def test_entry_creation(self):
entry1_1 = bed.BedEntry('X', 8000, 9000)
entry1_1.signalValue = 10
entry1_2 = bed.BedEntry('X', 80, 900)
entry1_2.signalValue = 3
bed1 = bed.BedFile([entry1_1, entry1_2])
entry2_1 = bed.BedEntry('X', 8500, 9000)
entry2_2 = bed.BedEntry('X', 80, 900)
entry2_1.signalValue = 10
entry2_2.signalValue = 3
bed2 = bed.BedFile([entry2_1, entry2_2])
entry3_1 = bed.BedEntry('X', 7500, 9000)
entry3_2 = bed.BedEntry('X', 80, 900)
entry3_1.signalValue = 10
entry3_2.signalValue = 3
bed3 = bed.BedFile([entry3_1, entry3_2])
entry4_1 = bed.BedEntry('X', 5000, 8999)
entry4_2 = bed.BedEntry('X', 80, 900)
entry4_1.signalValue = 10
entry4_2.signalValue = 3
bed4 = bed.BedFile([entry4_1, entry4_2])
unions = union([bed1, bed2, bed3, bed4], 2)
for fragment in unions[0]:
print(fragment)
def test_single_chrom(self):
med1 = bed.BedFile("test/data/med/med1_peaks.broadPeak", "Peaks").getChrom("chr17")
med2 = bed.BedFile("test/data/med/med2_peaks.broadPeak", "Peaks").getChrom("chr17")
med3 = bed.BedFile("test/data/med/med3_peaks.broadPeak", "Peaks").getChrom("chr17")
bedf = [med1, med2, med3]
minentries = 2
# First create intersection and rank the entries in each replicate and return the rankproduct values
ranks = rankreps(bedf, minentries, rankmethod='signalValue')
# Calculate rank product for each entry that contributes to a union entry
# Calculate the pvalues of the rank product values
print('Calculating rank product probabilities...')
rpb_up = rankprodbounds(ranks[1], len(ranks[1]), len(bedf), 'geometric')
print('Calculating binomial threshold...')
# Calculate rpb and binomial intersection point
Pks = thresholdCalc(rpb_up, k=len(bedf) - (minentries - 1))
if len(Pks[2]) != 0:
binomAlpha = round(min(Pks[2]), 3)
else:
print('No binomial convergence, defaulting to 0.1')
binomAlpha = 0.1
# Perform multiple hypothesis testing correction upon the pvals
fdr = multipletesting.fdrcorrection(rpb_up)
# Determine whether to remove entries that are called significant
print('Cleaning up output...')
for i, v in enumerate(ranks[0][0]):
p = rpb_up[i]
if p != 0.0:
ranks[0][0][i].addOption(name='TBD',
score=min([abs(int(125 * math.log2(rpb_up[i]))), 1000]),
strand='.',
pValue=rpb_up[i],
qValue=fdr[1][i])
else:
ranks[0][0][i].addOption(name='TBD',
score=1000,
strand='.',
pValue=2.5e-20,
qValue=2.5e-20)
collapsed = bed.BedFile(ranks[0][0], 'IDR')
connected = connect_entries(collapsed, bedf, 20, True)
self.assertNotEqual(len(collapsed), len(connected))
def joinChIA(chia):
interactions = []
for rep in chia:
for intr in rep:
interactions.append(intr)
pe = bed.BedFile(interactions, 'bedpe')
return pe
def run_1(self, args):
for i in args.input:
print(str(i.name))
print('Processing Input...')
