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
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
|
#include "likelihoodComputationGL.h"
#include "definitions.h"
#include "tree.h"
#include "likelihoodComputation.h"
#include <cmath>
#include <cassert>
using namespace likelihoodComputationGL;
// account for RateCat, GainCat,LossCat
// - For each RateCat an "external" multiplication is conducted - copy_et.multipleAllBranchesByFactor
// - the GainCat*LossCat SPs are covered by the "internal" mechanism of PijGam
/********************************************************************************************
*********************************************************************************************/
MDOUBLE likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(const tree& tr,
const sequenceContainer& sc,
const vector<vector<stochasticProcess*> >& spVVec,
const distribution * distGain, const distribution * distLoss,
const Vdouble * const weights,
unObservableData *unObservableData_p)
{
int numOfRateCategories = spVVec[0][0]->categories();
vector<computePijGam> pi_vec(numOfRateCategories);
vector<suffStatGlobalGam> ssc_vec(numOfRateCategories);
vector<computeUpAlg> cup_vec(numOfRateCategories);
likelihoodComputationGL::fillPijAndUp(tr,sc,spVVec,distGain,distLoss,pi_vec,ssc_vec,cup_vec);
MDOUBLE logLforMissingData;
MDOUBLE LforMissingData = 0;
if(unObservableData_p){
logLforMissingData = unObservableData_p->getlogLforMissingData();
LforMissingData = exp(logLforMissingData);
}
MDOUBLE res = 0.0;
for (int k=0; k < sc.seqLen(); ++k) {
MDOUBLE lnL = 0;
MDOUBLE resGivenRate = 0.0;
for(int rateIndex=0 ; rateIndex<numOfRateCategories; ++rateIndex){
lnL = likelihoodComputationGL::getProbOfPosUpIsFilledSelectionGam(k,//pos,
tr,//const tree&
sc,// sequenceContainer& sc,
spVVec, // only needed for sp.freq(let)
ssc_vec[rateIndex][k],//const computePijGam& ,
distGain, distLoss); // distributions ,
resGivenRate += lnL * spVVec[0][0]->ratesProb(rateIndex);
}
if(unObservableData_p){ // conditioning on observability for all rateCat.
resGivenRate = resGivenRate / (1- LforMissingData);
}
LOG(20,<<"pos= "<<k+1<<" resGivenRate= "<<resGivenRate<<endl);
//res += lnL;
res += log(resGivenRate) * (weights?(*weights)[k]:1);
}
//if(unObservableData_p){
// res = res - sc.seqLen()*log(1- exp(unObservableData_p->getlogLforMissingData()));
//}
return res;
}
/********************************************************************************************
*********************************************************************************************/
void likelihoodComputationGL::fillPijAndUp(const tree& tr,
const sequenceContainer& sc,
const vector<vector<stochasticProcess*> >& spVVec,
const distribution * distGain, const distribution * distLoss,
vector<computePijGam>& pi_vec,
vector<suffStatGlobalGam>& ssc_vec, // info filled into suffStat
vector<computeUpAlg>& cup_vec)
{
int numOfSPs = distGain->categories()*distLoss->categories();
int numOfRateCategories = spVVec[0][0]->categories();
for (int rateIndex=0 ; rateIndex<numOfRateCategories; ++rateIndex){
tree copy_et = tr;
copy_et.multipleAllBranchesByFactor(spVVec[0][0]->rates(rateIndex));
pi_vec[rateIndex]._V.resize(numOfSPs);
//Pij
for (int i=0; i < numOfSPs; ++i) {
int gainIndex =fromIndex2gainIndex(i,distGain->categories(),distLoss->categories());
int lossIndex =fromIndex2lossIndex(i,distGain->categories(),distLoss->categories());
pi_vec[rateIndex]._V[i].fillPij(copy_et,*spVVec[gainIndex][lossIndex]);
}
//ComputeUp
cup_vec[rateIndex].fillComputeUp(copy_et,sc,pi_vec[rateIndex],ssc_vec[rateIndex]);
}
}
/********************************************************************************************
*********************************************************************************************/
MDOUBLE likelihoodComputationGL::getProbOfPosUpIsFilledSelectionGam(const int pos,const tree& tr,
const sequenceContainer& sc,
const vector<vector<stochasticProcess*> >& spVVec,// only needed for sp.freq(let)
const suffStatGlobalGamPos& cup,
const distribution * distGain, const distribution * distLoss)
{
doubleRep res =0;
int numOfSPs = distGain->categories()*distLoss->categories();
for (int categor = 0; categor < numOfSPs; ++categor) {
doubleRep veryTmp =0.0;
int gainCategor = fromIndex2gainIndex(categor,distGain->categories(),distLoss->categories());
int lossCategor = fromIndex2lossIndex(categor,distGain->categories(),distLoss->categories());
for (int let =0; let < sc.alphabetSize(); ++let) {
veryTmp+=cup.get(categor,tr.getRoot()->id(),let) * spVVec[gainCategor][lossCategor]->freq(let); // Root character freq
}
res += veryTmp*(distGain->ratesProb(gainCategor)*distLoss->ratesProb(lossCategor));
