-
Notifications
You must be signed in to change notification settings - Fork 7
/
Spectacle_python.py
269 lines (223 loc) · 8.38 KB
/
Spectacle_python.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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
"""
Spectacle_python : An implementation of a spectral algorithm for Hidden Markov Models. The input file consists of sample (cross) moments computed from chromatin mark datasets (e.g. from ENCODE). The output file consists of Hidden Markov Model parameters (initial state distribution, emission matrix and transition matrix). This is a module inside the main Spectacle program which is written in Java and based on the ChromHMM software. Command line instructions:
python Spectacle_python.py file_directory fileID num_states num_marks (min_occurrence_observation)
* Example
java -jar Spectacle.jar LearnModel -nobed -nobrowser -noenrich -f "hg19_inputfilelist.txt" -i spectral -lambda 1 -p 4 -computesamplemomentonly SAMPLEDATA_HG19 OUTPUTSAMPLE_HG19 20 hg19
python Spectacle_python.py OUTPUTSAMPLE_HG19 spectral 20 8
java -jar Spectacle.jar MakeSegmentation -f "hg19_inputfilelist.txt" -i Spectacle -comb "OUTPUTSAMPLE_HG19/model_comb_20_spectral.txt" SAMPLEDATA_HG19 OUTPUTSAMPLE_HG19
"""
import re
import sys
import scipy as sp
import scipy.sparse as ss
import scipy.sparse.linalg as ssl
import numpy as np
import numpy.linalg as nl
chromosomes = ['chr1','chr2','chr3','chr4','chr5','chr6','chr7','chr8','chr9','chr10','chr11','chr12','chr13','chr14','chr15','chr16','chr17','chr18','chr19','chr20','chr21','chr22','chrX']
def getSampleMomentsFromFile(nobs,sample_moment_file,cutoff=1):
"""
Read in sample moments from the input file.
We remove observations whose occurrence in the genome is less than a cutoff.
If cutoff==-1, we include all possible observations.
"""
f = open(sample_moment_file)
line = f.readline()
line = f.readline()
nsegment = int(line.rstrip())
line = f.readline()
marks = []
for line in f:
if (line=='single_vector\n'): break
marks.append(line.rstrip())
M = 2**nobs
Single = ss.dok_matrix((M,1))
for line in f:
if (line=='initial_single_vector\n'): break
tokens = re.split("[\t\n]",line)
i = int(tokens[0])
Single[i,0] = float(tokens[1])
initSingle0 = ss.dok_matrix((M,1))
observation = []
for line in f:
if (line=='pair21_matrix\n'): break
tokens = re.split("[\t\n]",line)
i = int(tokens[0])
if (round(Single[i,0]*nsegment,0)<cutoff): continue
initSingle0[i,0] = float(tokens[1])
observation.append(i)
if cutoff==-1:
observation = range(0,M)
M2 = len(observation)
# print 'Number of non-zero combinations: '+str(M2)
print 'Number of >='+str(cutoff)+' combinations: '+str(M2)
transObs = {}
initSingle = ss.dok_matrix((M2,1))
for i in range(0,M2):
transObs[observation[i]] = i
initSingle[i,0] = initSingle0[observation[i],0]
Pair21 = ss.dok_matrix((M2,M2))
