5 * Created by westcott on 11/3/09.
6 * Copyright 2009 Schloss Lab. All rights reserved.
12 #include "phylosummary.h"
14 /**************************************************************************************************/
15 Bayesian::Bayesian(string tfile, string tempFile, string method, int ksize, int cutoff, int i) :
16 Classify(), kmerSize(ksize), confidenceThreshold(cutoff), iters(i) {
19 /************calculate the probablity that each word will be in a specific taxonomy*************/
20 string tfileroot = tfile.substr(0,tfile.find_last_of(".")+1);
21 string tempfileroot = m->getRootName(m->getSimpleName(tempFile));
22 string phyloTreeName = tfileroot + "tree.train";
23 string phyloTreeSumName = tfileroot + "tree.sum";
24 string probFileName = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.prob";
25 string probFileName2 = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.numNonZero";
30 ifstream phyloTreeTest(phyloTreeName.c_str());
31 ifstream probFileTest2(probFileName2.c_str());
32 ifstream probFileTest(probFileName.c_str());
33 ifstream probFileTest3(phyloTreeSumName.c_str());
35 int start = time(NULL);
37 //if they are there make sure they were created after this release date
38 bool FilesGood = false;
39 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3){
40 FilesGood = checkReleaseDate(probFileTest, probFileTest2, phyloTreeTest, probFileTest3);
43 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3 && FilesGood){
44 m->mothurOut("Reading template taxonomy... "); cout.flush();
46 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
48 m->mothurOut("DONE."); m->mothurOutEndLine();
50 genusNodes = phyloTree->getGenusNodes();
51 genusTotals = phyloTree->getGenusTotals();
53 m->mothurOut("Reading template probabilities... "); cout.flush();
54 readProbFile(probFileTest, probFileTest2, probFileName, probFileName2);
58 //create search database and names vector
59 generateDatabaseAndNames(tfile, tempFile, method, ksize, 0.0, 0.0, 0.0, 0.0);
61 //prevents errors caused by creating shortcut files if you had an error in the sanity check.
62 if (m->control_pressed) { remove(phyloTreeName.c_str()); remove(probFileName.c_str()); remove(probFileName2.c_str()); }
64 genusNodes = phyloTree->getGenusNodes();
65 genusTotals = phyloTree->getGenusTotals();
67 m->mothurOut("Calculating template taxonomy tree... "); cout.flush();
69 phyloTree->printTreeNodes(phyloTreeName);
71 m->mothurOut("DONE."); m->mothurOutEndLine();
73 m->mothurOut("Calculating template probabilities... "); cout.flush();
75 numKmers = database->getMaxKmer() + 1;
77 //initialze probabilities
78 wordGenusProb.resize(numKmers);
79 //cout << numKmers << '\t' << genusNodes.size() << endl;
80 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
81 //cout << numKmers << '\t' << genusNodes.size() << endl;
87 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
93 m->openOutputFile(probFileName, out);
95 //output mothur version
96 out << "#" << m->getVersion() << endl;
98 out << numKmers << endl;
100 m->openOutputFile(probFileName2, out2);
102 //output mothur version
103 out2 << "#" << m->getVersion() << endl;
111 for (int i = 0; i < numKmers; i++) {
112 if (m->control_pressed) { break; }
115 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
126 vector<int> seqsWithWordi = database->getSequencesWithKmer(i);
129 for (int k = 0; k < genusNodes.size(); k++) { count[genusNodes[k]] = 0; }
131 //for each sequence with that word
132 for (int j = 0; j < seqsWithWordi.size(); j++) {
133 int temp = phyloTree->getIndex(names[seqsWithWordi[j]]);
134 count[temp]++; //increment count of seq in this genus who have this word
137 //probabilityInTemplate = (# of seqs with that word in template + 0.05) / (total number of seqs in template + 1);
138 float probabilityInTemplate = (seqsWithWordi.size() + 0.50) / (float) (names.size() + 1);
141 for (int k = 0; k < genusNodes.size(); k++) {
142 //probabilityInThisTaxonomy = (# of seqs with that word in this taxonomy + probabilityInTemplate) / (total number of seqs in this taxonomy + 1);
143 wordGenusProb[i][k] = log((count[genusNodes[k]] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
144 if (count[genusNodes[k]] != 0) {
147 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
152 out << k << '\t' << wordGenusProb[i][k] << '\t';
163 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
169 out2 << probabilityInTemplate << '\t' << numNotZero << endl;
177 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
189 //read in new phylotree with less info. - its faster
190 ifstream phyloTreeTest(phyloTreeName.c_str());
193 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
197 m->mothurOut("DONE."); m->mothurOutEndLine();
198 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
200 catch(exception& e) {
201 m->errorOut(e, "Bayesian", "Bayesian");
205 /**************************************************************************************************/
206 Bayesian::~Bayesian() {
210 if (database != NULL) { delete database; }
212 catch(exception& e) {
213 m->errorOut(e, "Bayesian", "~Bayesian");
218 /**************************************************************************************************/
219 string Bayesian::getTaxonomy(Sequence* seq) {
224 //get words contained in query
225 //getKmerString returns a string where the index in the string is hte kmer number
226 //and the character at that index can be converted to be the number of times that kmer was seen
228 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
230 vector<int> queryKmers;
231 for (int i = 0; i < queryKmerString.length(); i++) {
232 if (queryKmerString[i] != '!') { //this kmer is in the query
233 queryKmers.push_back(i);
237 if (queryKmers.size() == 0) { m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); return "bad seq"; }