}
if ((res<-EPSILON)){
string err = "Error in likelihoodComputationGL::getProbOfPosUpIsFilledSelectionGam, non probability value (<0) Res=";
err+=double2string(convert(res));
errorMsg::reportError(err);
};
return convert(res);
}
/********************************************************************************************
*********************************************************************************************/
MDOUBLE likelihoodComputationGL::getTreeLikelihoodFromUp2(const tree& tr,
const sequenceContainer& sc,
const vector<vector<stochasticProcess*> >& spVVec,// only needed for sp.freq(let)
const suffStatGlobalGam& cup, //computing the likelihood from up:
const distribution * distGain, const distribution * distLoss,
const Vdouble * weights,
unObservableData *unObservableData_p,
Vdouble* posLike)
{
if(posLike)
posLike->clear();
MDOUBLE like = 0;
int numOfSPs = distGain->categories()*distLoss->categories();
for (int pos = 0; pos < sc.seqLen(); ++pos) {
doubleRep tmp=0;
for (int categor = 0; categor < numOfSPs; ++categor) {
doubleRep veryTmp =0;
int gainCategor = fromIndex2gainIndex(categor,distGain->categories(),distLoss->categories());
int lossCategor = fromIndex2lossIndex(categor,distGain->categories(),distLoss->categories());
for (int let =0; let < sc.alphabetSize(); ++let) {
veryTmp+=cup.get(pos,categor,tr.getRoot()->id(),let) * spVVec[gainCategor][lossCategor]->freq(let);
}
tmp += veryTmp*(distGain->ratesProb(gainCategor)*distLoss->ratesProb(lossCategor));
}
if(unObservableData_p)
tmp = tmp/(1- exp(unObservableData_p->getlogLforMissingData()));
if(posLike)
posLike->push_back(log(tmp));
like += log(tmp) * (weights?(*weights)[pos]:1);
}
return like;
}
/********************************************************************************************
*********************************************************************************************/
MDOUBLE likelihoodComputationGL::getTreeLikelihoodFromUp2(const tree& tr,
const sequenceContainer& sc,
const vector<vector<stochasticProcess*> >& spVVec,// only needed for sp.freq(let)
const vector<suffStatGlobalGam>& cup_vec, //computing the likelihood from up:
const distribution * distGain, const distribution * distLoss,
const Vdouble * weights,
unObservableData *unObservableData_p,
Vdouble* posLike )
{
if(posLike)
posLike->resize(sc.seqLen());
MDOUBLE like = 0;
int numOfRateCategories = spVVec[0][0]->categories();
for(int rateIndex=0 ; rateIndex<numOfRateCategories; ++rateIndex){
Vdouble posLikePerCat;
like += likelihoodComputationGL::getTreeLikelihoodFromUp2(tr,sc,spVVec,cup_vec[rateIndex], distGain,distLoss,weights,unObservableData_p,&posLikePerCat)
* spVVec[0][0]->ratesProb(rateIndex);
if(posLike){
for (int k=0; k < sc.seqLen(); ++k) {
(*posLike)[k]+= (posLikePerCat[k]* spVVec[0][0]->ratesProb(rateIndex));
}
}
}
return like;
}
/********************************************************************************************
*********************************************************************************************/
//MDOUBLE likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSameNoComputeUp(const tree& tr,
// const sequenceContainer& sc,
// const vector<vector<stochasticProcess*> >& spVVec,
// const distribution * distGain, const distribution * distLoss,
// unObservableData *unObservableData_p)
//{
// MDOUBLE res = 0.0;
// int numOfSPs = distGain->categories()*distLoss->categories();
// for (int i=0; i < numOfSPs; ++i) {
// int gainIndex =fromIndex2gainIndex(i,distGain->categories(),distLoss->categories());
// int lossIndex =fromIndex2lossIndex(i,distGain->categories(),distLoss->categories());
// res += likelihoodComputation::getTreeLikelihoodAllPosAlphTheSame(tr,sc,*spVVec[gainIndex][lossIndex])* distGain->ratesProb(gainIndex)*distLoss->ratesProb(lossIndex);
// }
// if(unObservableData_p){
// res = res - sc.seqLen()*log(1- exp(unObservableData_p->getlogLforMissingData()));
// }
// return res;
//}
/********************************************************************************************
un-obervable data
*********************************************************************************************/
/********************************************************************************************
used to fill the likelihood for the unobervable for each category
*********************************************************************************************/
//doubleRep likelihoodComputationGL::getLofPos(const int pos,
// const tree& tr,
// const sequenceContainer& sc,
// const computePijGam& pi,
// const stochasticProcess& sp,
// Vdouble& likePerCat) // all the likdelhoodsPerCat and rateProb are filled
//{
// // with the pi already computed.