for line in f:
if (line=='pair31_matrix\n'): break
tokens = re.split("[\t\n]",line)
if ((round(Single[int(tokens[0]),0]*nsegment,0)<cutoff)|(round(Single[int(tokens[1]),0]*nsegment,0)<cutoff)): continue
ID2 = transObs[int(tokens[0])]
ID1 = transObs[int(tokens[1])]
Pair21[ID2,ID1] = float(tokens[2])
Pair31 = ss.dok_matrix((M2,M2))
for line in f:
if (line=='triple31_matrix\n'): break
tokens = re.split("[\t\n]",line)
if ((round(Single[int(tokens[0]),0]*nsegment,0)<cutoff)|(round(Single[int(tokens[1]),0]*nsegment,0)<cutoff)): continue
ID3 = transObs[int(tokens[0])]
ID1 = transObs[int(tokens[1])]
Pair31[ID3,ID1] = float(tokens[2])
Triple = {}
t = 0
for line in f:
tokens = re.split("[\t\n]",line)
if ((round(Single[int(tokens[0]),0]*nsegment,0)<cutoff)|(round(Single[int(tokens[1]),0]*nsegment,0)<cutoff)|(round(Single[int(tokens[2]),0]*nsegment,0)<cutoff)): continue
ID2 = transObs[int(tokens[0])]
if (ID2 not in Triple): Triple[ID2] = ss.dok_matrix((M2,M2))
ID3 = transObs[int(tokens[1])]
ID1 = transObs[int(tokens[2])]
Triple[ID2][ID3,ID1] = float(tokens[3])
t += 1
f.close()
# print 'Number of pairs21: '+str(Pair21.getnnz())
# print 'Number of pairs31: '+str(Pair31.getnnz())
# print 'Number of triples: '+str(t)
print 'read sample moment data'
return (transObs,M2,marks,initSingle,Single,Pair21,Pair31,Triple)
def SpectralLearning(M2,K,initSingle,Single,Pair21,Pair31,Triple,transObs,marks,emission_file):
""" The main spectral learning algorithm """
transObs2 = {}
for i in transObs.keys():
transObs2[transObs[i]] = i
# 'Computes SVD of observation pairs matrix'
(U2,S2,V_T2) = ssl.svds(Pair31,K)
U_T2 = np.matrix(U2.transpose())
V_T2 = np.matrix(V_T2)
U_T = np.matrix(np.zeros((K,M2)))
S = np.zeros((K))
V = np.matrix(np.zeros((M2,K)))
for i in range(0,K):
U_T[i,:] = U_T2[K-i-1,:]*np.sign(U_T2[K-i-1,0])
S[i] = S2[K-i-1]
V[:,i] = (V_T2[K-i-1,:]*np.sign(V_T2[K-i-1,0])).transpose()
# Pair_inv3 = (U^T * Pair31)^+
Pair_inv3 = ss.dok_matrix((M2,K))
for j in range(0,K):
for i in range(0,M2):
Pair_inv3[i,j] = V[i,j]/S[j]
# 'Finds the major observation by using U'
mIdx = {}
exist = {}
for i in range(0,K):
mV = 0
mI = 0
for j in range(0,M2):
if (j in exist): continue
v = U_T[i,j]**2
if (v>mV):
mV = v
mI = j
mIdx[i] = mI
exist[mI] = True
# print i,transObs2[mIdx[i]]
# 'Computes eigenvectors'
eigVec = np.matrix(np.zeros((K,K)))
for i in range(0,K):
Cx = U_T * Triple[mIdx[i]] * Pair_inv3
# [w,v] = eigs(Cx,1,maxiter=miter)
[Ws,Vs] = nl.eig(Cx)
v = Vs[:,0]
eigVec[:,i] = np.matrix(np.real(v)*np.sign(np.real(v[0,0])))
eigVec_inv = nl.inv(eigVec)
# 'Computes emission matrix'
O = ss.dok_matrix((M2,K))
O1 = eigVec_inv * U_T
O2 = Pair_inv3 * eigVec
for i in range(0,M2):
if (i not in Triple): continue
Ox = O1 * Triple[i] * O2
for j in range(0,K): O[i,j] = abs(Ox[j,j])
for j in range(0,K): O[:,j] /= O[:,j].sum()
