forked from brantp/rtd
-
Notifications
You must be signed in to change notification settings - Fork 0
/
RE_site_dropout.py
executable file
·143 lines (119 loc) · 6.91 KB
/
RE_site_dropout.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
#!/usr/bin/env python
from scipy.stats.distributions import norm,poisson
import sys,numpy,time
def count_likelihood_standard(this_counts,tot_counts,this_port,this_cv,this_fract_recov=1):
L = [norm.pdf(this_c,loc=tot_c*this_port*this_fract_recov,scale=(tot_c*this_port*this_fract_recov)*this_cv) \
for this_c,tot_c in zip(this_counts,tot_counts)]
logL = numpy.log2(numpy.array(L))
return sum(logL)
# <codecell>
def count_likelihood_poisson(this_counts,tot_counts,this_port,this_cv,this_fract_recov=1):
L = [poisson.pmf(this_c,tot_c*this_port*this_fract_recov) \
for this_c,tot_c in zip(this_counts,tot_counts)]
logL = numpy.log2(numpy.array(L))
return sum(logL)
# <codecell>
def fake_null_cuts(this_counts,freq_cut_null):
num_absent = int(round(freq_cut_null**2 * len(this_counts)))
num_het = int(round((2 * freq_cut_null * (1 - freq_cut_null)) * len(this_counts)))
ret = []
for i in range(num_absent):
ret.append(0)
for i in range(num_absent,num_absent+num_het):
ret.append(int(this_counts[i]*numpy.random.normal(0.5,0.01)))
for i in range(num_absent+num_het,len(this_counts)):
ret.append(this_counts[i])
return numpy.array(ret)
# <codecell>
def count_likelihood_skew(this_counts,tot_counts,this_port,this_cv,freq_cut_null,err=1e-5,Lfn=count_likelihood_standard):
num_absent = int(round(freq_cut_null**2 * len(this_counts)))
num_het = int(round((2 * freq_cut_null * (1 - freq_cut_null)) * len(this_counts)))
ind_absent = [i for i,ncounts in sorted(enumerate(this_counts),key = lambda x:x[1])[:num_absent]]
ind_het = []
for i, portion in sorted(enumerate(this_counts/tot_counts.astype('float')),key = lambda x:x[1]):
if len(ind_het) == num_het:
break
if i in ind_absent:
continue
ind_het.append(i)
ind_full = list((set(range(len(this_counts)))-set(ind_het))-set(ind_absent))
#print [(i,this_counts[i],(this_counts/tot_counts.astype('float'))[i]) for i in ind_absent]
#print [(i,this_counts[i],(this_counts/tot_counts.astype('float'))[i]) for i in ind_het]
#print [(i,this_counts[i],(this_counts/tot_counts.astype('float'))[i]) for i in ind_full]
#print 'absent %s\nhet %s\nfull %s' % (ind_absent,ind_het,ind_full)
L0li = [err**this_counts[i] for i in ind_absent]
L1 = Lfn(this_counts[ind_het],tot_counts[ind_het],this_port,this_cv,0.5)
L2 = Lfn(this_counts[ind_full],tot_counts[ind_full],this_port,this_cv,1)
return sum(numpy.log2(L0li)) + L1 + L2
# <codecell>
def max_likelihood_standard(this_counts,tot_counts, \
num_portions_to_test=10,mincv=0.1,maxcv=2.0,num_cv_to_test=10, \
npass=2,take_top=0.1,Lfn=count_likelihood_standard):
this_portions = this_counts/tot_counts.astype('float')
test_ports = numpy.linspace(this_portions.min(),this_portions.max(),num_portions_to_test)
if Lfn == count_likelihood_poisson:
test_cvs = [1]
else:
test_cvs = numpy.linspace(mincv,maxcv,num_cv_to_test)
#print this_portions,test_ports,test_cvs
for i in range(npass):
l_out = sorted([ (Lfn(this_counts,tot_counts,test_port,test_cv),\
(test_port,test_cv)) \
for test_port in test_ports for test_cv in test_cvs])
top = int(len(test_ports)*len(test_cvs)*take_top)