// int numOfCat = sp.categories();
// doubleRep tmp=0;
// for (int i=0; i < numOfCat;++i) {
// likePerCat[i] = getLofPos(pos,tr,sc,pi[i],sp)*sp.ratesProb(i);
// likePerCat[i+numOfCat] = sp.ratesProb(i);
// tmp += likePerCat[i];
// }
// return tmp;
//}
///********************************************************************************************
//likelihood computation - full data (1)
//*********************************************************************************************/
//MDOUBLE likelihoodComputationGL::getTreeLikelihoodAllPosAlphTheSame(const tree& tr,
// const sequenceContainer& sc,
// const stochasticProcess& sp,
// const Vdouble * const weights,
// Vdouble *pLforMissingDataPerCat)
//{
// computePijGam pi;
// pi.fillPij(tr,sp);
// MDOUBLE res =0;
// doubleRep LofPos;
// int k;
// for (k=0; k < sc.seqLen(); ++k) {
// LofPos = likelihoodComputationGL::getLofPos(k,//pos,
// tr,//const tree&
// sc,// sequenceContainer& sc,
// pi,//const computePijGam& ,
// sp,
// pLforMissingDataPerCat);
// res += log(LofPos) * (weights?(*weights)[k]:1);//const stochasticProcess& );
// }
// return res;
//}
//
///********************************************************************************************
//likelihood computation - per pos (1.1)
//*********************************************************************************************/
//doubleRep likelihoodComputationGL::getLofPos(const int pos,
// const tree& tr,
// const sequenceContainer& sc,
// const computePijGam& pi,
// const stochasticProcess& sp,
// Vdouble *pLforMissingDataPerCat)
//{
//// with the pi already computed.
// doubleRep tmp=0;
// int numOfCat = sp.categories();
// Vdouble tmpPerCat;
// tmpPerCat.resize(numOfCat);
//
// for (int i=0; i < sp.categories();++i) {
// tmpPerCat[i] = getLofPos(pos,tr,sc,pi[i],sp);
// if(pLforMissingDataPerCat){
// LOG(11,<<"res before MissingData correction= "<<tmpPerCat[i]);
// tmpPerCat[i] = tmpPerCat[i]/(1- (*pLforMissingDataPerCat)[i]);
// LOG(11,<<" after= "<<tmpPerCat[i]<<endl);
// }
// tmp += tmpPerCat[i]*sp.ratesProb(i);
// }
// return tmp;
//}
//
///********************************************************************************************
//likelihood computation - per pos, per cat (1.1.1)
//*********************************************************************************************/
//doubleRep likelihoodComputationGL::getLofPos(const int pos,
// const tree& tr,
// const sequenceContainer& sc,
// const computePijHom& pi,
// const stochasticProcess& sp)
//{
// computeUpAlg cup;
// suffStatGlobalHomPos ssc;
// cup.fillComputeUp(tr,sc,pos,pi,ssc);
//
// doubleRep tmp = 0.0;
// for (int let = 0; let < sp.alphabetSize(); ++let) {
// doubleRep tmpLcat=
// ssc.get(tr.getRoot()->id(),let)*
// sp.freq(let);
// if (!DBIG_EQUAL(convert(tmpLcat), 0.0))
// {
// cerr<<"tmpLcat = "<<tmpLcat<<endl;
// errorMsg::reportError("error in likelihoodComputation::getLofPos. likelihood is smaller than zero");
// }
//
// //assert(tmpLcat>=0.0);
// tmp+=tmpLcat;
// }
//// cout<<"likelihoodComputation::getLofPos: tmp = "; tmp.outputn(cout); // DEBUG EP
// if (!(tmp>0.0)){
// LOG(5,<<"likelihoodComputation::getLofPos: "<< tmp<<endl;);
// LOG(5,<<"pos = "<< pos <<endl;);
// tmp = EPSILON;
// //errorMsg::reportError("likelihoodComputation::getLofPos: likelihood of pos was zero!",1);
//
// }
// return tmp;
//}
//
//Vdouble likelihoodComputationGL::getLofPosPerCat(const int pos,
// const tree& tr,
// const sequenceContainer& sc,
// const computePijGam& pi,
// const stochasticProcess& sp)
//{
//// with the pi already computed.
// int numOfCat = sp.categories();
// Vdouble tmp;
// tmp.resize(numOfCat*2);
// for (int i=0; i < numOfCat;++i) {
// tmp[i] = getLofPos(pos,tr,sc,pi[i],sp)*sp.ratesProb(i);
// tmp[i+numOfCat] = sp.ratesProb(i);
// }
// return tmp;
//}
|