# 'Write emission matrix to the output file'
f = open(emission_file,'w+')
f.write(str(K)+'\t'+str(transObs.__len__())+'\t(Emission order)')
for mark in marks: f.write('\t'+mark)
f.write('\n')
for j in range(0,K):
for i in sorted(transObs.keys()):
f.write(str(j+1)+'\t'+str(i)+'\t'+str(O[transObs[i],j])+'\n')
f.close()
# U_T_O = U_T * O
# 'Computes Inverse of Emission Matrix'
# O_inv = matrix(pinv(O.todense()))
# Pi = O_inv * initSingle
A = ssl.inv((O.transpose()*O).tocsc())
# 'Computes initial state distribution vector'
Pi = A * (O.transpose()*initSingle)
s = abs(Pi).sum()
for i in range(0,K):
Pi[i,0] = abs(Pi[i,0])/s
# 'Computes transition matrix'
# T = abs(O_inv*Pair21*O_inv.transpose()*inv(diagMatrix(Pi).todense()))
# T1 = O_inv*Pair31
# T2 = pinv(O_inv*Pair21)
# T = abs(T1 * T2)
T3 = O.transpose()*Pair31
T2 = O.transpose()*Pair21
T2_inv = ssl.inv((T2*T2.transpose()).tocsc())
A_inv = ssl.inv(A)
T = abs(A*(T3*T2.transpose())*T2_inv*A_inv)
for j in range(0,K): T[:,j] /= T[:,j].sum()
return (T,O,Pi)
def diagMatrix(V):
n = max(np.shape(V)[0],np.shape(V)[1])
n2 = min(np.shape(V)[0],np.shape(V)[1])
if (n2!=1):
print 'Error: it is not a vector'
return V
if (np.shape(V)[1]==n): V = V.transpose()
M = ss.dok_matrix((n,n))
for i in range(0,n): M[i,i] = V[i,0]
return M
def diagSum(M):
n = min(np.shape(M)[0],np.shape(M)[1])
c = 0
for i in range(0,n): c += M[i,i]
return c
def writeModel(T,O,Pi,nobs,marks,transObs,model_file):
(M2,K) = np.shape(O)
f = open(model_file,'w+')
f.write(str(K)+'\t'+str(nobs)+'\tU\t0\t0\n')
for i in range(0,K):
f.write('probinit\t'+str(i+1)+'\t'+str(Pi[i,0])+'\n')
for j in range(0,K):
for i in range(0,K):
f.write('transitionprobs\t'+str(j+1)+'\t'+str(i+1)+'\t'+str(T[i,j])+'\n')
for j in range(0,K):
for i in sorted(transObs.keys()):
f.write('emissionprobs\t'+str(j+1)+'\t'+str(i)+'\t'+str(O[transObs[i],j])+'\n')
f.close()
if __name__ == "__main__":
nargv = sys.argv.__len__()
if ((nargv!=5)&(nargv!=6)):
print 'Error: wrong number of arguments'
print 'python Spectacle_python.py file_directory fileID num_states num_marks (min_occurrence_observation)'
exit(1)
filedir = sys.argv[1]
outputfileID = sys.argv[2]
K = int(sys.argv[3])
nobs = int(sys.argv[4])
if (nargv==6): cutoff = int(sys.argv[5])
else: cutoff = 1
sample_moment_file = filedir+'/sample_moments_'+outputfileID+'_'+str(nobs)+'.txt'
(transObs,M2,marks,initSingle,Single,Pair21,Pair31,Triple) = getSampleMomentsFromFile(nobs,sample_moment_file,cutoff)
if (cutoff>1): outputfileID2 = outputfileID+'_moo'+str(cutoff)
else: outputfileID2 = outputfileID
emission_file = filedir+'/emissions_comb_'+str(K)+'_'+outputfileID2+'.txt'
(T,O,Pi) = SpectralLearning(M2,K,initSingle,Single,Pair21,Pair31,Triple,transObs,marks,emission_file)
model_file = filedir+'/model_comb_'+str(K)+'_'+outputfileID2+'.txt'
writeModel(T,O,Pi,nobs,marks,transObs,model_file)