forked from soedinglab/MMseqs2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
IndexBuilder.cpp
262 lines (224 loc) · 10.5 KB
/
IndexBuilder.cpp
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
#include "IndexBuilder.h"
#include "tantan.h"
#ifdef OPENMP
#include <omp.h>
#endif
char* getScoreLookup(BaseMatrix &matrix) {
char *idScoreLookup = NULL;
idScoreLookup = new char[matrix.alphabetSize];
for (int aa = 0; aa < matrix.alphabetSize; aa++){
short score = matrix.subMatrix[aa][aa];
if (score > SCHAR_MAX || score < SCHAR_MIN) {
Debug(Debug::WARNING) << "Truncating substitution matrix diagonal score!";
}
idScoreLookup[aa] = (char) score;
}
return idScoreLookup;
}
class DbInfo {
public:
DbInfo(size_t dbFrom, size_t dbTo, unsigned int effectiveKmerSize, DBReader<unsigned int> & reader) {
tableSize = 0;
aaDbSize = 0;
size_t dbSize = dbTo - dbFrom;
sequenceOffsets = new size_t[dbSize];
sequenceOffsets[0] = 0;
for (size_t id = dbFrom; id < dbTo; id++) {
const int seqLen = reader.getSeqLen(id);
aaDbSize += seqLen;
size_t idFromNull = (id - dbFrom);
if (id < dbTo - 1) {
sequenceOffsets[idFromNull + 1] = sequenceOffsets[idFromNull] + seqLen;
}
if (Util::overlappingKmers(seqLen, effectiveKmerSize > 0)) {
tableSize += 1;
}
}
}
~DbInfo() {
delete[] sequenceOffsets;
}
size_t tableSize;
size_t aaDbSize;
size_t *sequenceOffsets;
};
void IndexBuilder::fillDatabase(IndexTable *indexTable, SequenceLookup **maskedLookup,
SequenceLookup **unmaskedLookup,BaseMatrix &subMat, Sequence *seq,
DBReader<unsigned int> *dbr, size_t dbFrom, size_t dbTo, int kmerThr,
bool mask, bool maskLowerCaseMode) {
Debug(Debug::INFO) << "Index table: counting k-mers\n";
const bool isProfile = Parameters::isEqualDbtype(seq->getSeqType(), Parameters::DBTYPE_HMM_PROFILE);
dbTo = std::min(dbTo, dbr->getSize());
size_t dbSize = dbTo - dbFrom;
DbInfo* info = new DbInfo(dbFrom, dbTo, seq->getEffectiveKmerSize(), *dbr);
SequenceLookup *sequenceLookup;
if (unmaskedLookup != NULL && maskedLookup == NULL) {
*unmaskedLookup = new SequenceLookup(dbSize, info->aaDbSize);
sequenceLookup = *unmaskedLookup;
} else if (unmaskedLookup == NULL && maskedLookup != NULL) {
*maskedLookup = new SequenceLookup(dbSize, info->aaDbSize);
sequenceLookup = *maskedLookup;
} else if (unmaskedLookup != NULL && maskedLookup != NULL) {
*unmaskedLookup = new SequenceLookup(dbSize, info->aaDbSize);
*maskedLookup = new SequenceLookup(dbSize, info->aaDbSize);
sequenceLookup = *maskedLookup;
} else{
Debug(Debug::ERROR) << "This should not happen\n";
EXIT(EXIT_FAILURE);
}
// need to prune low scoring k-mers through masking
ProbabilityMatrix *probMatrix = NULL;
if (maskedLookup != NULL) {
probMatrix = new ProbabilityMatrix(subMat);
}
// identical scores for memory reduction code
char *idScoreLookup = NULL;
if (Parameters::isEqualDbtype(seq->getSeqType(), Parameters::DBTYPE_PROFILE_STATE_SEQ) == false) {
idScoreLookup = getScoreLookup(subMat);
}
Debug::Progress progress(dbTo-dbFrom);
size_t maskedResidues = 0;
size_t totalKmerCount = 0;
#pragma omp parallel
{
unsigned int thread_idx = 0;
#ifdef OPENMP
thread_idx = static_cast<unsigned int>(omp_get_thread_num());
#endif
Indexer idxer(static_cast<unsigned int>(indexTable->getAlphabetSize()), seq->getKmerSize());
Sequence s(seq->getMaxLen(), seq->getSeqType(), &subMat, seq->getKmerSize(), seq->isSpaced(), false, true, seq->getSpacedKmerPattern());
KmerGenerator *generator = NULL;
if (isProfile) {
generator = new KmerGenerator(seq->getKmerSize(), indexTable->getAlphabetSize(), kmerThr);
generator->setDivideStrategy(s.profile_matrix);
}
unsigned int *buffer = new unsigned int[seq->getMaxLen()];
char *charSequence = new char[seq->getMaxLen()];
#pragma omp for schedule(dynamic, 100) reduction(+:totalKmerCount, maskedResidues)
for (size_t id = dbFrom; id < dbTo; id++) {
progress.updateProgress();
s.resetCurrPos();
char *seqData = dbr->getData(id, thread_idx);
unsigned int qKey = dbr->getDbKey(id);
s.mapSequence(id - dbFrom, qKey, seqData, dbr->getSeqLen(id));
// count similar or exact k-mers based on sequence type
if (isProfile) {
// Find out if we should also mask profiles
totalKmerCount += indexTable->addSimilarKmerCount(&s, generator);