bedfs = [bed.BedFile(str(i.name), 'Peaks') for i in args.input]
rankprod.performrankprod(bedfs, args.minentries, args.rankmethod,
'all', args.duphandling, args.random_seed,
args.alpha, args.output, args.size, False,
args.fragment)
def test_from_bed(self):
med1 = bed.BedFile("test/data/med/med1_peaks.broadPeak", "Peaks")
med2 = bed.BedFile("test/data/med/med2_peaks.broadPeak", "Peaks")
med3 = bed.BedFile("test/data/med/med3_peaks.broadPeak", "Peaks")
bedf = [med1, med2, med3]
performrankprod(bedf, minentries=2, rankmethod="pvalue", specifyMax=None,
duphandling='average', random_seed=0.5,
alpha=0.05,
filename="test_fragments_true",
default_min_peak=20,
print_pvals=True,
fragment=True)
performrankprod(bedf, minentries=2, rankmethod="pvalue", specifyMax=None,
duphandling='average', random_seed=0.5,
alpha=0.05,
filename="test_fragments_false",
default_min_peak=20,
print_pvals=True,
fragment=False)
def ChIAreproducibility(chia,
chip,
minentries=None,
rank='signalValue',
threshold='all',
alpha=0.05,
filename='NA'):
if minentries is None:
minentries = len(chip)
if len(chia) == 1 and len(chip) == 1:
return verifyChIAPeaks(chia[0],
chip[0],
filename=filename,
alpha=alpha)
else:
try: # Check to see if multiple chiapet files are entered
chia[0]
inters = joinChIA(chia)
except TypeError:
inters = chia
rankprod.performrankprod(chip,
minentries=minentries,
rankmethod=rank,
alpha=alpha,
filename=filename + '.bed')
if threshold == 'binom':
RP = bed.BedFile('T2_' + filename + '.bed', 'Peaks')
return verifyChIAPeaks(inters, RP, filename=filename, alpha=alpha)
elif threshold == 'alpha':
RP = bed.BedFile('T1_' + filename + '.bed', 'Peaks')
return verifyChIAPeaks(inters, RP, filename=filename, alpha=alpha)
elif threshold == 'all':
RP = bed.BedFile('ALL_' + filename + '.bed', 'Peaks')
return verifyChIAPeaks(inters, RP, filename=filename, alpha=alpha)
def test_bed(self):
bf = bed.BedFile("test/data/med/med1_peaks.broadPeak", "Peaks")
print(bf.chroms.keys())
g = bf.generate('chr1')
print(next(g))
print(next(g))
print(next(g))
cnt = 0
for entry in bf:
cnt += 1
print(str(cnt) + '\t' + str(entry))
if cnt == 100:
break
entry1 = bed.BedEntry('chrX', 3858266, 3858530)
print(entry1 in bf)
entry2 = bed.BedEntry('chrX', 10047550, 10067694)
for x in bf.getOverlap(entry2):
print(x)
entry3 = bed.BedEntry('chr9', 102699903, 102700167)
for x in bf.getClosest(entry3):
print(x)
for y in x:
print(y)
def verifyChIAPeaks(chiaData, chipData, filename, alpha):
listinters = []
PETs = []
probs = []
inters = {}
PETgroups = {}
prs = []
fdr = []
RP = []
if getattr(chiaData, 'format') == 'BED12':
chiaData = bed.BED12toBEDPE(chiaData)
for link in chiaData:
if link.PETs <= 2:
continue
o1 = chipData.getOverlap(link.partner1)
o2 = chipData.getOverlap(link.partner2)
try:
PETgroups[link.PETs].append(link)
except KeyError:
PETgroups[link.PETs] = [link]
try:
if len(o1) > 0 and len(o2) > 0:
for peaks1, peaks2 in zip(o1, o2):
# Create entries in dictionary with peaks as support
try:
inters[link].append([peaks1, peaks2])
except KeyError:
inters[link] = [peaks1, peaks2]
except TypeError:
continue
dists = []
for k, v in inters.items():
pvals = []
rs = []
# if k.PETs <= 1:
# continue
# kvals = []
for peak in v:
try:
pvals.append(peak.pValue)
rs.append(peak.signalValue)
# combined = scipy.stats.combine_pvalues([peak1.pValue, peak2.pValue], 'fisher')
# pvals.append(combined[1])
except AttributeError:
if type(peak) == 'list':
for p in peak:
try:
pvals.append(p.pValue)
rs.append(peak.signalValue)
except AttributeError:
continue
else:
continue
if len(pvals) != 0:
combined = scipy.stats.combine_pvalues(pvals, 'stouffer', rs)
probs.append(combined[1])
# rs = iterflatten(rs)
# RP.append(rs)
listinters.append(k)
PETs.append(k.PETs)
dists.append(k.getDistance())
# probcor = multipletesting.fdrcorrection(probs, alpha)
PETs, listinters, probcor, dists = (list(x) for x in zip(
*sorted(zip(PETs, listinters, probs, dists),
key=lambda pair: pair[0],
reverse=True)))
bins = np.arange(min(dists), max(dists), max(dists) / 50, dtype=int)
brobs = []
ints = []
pets2 = []
ainters = []
tags = 0
# for bin in range(1, 50):
# pets = []
# lints = []
# ps = []
# ds = []
# # rp = []
# for idx, j in enumerate(np.digitize(dists, bins)):
# if j == bin:
# pets.append(PETs[idx])
# lints.append(listinters[idx])
# ps.append(probcor[idx])
# ds.append(dists[idx])
# rp.append(RP[idx])
# try:
# pets, lints, ps, ds, rp = (list(x) for x in zip(*sorted(zip(pets, lints, ps, ds, rp),
# key=lambda pair: pair[0], reverse=True)))
# except ValueError:
# continue
#
n = len(RP)
# k = sum(pets)
t = sum(PETs)
# brobs = []
tags = 0
# RPranks = scipy.stats.rankdata(RP)
# mRPrank = max(RPranks)
# RPranks1 = [mRPrank+1 - x for x in RPranks]
# probranks = scipy.stats.rankdata(probcor)
# mprobs = max(probranks)
# probranks1 = [mprobs+1 -x for x in probranks]
# # distranks = scipy.stats.rankdata(dists)
# # mdist = max(distranks)
# # distranks1 = [mdist+1 - x for x in distranks]
# # PETranks1 = scipy.stats.rankdata(PETs)
# # mpet = max(PETranks1)
# rankprods = []
# # PETranks = [mpet + 1 - x for x in PETranks1]
# PETdists = [x/y for x,y in zip(dists, PETs)]
# PDranks = scipy.stats.rankdata(PETdists)
# mPD = max(PDranks)
# PDranks1 = [mPD+1 - x for x in PDranks]
# for rank in range(len(RP)):
# print(PDranks[rank], PETdists[rank])
# try:
# rankprods.append(PDranks1[rank]*RPranks1[rank])
# pets2.append(PETs[rank])
# except IndexError:
# continue
# print(rankprods)
# rpb = rankprodbounds(rankprods, len(rankprods), 2, 'geometric')
# for p in rpb:
# brobs.append(p)
# print(rpb)
# mRP = max(rankprods)
# mPETs = max(PETs)
# mdists = max(dists)
for i, p in enumerate(zip(probcor, PETs)):
tags += p[1]
b = scipy.stats.binom.cdf(tags, t, 1 - p[0])
# b2 = scipy.stats.combine_pvalues([b, [i]], 'stouffer', [(mPETs-PETs[i])/mPETs, (mRP-rankprods[i])/mRP])
print(p[0], p[1])
brobs.append(b)
pets2.append(p[1])
ints.append(listinters[i])
# # secinters = []
# # secPETs = []
# # secprobcor = []
# print(rpb)
# ints, probs = (list(x) for x in zip(*sorted(zip(ainters, probs), key=lambda pair: pair[1], reverse=False)))
corrected = multipletesting.fdrcorrection(brobs, alpha)
thresh = rankprod.thresholdCalc(brobs)
try:
binomAlpha = round(min(thresh[2]), 3)
if binomAlpha == 0:
binomAlpha = 0.05
except ValueError:
binomAlpha = 0.05
# brobs.append(rpb)
print(binomAlpha)
for i, p in zip(ints, corrected[1]):
i.addOption(pValue=p)
prs.append(p)
ainters.append(i)
# print(p, binomAlpha)
if p <= 0.05:
if p == 0:
i.addOption(pValue=0.00000000000001)
fdr.append(i)
# else:
# secinters.append(i)
# secPETs.append(pet)
# secprobcor.append(pc)
# n = sum(secPETs)
# brobs = []
# tags = 0
# for i, p in zip(secPETs, secprobcor):
# tags += i
# b = scipy.stats.binom.cdf(tags, n, p)
# brobs.append(b)
#
# corrected = multipletesting.fdrcorrection(brobs, alpha)
# for i, p in zip(listinters, brobs):
# i.addOption(pValue=p)
# if p <= alpha:
# if p == 0:
# i.addOption(pValue=0.00000000000001)
# fdr.append(i)
fdr = bed.BedFile(fdr, "BEDPE")
