-
Notifications
You must be signed in to change notification settings - Fork 1
/
hmm.cc
261 lines (227 loc) · 13.9 KB
/
hmm.cc
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
#include "hmm.h"
#include <iomanip>
hmm::hmm( obsSequence *seq)
{
mObsSequence = seq;
mModelProb = NULL;
mHmmResults = NULL;
}
double hmm::prob_obs_internal( obsSite dat, probSegregDInternal prob )
{
if ( dat.getDerived() < dat.getCoverage() )
if ( dat.getUlindi() == 'D' )
return prob.getProb( dat.getDerived() ) ;
else return 1.-prob.getProb( dat.getDerived() ) ;
else
if ( dat.getUlindi() == 'D' )
return mModelProb->getProbFixDInternal() ;
else return 1.-mModelProb->getProbFixDInternal() ;
}
double hmm::prob_obs_external( obsSite dat )
{
if ( dat.getDerived() < dat.getCoverage() )
if ( dat.getUlindi() == 'D' )
return mModelProb->getProbSegregDExternal() ;
else return 1.-mModelProb->getProbSegregDExternal() ;
else
if ( dat.getUlindi() == 'D' )
return mModelProb->getProbFixDExternal() ;
else return 1.-mModelProb->getProbFixDExternal() ;
}
void hmm::computeFwdBwd(modelProb *model, hmmResults *result)
{
mModelProb = model;
mHmmResults = result;
//////////
// forward probabilities
mHmmResults->mfwd_stateE_scaled.resize( mObsSequence->size(), 0.0 ) ; // external
mHmmResults->mfwd_stateI_scaled.resize( mObsSequence->size(), 0.0 ) ; // internal
// first one by hand
vector<double> fwd_scaling( mObsSequence->size(), 0.0 ) ;
// Initial state = internal
fwd_scaling[0] = prob_obs_internal( mObsSequence->at(0), mModelProb->getProbSegregDInternal() ) ;
mHmmResults->mfwd_stateI_scaled[0] = 1. ; // == fwd_stateI[0]/fwd_scaling[0] ;
// rest
for ( int i = 1 ; i < mObsSequence->size() ; ++i ) {
double I_tmp = prob_obs_internal( mObsSequence->at(i), mModelProb->getProbSegregDInternal() )
*mHmmResults->mfwd_stateI_scaled[i-1]*1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayInternal()) +
prob_obs_internal( mObsSequence->at(i), mModelProb->getProbSegregDInternal() )
*mHmmResults->mfwd_stateE_scaled[i-1]*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayExternal())) ;
double E_tmp = ( prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateI_scaled[i-1]*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayInternal())) +
prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateE_scaled[i-1]*1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayExternal()) ) ;
fwd_scaling[i] = I_tmp + E_tmp ;
mHmmResults->mfwd_stateI_scaled[i] = I_tmp/fwd_scaling[i] ;
mHmmResults->mfwd_stateE_scaled[i] = E_tmp/fwd_scaling[i] ;
}
//////////////
// Log-likelihood
mHmmResults->mlogLikelihood = 0.0;
for ( int i = 1 ; i < mObsSequence->size() ; ++i ) {
mHmmResults->mlogLikelihood += log(fwd_scaling[i]);
}
//////////////////
// backward probabilities
mHmmResults->mbwd_stateI_scaled.resize( mObsSequence->size(), 0.0 ) ;
mHmmResults->mbwd_stateE_scaled.resize( mObsSequence->size(), 0.0 ) ;
mHmmResults->mbwd_stateI_scaled[mObsSequence->size()-1] = 1. ;
mHmmResults->mbwd_stateE_scaled[mObsSequence->size()-1] = 1. ;
for ( int i = mObsSequence->size()-2 ; i >= 0 ; i-- ) {
double scale = fwd_scaling[i+1] ;
mHmmResults->mbwd_stateI_scaled[i] = (exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayInternal())
* prob_obs_internal( mObsSequence->at(i+1), mModelProb->getProbSegregDInternal() )*mHmmResults->mbwd_stateI_scaled[i+1] +
(1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayInternal()))
* prob_obs_external( mObsSequence->at(i+1) )*mHmmResults->mbwd_stateE_scaled[i+1])/scale ;
mHmmResults->mbwd_stateE_scaled[i] = ((1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayExternal()))
* prob_obs_internal( mObsSequence->at(i+1), mModelProb->getProbSegregDInternal() )*mHmmResults->mbwd_stateI_scaled[i+1] +
exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayExternal()) * prob_obs_external( mObsSequence->at(i+1) )
*mHmmResults->mbwd_stateE_scaled[i+1])/scale ;
}
//////////////////////
// posterior decoding
mHmmResults->mExternal.resize( mObsSequence->size(), 0 ) ; // == 1 if external
for ( size_t i = 0 ; i < mObsSequence->size() ; ++i ) {
double val = (mHmmResults->mfwd_stateI_scaled[i]*mHmmResults->mbwd_stateI_scaled[i]) ;
if ( val < 0.2 ) mHmmResults->mExternal[i] = 1 ;
else mHmmResults->mExternal[i] = 0 ;
}
}
void hmm::computeFwdBwd3states(modelProb *model, hmmResults *result)
{
mModelProb = model;
mHmmResults = result;
//////////
// forward probabilities
mHmmResults->mfwd_stateE_scaled.resize( mObsSequence->size(), 0.0 ) ; // external
mHmmResults->mfwd_stateLE_scaled.resize( mObsSequence->size(), 0.0 ) ; // long external
mHmmResults->mfwd_stateI_scaled.resize( mObsSequence->size(), 0.0 ) ; // internal
// first one by hand
vector<double> fwd_scaling( mObsSequence->size(), 0.0 ) ;
// Initial state = internal
fwd_scaling[0] = prob_obs_internal( mObsSequence->at(0), mModelProb->getProbSegregDInternal() ) ;
mHmmResults->mfwd_stateI_scaled[0] = 1. ; // == fwd_stateI[0]/fwd_scaling[0] ;
// rest
for ( int i = 1 ; i < mObsSequence->size() ; ++i ) {
double I_tmp = prob_obs_internal( mObsSequence->at(i), mModelProb->getProbSegregDInternal() )
*mHmmResults->mfwd_stateI_scaled[i-1]*1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayInternal()) +
prob_obs_internal( mObsSequence->at(i), mModelProb->getProbSegregDInternal() )
*mHmmResults->mfwd_stateE_scaled[i-1]*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayExternal())) +
prob_obs_internal( mObsSequence->at(i), mModelProb->getProbSegregDInternal() )
*mHmmResults->mfwd_stateLE_scaled[i-1]*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayLongExternal())) ;
double E_tmp = ( prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateI_scaled[i-1]*(1-mModelProb->getLErate())*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayInternal())) +
prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateE_scaled[i-1]*1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayExternal()) ) ;
double LE_tmp = ( prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateI_scaled[i-1]*mModelProb->getLErate()*(1.-1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayInternal())) +
prob_obs_external( mObsSequence->at(i) )
* mHmmResults->mfwd_stateLE_scaled[i-1]*1./exp(mObsSequence->at(i).getDist()/mModelProb->getStayLongExternal()) ) ;
fwd_scaling[i] = I_tmp + E_tmp + LE_tmp ;
mHmmResults->mfwd_stateI_scaled[i] = I_tmp/fwd_scaling[i] ;
mHmmResults->mfwd_stateE_scaled[i] = E_tmp/fwd_scaling[i] ;
mHmmResults->mfwd_stateLE_scaled[i] = LE_tmp/fwd_scaling[i] ;
}
//////////////
// Log-likelihood
mHmmResults->mlogLikelihood = 0.0;
for ( int i = 1 ; i < mObsSequence->size() ; ++i ) {
mHmmResults->mlogLikelihood += log(fwd_scaling[i]);
}
//////////////////
// backward probabilities
mHmmResults->mbwd_stateI_scaled.resize( mObsSequence->size(), 0.0 ) ;
mHmmResults->mbwd_stateE_scaled.resize( mObsSequence->size(), 0.0 ) ;
mHmmResults->mbwd_stateLE_scaled.resize( mObsSequence->size(), 0.0 ) ;
mHmmResults->mbwd_stateI_scaled[mObsSequence->size()-1] = 1. ;
mHmmResults->mbwd_stateE_scaled[mObsSequence->size()-1] = 1. ;
mHmmResults->mbwd_stateLE_scaled[mObsSequence->size()-1] = 1. ;
for ( int i = mObsSequence->size()-2 ; i >= 0 ; i-- ) {
double scale = fwd_scaling[i+1] ;
mHmmResults->mbwd_stateI_scaled[i] = (exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayInternal())
* prob_obs_internal( mObsSequence->at(i+1), mModelProb->getProbSegregDInternal() )*mHmmResults->mbwd_stateI_scaled[i+1] +
(1-mModelProb->getLErate())*(1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayInternal()))