(*unmaskedLookup)->addSequence(s.int_consensus_sequence, s.L, id - dbFrom, info->sequenceOffsets[id - dbFrom]);
} else {
// Do not mask if column state sequences are used
if (unmaskedLookup != NULL) {
(*unmaskedLookup)->addSequence(s.int_sequence, s.L, id - dbFrom, info->sequenceOffsets[id - dbFrom]);
}
if (mask == true) {
for (int i = 0; i < s.L; ++i) {
charSequence[i] = (char) s.int_sequence[i];
}
// s.print();
maskedResidues += tantan::maskSequences(charSequence,
charSequence + s.L,
50 /*options.maxCycleLength*/,
probMatrix->probMatrixPointers,
0.005 /*options.repeatProb*/,
0.05 /*options.repeatEndProb*/,
0.9 /*options.repeatOffsetProbDecay*/,
0, 0,
0.9 /*options.minMaskProb*/,
probMatrix->hardMaskTable);
for (int i = 0; i < s.L; i++) {
s.int_sequence[i] = charSequence[i];
}
}
if(maskLowerCaseMode == true && (Parameters::isEqualDbtype(s.getSequenceType(), Parameters::DBTYPE_AMINO_ACIDS) ||
Parameters::isEqualDbtype(s.getSequenceType(), Parameters::DBTYPE_NUCLEOTIDES))) {
const char * charSeq = s.getSeqData();
int maskLetter = subMat.aa2int[(int)'X'];
for (int i = 0; i < s.L; i++) {
bool isLowerCase = (islower(charSeq[i]));
maskedResidues += isLowerCase;
s.int_sequence[i] = isLowerCase ? maskLetter : s.int_sequence[i];
}
}
if(maskedLookup != NULL){
(*maskedLookup)->addSequence(s.int_sequence, s.L, id - dbFrom, info->sequenceOffsets[id - dbFrom]);
}
totalKmerCount += indexTable->addKmerCount(&s, &idxer, buffer, kmerThr, idScoreLookup);
}
}
delete[] charSequence;
delete[] buffer;
if (generator != NULL) {
delete generator;
}
}
if(probMatrix != NULL) {
delete probMatrix;
}
Debug(Debug::INFO) << "Index table: Masked residues: " << maskedResidues << "\n";
if(totalKmerCount == 0) {
Debug(Debug::ERROR) << "No k-mer could be extracted for the database " << dbr->getDataFileName() << ".\n"
<< "Maybe the sequences length is less than 14 residues.\n";
if (maskedResidues == true){
Debug(Debug::ERROR) << " or contains only low complexity regions.";
Debug(Debug::ERROR) << "Use --mask 0 to deactivate the low complexity filter.\n";
}
EXIT(EXIT_FAILURE);
}
dbr->remapData();
//TODO find smart way to remove extrem k-mers without harming huge protein families
// size_t lowSelectiveResidues = 0;
// const float dbSize = static_cast<float>(dbTo - dbFrom);
// for(size_t kmerIdx = 0; kmerIdx < indexTable->getTableSize(); kmerIdx++){
// size_t res = (size_t) indexTable->getOffset(kmerIdx);
// float selectivityOfKmer = (static_cast<float>(res)/dbSize);
// if(selectivityOfKmer > 0.005){
// indexTable->getOffset()[kmerIdx] = 0;
// lowSelectiveResidues += res;
// }
// }
// Debug(Debug::INFO) << "Index table: Remove "<< lowSelectiveResidues <<" none selective residues\n";
// Debug(Debug::INFO) << "Index table: init... from "<< dbFrom << " to "<< dbTo << "\n";
indexTable->initMemory(info->tableSize);
indexTable->init();
delete info;
Debug::Progress progress2(dbTo-dbFrom);
Debug(Debug::INFO) << "Index table: fill\n";
#pragma omp parallel
{
unsigned int thread_idx = 0;
#ifdef OPENMP
thread_idx = static_cast<unsigned int>(omp_get_thread_num());
#endif
Sequence s(seq->getMaxLen(), seq->getSeqType(), &subMat, seq->getKmerSize(), seq->isSpaced(), false, true, seq->getSpacedKmerPattern());
Indexer idxer(static_cast<unsigned int>(indexTable->getAlphabetSize()), seq->getKmerSize());
IndexEntryLocalTmp *buffer = new IndexEntryLocalTmp[seq->getMaxLen()];
KmerGenerator *generator = NULL;
if (isProfile) {
generator = new KmerGenerator(seq->getKmerSize(), indexTable->getAlphabetSize(), kmerThr);
generator->setDivideStrategy(s.profile_matrix);
}
#pragma omp for schedule(dynamic, 100)
for (size_t id = dbFrom; id < dbTo; id++) {
s.resetCurrPos();
progress2.updateProgress();
unsigned int qKey = dbr->getDbKey(id);
if (isProfile) {
s.mapSequence(id - dbFrom, qKey, dbr->getData(id, thread_idx), dbr->getSeqLen(id));
indexTable->addSimilarSequence(&s, generator, &idxer);
} else {
s.mapSequence(id - dbFrom, qKey, sequenceLookup->getSequence(id - dbFrom));
indexTable->addSequence(&s, &idxer, buffer, kmerThr, idScoreLookup);
}
}
if (generator != NULL) {
delete generator;
}
delete [] buffer;
}
if(idScoreLookup!=NULL){
delete[] idScoreLookup;
}
indexTable->revertPointer();
indexTable->sortDBSeqLists();
}