allinters = bed.BedFile(ainters, "BEDPE")
print(len(fdr), 'Interactions pass FDR threshold')
bed.writeBedFile(allinters, filename + '.bedpe', format='BEDPE')
bed.writeBedFile(fdr, filename + '.bedpe', format="BEDPE")
intersBED12 = bed.BEDPEtoBED12(allinters)
bed.writeBedFile(intersBED12, filename + '.bed', format='BED12')
return ainters, fdr, allinters, PETs, dists, brobs, corrected[1]
def verifyChIAPeaks(chiaData, chipData, filename, alpha):
listinters = []
PETs = []
probs = []
inters = {}
PETgroups = {}
prs = []
fdr = []
RP = []
if getattr(chiaData, 'format') == 'BED12':
chiaData = bed.BED12toBEDPE(chiaData)
for link in chiaData:
if link.PETs <= 2:
continue
o1 = chipData.getOverlap(link.partner1)
o2 = chipData.getOverlap(link.partner2)
try:
PETgroups[link.PETs].append(link)
except KeyError:
PETgroups[link.PETs] = [link]
try:
if len(o1) > 0 and len(o2) > 0:
for peaks1, peaks2 in zip(o1, o2):
# Create entries in dictionary with peaks as support
try:
inters[link].append([peaks1, peaks2])
except KeyError:
inters[link] = [peaks1, peaks2]
except TypeError:
continue
dists = []
for k, v in inters.items():
pvals = []
rs = []
# if k.PETs <= 1:
# continue
# kvals = []
for peak in v:
try:
pvals.append(peak.pValue)
rs.append(peak.signalValue)
# combined = scipy.stats.combine_pvalues([peak1.pValue, peak2.pValue], 'fisher')
# pvals.append(combined[1])
except AttributeError:
if type(peak) == 'list':
for p in peak:
try:
pvals.append(p.pValue)
rs.append(peak.signalValue)
except AttributeError:
continue
else:
continue
if len(pvals) != 0:
combined = scipy.stats.combine_pvalues(pvals, 'stouffer', rs)
probs.append(combined[1])
# rs = iterflatten(rs)
# RP.append(rs)
listinters.append(k)
PETs.append(k.PETs)
dists.append(k.getDistance())
# probcor = multipletesting.fdrcorrection(probs, alpha)
PETs, listinters, probcor, dists = (list(x) for x in zip(
*sorted(zip(PETs, listinters, probs, dists),
key=lambda pair: pair[0],
reverse=True)))
bins = np.arange(min(dists), max(dists), max(dists) / 50, dtype=int)
brobs = []
ints = []
pets2 = []
ainters = []
tags = 0
n = len(RP)
# k = sum(pets)
t = sum(PETs)
# brobs = []
tags = 0
for i, p in enumerate(zip(probcor, PETs)):
tags += p[1]
b = scipy.stats.binom.cdf(tags, t, 1 - p[0])
print(p[0], p[1])
brobs.append(b)
pets2.append(p[1])
ints.append(listinters[i])
corrected = multipletesting.fdrcorrection(brobs, alpha)
thresh = rankprod.thresholdCalc(brobs)
try:
binomAlpha = round(min(thresh[2]), 3)
if binomAlpha == 0:
binomAlpha = 0.05
except ValueError:
binomAlpha = 0.05
# brobs.append(rpb)
print(binomAlpha)
for i, p in zip(ints, corrected[1]):
i.addOption(pValue=p)
prs.append(p)
ainters.append(i)
# print(p, binomAlpha)
if p <= 0.05:
if p == 0:
i.addOption(pValue=0.00000000000001)
fdr.append(i)
fdr = bed.BedFile(fdr, "BEDPE")
allinters = bed.BedFile(ainters, "BEDPE")
print(len(fdr), 'Interactions pass FDR threshold')
bed.writeBedFile(allinters, filename + '.bedpe', format='BEDPE')
bed.writeBedFile(fdr, filename + '.bedpe', format="BEDPE")
intersBED12 = bed.BEDPEtoBED12(allinters)
bed.writeBedFile(intersBED12, filename + '.bed', format='BED12')
return ainters, fdr, allinters, PETs, dists, brobs, corrected[1]