* prob_obs_external( mObsSequence->at(i+1) )*mHmmResults->mbwd_stateE_scaled[i+1] +
mModelProb->getLErate()*(1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayInternal()))
* prob_obs_external( mObsSequence->at(i+1) )*mHmmResults->mbwd_stateLE_scaled[i+1])/scale ;
mHmmResults->mbwd_stateE_scaled[i] = ((1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayExternal()))
* prob_obs_internal( mObsSequence->at(i+1), mModelProb->getProbSegregDInternal() )*mHmmResults->mbwd_stateI_scaled[i+1] +
exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayExternal()) * prob_obs_external( mObsSequence->at(i+1) )
*mHmmResults->mbwd_stateE_scaled[i+1])/scale ;
mHmmResults->mbwd_stateLE_scaled[i] = ((1.-exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayLongExternal()))
* prob_obs_internal( mObsSequence->at(i+1), mModelProb->getProbSegregDInternal() )*mHmmResults->mbwd_stateI_scaled[i+1] +
exp(-1.*mObsSequence->at(i+1).getDist()/mModelProb->getStayLongExternal()) * prob_obs_external( mObsSequence->at(i+1) )
*mHmmResults->mbwd_stateLE_scaled[i+1])/scale ;
}
//////////////////////
// posterior decoding
mHmmResults->mExternal.resize( mObsSequence->size(), 0 ) ; // == 1 if external
for ( size_t i = 0 ; i < mObsSequence->size() ; ++i ) {
double val = (mHmmResults->mfwd_stateE_scaled[i]*mHmmResults->mbwd_stateE_scaled[i]) ;
double val2 = (mHmmResults->mfwd_stateLE_scaled[i]*mHmmResults->mbwd_stateLE_scaled[i]) ;
if ( val > 0.8 ) mHmmResults->mExternal[i] = 1 ;
else if ( val2 > 0.8 ) mHmmResults->mExternal[i] = 2 ;
else mHmmResults->mExternal[i] = 0 ;
}
}
void hmm::writeLogFile(const char* fileName, string logInfo)
{
ofstream myfile;
myfile.open(fileName, std::ios_base::app);
myfile << std::setprecision (15) << logInfo << "\t"
<< mHmmResults->mlogLikelihood << "\t"
<< mModelProb->getStayInternal() << "\t"
<< mModelProb->getStayExternal() << "\t"
<< mModelProb->getProbFixDInternal() << "\t"
<< mModelProb->getProbFixDExternal() << "\t"
<< mModelProb->getProbSegregDExternal() << "\t"
<< mModelProb->getProbSegregDInternal().printAll()
<< "\n";
myfile.close();
}
void hmm::writeLogFile3states(const char* fileName, string logInfo)
{
ofstream myfile;
myfile.open(fileName, std::ios_base::app);
myfile << std::setprecision (15) << logInfo << "\t"
<< mHmmResults->mlogLikelihood << "\t"
<< mModelProb->getLErate() << "\t"
<< mModelProb->getStayInternal() << "\t"
<< mModelProb->getStayExternal() << "\t"
<< mModelProb->getStayLongExternal() << "\t"
<< mModelProb->getProbFixDInternal() << "\t"
<< mModelProb->getProbFixDExternal() << "\t"
<< mModelProb->getProbSegregDExternal() << "\t"
<< mModelProb->getProbSegregDInternal().printAll()
<< "\n";
myfile.close();
}
void hmm::writeOutputFile(void)
{
for(size_t i=1; i<mObsSequence->size(); ++i) {
cout << mObsSequence->at(i).getChr() << "\t"
<< mObsSequence->at(i).getLocation() << "\t"
<< mObsSequence->at(i).getClint() << "\t"
<< mObsSequence->at(i).getCoverage() << "\t"
<< mObsSequence->at(i).getDerived() << "\t"
<< mObsSequence->at(i).getUlindi() << "\t"
<< mObsSequence->at(i).getDist() << "\t"
<< mHmmResults->mExternal[i] << "\t"
<< mHmmResults->mfwd_stateE_scaled[i]*mHmmResults->mbwd_stateE_scaled[i] << "\n" ;
}
}
void hmm::writeOutputFile3states(void)
{
for(size_t i=1; i<mObsSequence->size(); ++i) {
cout << mObsSequence->at(i).getChr() << "\t"
<< mObsSequence->at(i).getLocation() << "\t"
<< mObsSequence->at(i).getClint() << "\t"
<< mObsSequence->at(i).getCoverage() << "\t"
<< mObsSequence->at(i).getDerived() << "\t"
<< mObsSequence->at(i).getUlindi() << "\t"
<< mObsSequence->at(i).getDist() << "\t"
<< mHmmResults->mExternal[i] << "\t"
<< mHmmResults->mfwd_stateE_scaled[i]*mHmmResults->mbwd_stateE_scaled[i] << "\t"
<< mHmmResults->mfwd_stateLE_scaled[i]*mHmmResults->mbwd_stateLE_scaled[i] << "\n" ;
}
}