def create_seghists_from_eventsegs(esegs, histScores): #Note that the edges must be sorted by segstr.
totalp=ecycles.compute_totalp(histScores)
myseghists=[]
firstseg=SegmentHistory(esegs[0])
working_seghists=[firstseg]
lastseg=None #keep track of the last location for segments
if firstseg.start < firstseg.end: lastseg=firstseg
for e in esegs[1:]:
# sys.stderr.write("eseg\t%sworking_seghists: %d\tmyseghists: %d\n" % (str(e), len(working_seghists), len(myseghists)))
tmpseghists=[]
merged=False
for seghist in working_seghists:
# sys.stderr.write("working:" + str(seghist))
# if it's a break the start and end will be equal.
# We don't want to merge breaks with segments, only with equivalent breaks.
if ((e.start == e.end and seghist.samelocus(e)) or
((e.start < e.end) and (seghist.start < seghist.end) and seghist.overlaps(e))):
new_seghists=merge_in_edge(e, seghist)
tmpseghists+=new_seghists
merged=True
# sys.stderr.write("samelocus %d\n%s%s" % (len(tmpseghists), str(e), str(seghist)))
elif seghist.overlaps(e) or seghist.comes_after(e):
tmpseghists.append(seghist)
elif seghist.comes_before(e):
myseghists.append(seghist)
if merged==True and (e.end > e.start) and (e.end > lastseg.end):
newseg=SegmentHistory(e)
newseg.start=lastseg.end+1
tmpseghists.append(newseg)
lastseg=newseg
merged=True
elif merged==False:
newseg=SegmentHistory(e)
tmpseghists.append(newseg)
if ((e.start < e.end) and
(((lastseg is not None) and (lastseg.comes_before(e))) or
(lastseg is None))):
lastseg=newseg
working_seghists=tmpseghists
myseghists+=working_seghists
# return myseghists
sys.stderr.write("done making seghists, now creating the CN profiles.\n")
datfh=open("seghists.dat", 'w')
for s in myseghists:
datfh.write("%s\t%d\t%d\t%d\n" % (s.chr, s.start, s.end, len(s.esegs)))
create_CNprofiles_from_Edges(s, histScores, totalp)
sys.stderr.write("done making CN profiles\n")
datfh.close()
return(myseghists)
def create_seghists_from_eventsegs(esegs, histScores): # Note that the edges must be sorted by segstr.
totalp = ecycles.compute_totalp(histScores)
myseghists = []
firstseg = SegmentHistory(esegs[0])
working_seghists = [firstseg]
lastseg = None # keep track of the last location for segments
if firstseg.start < firstseg.end:
lastseg = firstseg
for e in esegs[1:]:
# sys.stderr.write("eseg\t%sworking_seghists: %d\tmyseghists: %d\n" % (str(e), len(working_seghists), len(myseghists)))
tmpseghists = []
merged = False
for seghist in working_seghists:
# sys.stderr.write("working:" + str(seghist))
# if it's a break the start and end will be equal.
# We don't want to merge breaks with segments, only with equivalent breaks.
if (e.start == e.end and seghist.samelocus(e)) or (
(e.start < e.end) and (seghist.start < seghist.end) and seghist.overlaps(e)
):
new_seghists = merge_in_edge(e, seghist)
tmpseghists += new_seghists
merged = True
# sys.stderr.write("samelocus %d\n%s%s" % (len(tmpseghists), str(e), str(seghist)))
elif seghist.overlaps(e) or seghist.comes_after(e):
tmpseghists.append(seghist)
elif seghist.comes_before(e):
myseghists.append(seghist)
if merged == True and (e.end > e.start) and (e.end > lastseg.end):
newseg = SegmentHistory(e)
newseg.start = lastseg.end + 1
tmpseghists.append(newseg)
lastseg = newseg
merged = True
elif merged == False:
newseg = SegmentHistory(e)
tmpseghists.append(newseg)
if (e.start < e.end) and (((lastseg is not None) and (lastseg.comes_before(e))) or (lastseg is None)):
lastseg = newseg
working_seghists = tmpseghists
myseghists += working_seghists
# return myseghists
sys.stderr.write("done making seghists, now creating the CN profiles.\n")
datfh = open("seghists.dat", "w")
for s in myseghists:
datfh.write("%s\t%d\t%d\t%d\n" % (s.chr, s.start, s.end, len(s.esegs)))
create_CNprofiles_from_Edges(s, histScores, totalp)
sys.stderr.write("done making CN profiles\n")
datfh.close()
return myseghists
def create_CNprofiles_from_Edges(self, histScores, totalp=0):
edgelist = self.esegs
mycnprofs = []
if totalp == 0:
totalp = ecycles.compute_totalp(historyScores)
(hprofiles, pprofiles) = create_profile_matrices(edgelist, histScores)
(ridx, profiles) = get_unique_rows(hprofiles)
goodi = np.where(np.sum(profiles != 0, axis=1) > 0)[0]
for i in goodi:
mycnp = CNprofile()
mycnp.likelihood = ecycles.compute_likelihood_histories(histScores[ridx == i, 0], histScores, totalp)
mycnp.numhists = np.sum(ridx == i)
cnvals = profiles[i, :]
pvals = pprofiles[ridx == i, :]
pvalsd = np.std(pvals, axis=0)
pvalsm = np.mean(pvals, axis=0)
mycnp.pvals = pvalsm[cnvals != 0].tolist()
mycnp.pvalsd = pvalsd[cnvals != 0].tolist()
mycnp.cnvals = cnvals[cnvals != 0].tolist()
mycnprofs.append(mycnp)
self.CNprofiles = mycnprofs
def create_CNprofiles_from_Edges(self, histScores, totalp=0):
edgelist = self.esegs
mycnprofs = []
if totalp == 0: totalp = ecycles.compute_totalp(historyScores)
(hprofiles, pprofiles) = create_profile_matrices(edgelist, histScores)
(ridx, profiles) = get_unique_rows(hprofiles)
goodi = np.where(np.sum(profiles != 0, axis=1) > 0)[0]
for i in goodi:
mycnp = CNprofile()
mycnp.likelihood = ecycles.compute_likelihood_histories(
histScores[ridx == i, 0], histScores, totalp)
mycnp.numhists = np.sum(ridx == i)
cnvals = profiles[i, :]
pvals = pprofiles[ridx == i, :]
pvalsd = np.std(pvals, axis=0)
pvalsm = np.mean(pvals, axis=0)
mycnp.pvals = pvalsm[cnvals != 0].tolist()
mycnp.pvalsd = pvalsd[cnvals != 0].tolist()
mycnp.cnvals = cnvals[cnvals != 0].tolist()
mycnprofs.append(mycnp)
self.CNprofiles = mycnprofs
