaks():
for f in peak.fragCluster:
#results = makeAnnotations(adducts_to_check, adducts, f.mass(), ppm)
for annot in adducts.iterkeys():
p = f.mass() / annot[1] + annot[0]
diff = peak.mass() * ppm if mode == 'HighRes' else 1
if peak.mass() > p - diff and peak.mass() < p + diff:
f.annotation[annot] = adducts[annot]
break
finalPeaks = checkingSons(finalPeaks)
#5,second filter, correlation on the isotopic cluster between samples
# if useCorrelation:
# print "Calculating correlation between samples..."
# interSamplesCorr(spl, **kwargs)
# print "Calculating correlation intra sample..."
# intraSampleCorr(spl)
# #6 merging
print "Merging interesting peaks"
for peak in peaks_with_iso.ipeaks(
): #wring merging must take out those which allow to construct this peak
if peak not in finalPeaks and peak not in adds: #matching_peaks:
finalPeaks.append(peak) #matching_peaks to
if not finalPeaks:
print(
"no cluster found, please increase the ppm, or rt drift parameters"
)
print("finished, time elapsed:", time.clock() - t)
return MSPeakList(
sorted(finalPeaks,
key=lambda x: x.mass)), adds #checkingSons(finalPeaks), adds
def isotopicPeakListFinder(peaks, isomasses, **kwargs):
"""
assign an isotopic cluster for each peak, and try to find an idms
we may use a system like the CAMERA algorithm to see...
input:
list of peak must an obj.MSPeakList object
clusterLength = 6 never go to six in LOW_RES
size expected of an isotopic cluster
rtError: maximum drift of the retention time
decreaseOrder: allow or not allow that the successive peak of the isotopic cluster
intensity are going down, can be confusing for finding idms
output:
two MSPeakList, the first one corresponding to the peaks with an isotopic cluster
and the other one all peaks belonging to an isotopic cluster
"""
#unpacking parameters
print "Isotopic cluster calculation..."
rtError = np.float(kwargs.get('rtError', 6))
ppm = np.float(peaks[0].sample.ppm / 1e6)
MAX_GAP_ALLOWED = np.int(len(isomasses))
decreaseOrder = kwargs.get('decreaseOrder',
True) #we use the less restrictive...
mode = kwargs.get('mode', 'Highres')
#sort isomasses
#isomasses = sorted(isomasses, key=lambda x:x[0])
peaks_with_iso = MSPeakList()
peaks_without_iso = MSPeakList(
) #peaks without isotopic cluster but which does not have a isotopic cluster
list_iso = set() #MSPeakList()
t = time.clock()
for peak in peaks.ipeaks(): #iterating over peaks
if peak in list_iso:
continue #avoid to calculate for every peaks
isoCluster = MSClusterList()
gap = 0
#isos = resolutionAdjustment(isomasses, peak.mass()*ppm) if mode=='HighRes' else isomasses
for i, isomass in enumerate(sorted(isomasses, key=lambda x: x[0])):
#pic = _getMatchingPeaks(peaks, peak, isomass[0], ppm, rtError)
mass = isomass[0]
massToCheck = peak.mass() + mass
p = peaks.peaksInMZRange(
massToCheck,
deltam=ppm *
massToCheck if mode == 'HighRes' else 1.) #deltart
matchingRtPeaks = MSPeakList(
) #will contain all matching peak in rt
for pk in p.ipeaks():
if pk != peak:
if abs(peak.rt - pk.rt) <= rtError:
matchingRtPeaks.append(pk)
if matchingRtPeaks:
pic = sorted(matchingRtPeaks,
key=lambda pics: abs(pics.mass() - peak.mass()))[
0] #take the closest in mass
if pic is not None:
if decreaseOrder: #we want peak area inferior a peak
#if isoCluster:
areaToCompare = isoCluster[
-1].area if isoCluster else peak.area
if areaToCompare < pic.area: #idms found ???
break
if pic not in list_iso: #pic not in isoCluster and
isoCluster.append(pic)
list_iso.add(pic)
else:
gap += 1
if gap >= MAX_GAP_ALLOWED:
break
# #set parent for all peaks found
if isoCluster:
for pics in isoCluster:
#pics.parentPeak=peak
pics.parentPeak.append(peak)
peak.isoCluster = isoCluster
peaks_with_iso.addPeak(peak)
else:
peaks_without_iso.addPeak(peak)
# for p in peaks.ipeaks():
# if p not in peaks_with_iso and p not in list_iso:
# peaks_without_iso.addPeak(p)
print time.clock() - t
print "peaks with isotopes: ", len(peaks_with_iso)
print "list isotopes: ", len(list_iso)
print "peaks without isotopes: ", len(peaks_without_iso)
return peaks_with_iso + peaks_without_iso, list_iso