#print >> sys.stderr, l_out[-1]
top_ports,top_cv = zip(*[el[1] for el in l_out[-top:]])
test_ports = numpy.linspace(min(top_ports),max(top_ports),num_portions_to_test)
test_cvs = numpy.linspace(min(top_cv),max(top_cv),num_cv_to_test)
return l_out[-1]
# <codecell>
def max_likelihood_skew(this_counts,tot_counts, \
num_portions_to_test=10,mincv=0.1,maxcv=2.0,num_cv_to_test=10, \
num_freq_to_test=10,npass=2,take_top=0.1,Lfn=count_likelihood_standard):
this_portions = this_counts/tot_counts.astype('float')
test_ports = numpy.linspace(this_portions.min(),this_portions.max(),num_portions_to_test)
if Lfn == count_likelihood_poisson:
test_cvs = [1]
else:
test_cvs = numpy.linspace(mincv,maxcv,num_cv_to_test)
test_NFs = numpy.linspace(0,1,num_freq_to_test)
for i in range(npass):
l_out = sorted([ (count_likelihood_skew(this_counts,tot_counts,test_port,test_cv,test_NF,Lfn=Lfn),\
(test_port,test_cv,test_NF)) \
for test_port in test_ports for test_cv in test_cvs for test_NF in test_NFs])
top = int(len(test_ports)*len(test_cvs)*take_top)
#print >> sys.stderr, l_out[-1]
top_ports,top_cv,top_NF = zip(*[el[1] for el in l_out[-top:]])
test_ports = numpy.linspace(min(top_ports),max(top_ports),num_portions_to_test)
test_cvs = numpy.linspace(min(top_cv),max(top_cv),num_cv_to_test)
test_NFs = numpy.linspace(min(top_NF),max(top_NF),num_cv_to_test)
return l_out[-1]
def run_main(tot_mean,tot_sd,n_ind,this_mean,this_cv,this_NF,n_steps,n_pass,Lfn=count_likelihood_standard):
#tot_mean = 300000
#tot_sd = 150000
#n_ind = 48
tot_counts = numpy.array([x>10000 and x or 10000 \
for x in numpy.random.normal(tot_mean,tot_sd,size=n_ind).astype('int')])
#this_mean = 10
port = this_mean/float(tot_mean)
#this_cv = 0.5
this_counts = numpy.array([x>0 and x or 0 for x in tot_counts*numpy.random.normal(port,port*this_cv,n_ind)]).astype('int')
print >> sys.stderr, tot_counts,numpy.mean(tot_counts),numpy.std(tot_counts)
print >> sys.stderr, this_counts,numpy.mean(this_counts),numpy.std(this_counts)
mean_port = numpy.mean(this_counts/tot_counts.astype('float'))
std_port = numpy.std(this_counts/tot_counts.astype('float'))
print >> sys.stderr, mean_port, std_port/mean_port
L0,(L0_portion,L0_cv) = max_likelihood_standard(this_counts,tot_counts,num_portions_to_test=n_steps,num_cv_to_test=n_steps,npass=n_pass,Lfn=Lfn)
L1,(L1_portion,L1_cv,F) = max_likelihood_skew(this_counts,tot_counts,num_portions_to_test=n_steps,num_cv_to_test=n_steps,num_freq_to_test=n_steps,npass=n_pass,Lfn=Lfn)
print >> sys.stderr, 2*(-L0 - -L1), L0_portion,L0_cv
LRTneg = 2*(-L0 - -L1)
skew_counts = fake_null_cuts(this_counts,this_NF)
L0,(L0_portion,L0_cv) = max_likelihood_standard(skew_counts,tot_counts,num_portions_to_test=n_steps,num_cv_to_test=n_steps,npass=n_pass,Lfn=Lfn)
L1,(L1_portion,L1_cv,F) = max_likelihood_skew(skew_counts,tot_counts,num_portions_to_test=n_steps,num_cv_to_test=n_steps,num_freq_to_test=n_steps,npass=n_pass,Lfn=Lfn)
print >> sys.stderr, 2*(-L0 - -L1), L1_portion,L1_cv,F
LRTpos = 2*(-L0 - -L1)
return {'LRT':{'neg':LRTneg,'pos':LRTpos},'F':F}
if __name__ == "__main__":
tot_mean,tot_sd,n_ind,this_mean,this_cv,this_NF,n_steps,n_pass = map(float,sys.argv[1:])
n_ind, n_steps, n_pass = map(int,[n_ind, n_steps, n_pass])
print run_main(tot_mean,tot_sd,n_ind,this_mean,this_cv,this_NF,n_steps,n_pass)