def binProbGt(k, size, prob):
""" binomial probability that x is > k (up to n). The corresponding R code for this is: pbinom(k, size = n, prob = p, lower=F) """
# -- manually, not exact enough
#sum = 0.0
#for i in range(0, k):
#sum+=binProb(n, p, i)
#return sum
# -- using scipy, not exact enough:
# 1.0 - cdf is not as exact as sf
#return 1.0 - scipy.stats.distributions.binom.cdf(k-1, n, p)
# scipy is too complicated to compile on the cluster
#return scipy.stats.distributions.binom.sf(k, size, prob)
return dist.pbinom(k, size,prob)
def binProbGt(k, size, prob):
""" binomial probability that x is > k (up to n). The corresponding R code for this is: pbinom(k, size = n, prob = p, lower=F) """
# -- manually, not exact enough
#sum = 0.0
#for i in range(0, k):
#sum+=binProb(n, p, i)
#return sum
# -- using scipy, not exact enough:
# 1.0 - cdf is not as exact as sf
#return 1.0 - scipy.stats.distributions.binom.cdf(k-1, n, p)
# scipy is too complicated to compile on the cluster
#return scipy.stats.distributions.binom.sf(k, size, prob)
return dist.pbinom(k, size, prob)
def __next(self, a, b, obsOcc):
width = b - a + 1
if width < self.MIN_WIDTH or width > self.MAX_WIDTH:
return
if obsOcc < self.MIN_OCC or obsOcc > self.MAX_WIDTH:
return
n = sum(self.N[len(self.w)][a - self.location[0]:b - self.location[0] +
1])
try:
obsFreq = obsOcc / float(n)
except:
cli.warning('n error')
expFreq = self.bg.freq(self.w, (a, b))
expOcc = expFreq * n
#pv = ppois(obsOcc, expOcc)
#pv = ppois_cached(obsOcc, expOcc)
#pv = pbinom_right_left_cached(obsOcc, n, expFreq)
if self.params['under']:
pv = pbinom_left(obsOcc, n, expFreq)
else:
pv = pbinom(obsOcc, n, expFreq)
ev = 1.0
label = '%s|%s' % (self.w, reverse_complement(self.w))
w = self.w
spaces = self.w.count('N')
if spaces >= 1:
label = label.replace('N' * spaces, 'n{%d}' % spaces)
w = self.w.replace('N' * spaces, 'n{%d}' % spaces)
self.R.append([
w, label, obsFreq, expFreq, obsOcc, expOcc, pv, ev, -log10(ev), a,
b, b - a + 1, 0, n, 0, 0
])
def calcBinomScore(background, foreground, genes, backgroundProb):
TP = len(genes.intersection(foreground))
binomProb = dist.pbinom(TP, len(genes), backgroundProb)
binomScore = -math.log10(binomProb)
return binomScore
def calcBinomScore(background, foreground, genes, backgroundProb):
TP = len(genes.intersection(foreground))
binomProb = dist.pbinom(TP, len(genes), backgroundProb)
binomScore = -math.log10(binomProb)
return binomScore