240 int index = getMostProbableTaxonomy(queryKmers);
242 if (m->control_pressed) { return tax; }
244 //bootstrap - to set confidenceScore
245 int numToSelect = queryKmers.size() / 8;
247 tax = bootstrapResults(queryKmers, index, numToSelect);
251 catch(exception& e) {
252 m->errorOut(e, "Bayesian", "getTaxonomy");
256 /**************************************************************************************************/
257 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
260 map<int, int> confidenceScores;
262 //initialize confidences to 0
264 TaxNode seq = phyloTree->get(tax);
265 confidenceScores[tax] = 0;
267 while (seq.level != 0) { //while you are not at the root
268 seqIndex = seq.parent;
269 confidenceScores[seqIndex] = 0;
270 seq = phyloTree->get(seq.parent);
273 map<int, int>::iterator itBoot;
274 map<int, int>::iterator itBoot2;
275 map<int, int>::iterator itConvert;
277 for (int i = 0; i < iters; i++) {
278 if (m->control_pressed) { return "control"; }
281 for (int j = 0; j < numToSelect; j++) {
282 int index = int(rand() % kmers.size());
285 temp.push_back(kmers[index]);
289 int newTax = getMostProbableTaxonomy(temp);
291 TaxNode taxonomyTemp = phyloTree->get(newTax);
293 //add to confidence results
294 while (taxonomyTemp.level != 0) { //while you are not at the root
295 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
297 if (itBoot2 != confidenceScores.end()) { //this is a classification we need a confidence for
301 newTax = taxonomyTemp.parent;
302 taxonomyTemp = phyloTree->get(newTax);
307 string confidenceTax = "";
310 int seqTaxIndex = tax;
311 TaxNode seqTax = phyloTree->get(tax);
313 while (seqTax.level != 0) { //while you are not at the root
315 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
318 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
319 confidence = itBoot2->second;
322 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
323 confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
324 simpleTax = seqTax.name + ";" + simpleTax;
327 seqTaxIndex = seqTax.parent;
328 seqTax = phyloTree->get(seqTax.parent);
331 if (confidenceTax == "") { confidenceTax = "unclassified;"; simpleTax = "unclassified;"; }
332 return confidenceTax;
335 catch(exception& e) {
336 m->errorOut(e, "Bayesian", "bootstrapResults");
340 /**************************************************************************************************/
341 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
343 int indexofGenus = 0;
345 double maxProbability = -1000000.0;
346 //find taxonomy with highest probability that this sequence is from it
347 for (int k = 0; k < genusNodes.size(); k++) {
348 //for each taxonomy calc its probability
350 for (int i = 0; i < queryKmer.size(); i++) {
351 prob += wordGenusProb[queryKmer[i]][k];
354 //is this the taxonomy with the greatest probability?
355 if (prob > maxProbability) {
356 indexofGenus = genusNodes[k];
357 maxProbability = prob;
360 // cout << phyloTree->get(indexofGenus).name << '\t' << maxProbability << endl;
363 catch(exception& e) {
364 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
368 /*************************************************************************************************
369 map<string, int> Bayesian::parseTaxMap(string newTax) {
372 map<string, int> parsed;
374 newTax = newTax.substr(0, newTax.length()-1); //get rid of last ';'
378 while (newTax.find_first_of(';') != -1) {
379 individual = newTax.substr(0,newTax.find_first_of(';'));
380 newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
381 parsed[individual] = 1;
390 catch(exception& e) {
391 m->errorOut(e, "Bayesian", "parseTax");
395 /**************************************************************************************************/
396 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
401 int pid, num, num2, processors;
402 vector<unsigned long int> positions;
403 vector<unsigned long int> positions2;
408 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
409 MPI_Comm_size(MPI_COMM_WORLD, &processors);
412 char inFileName[1024];
413 strcpy(inFileName, inNumName.c_str());
415 char inFileName2[1024];
416 strcpy(inFileName2, inName.c_str());
418 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
419 MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2); //comm, filename, mode, info, filepointer
422 positions = m->setFilePosEachLine(inNumName, num);
423 positions2 = m->setFilePosEachLine(inName, num2);
425 for(int i = 1; i < processors; i++) {
426 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
427 MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
429 MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
430 MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
434 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
435 positions.resize(num+1);
436 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
438 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
439 positions2.resize(num2+1);
440 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
444 int length = positions2[1] - positions2[0];
445 char* buf5 = new char[length];
447 MPI_File_read_at(inMPI2, positions2[0], buf5, length, MPI_CHAR, &status);
451 length = positions2[2] - positions2[1];
452 char* buf = new char[length];
454 MPI_File_read_at(inMPI2, positions2[1], buf, length, MPI_CHAR, &status);
456 string tempBuf = buf;
457 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
460 istringstream iss (tempBuf,istringstream::in);
463 //initialze probabilities
464 wordGenusProb.resize(numKmers);
466 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
469 vector<int> numbers; numbers.resize(numKmers);
471 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
474 length = positions[1] - positions[0];
475 char* buf6 = new char[length];
477 MPI_File_read_at(inMPI2, positions[0], buf6, length, MPI_CHAR, &status);
481 for(int i=1;i<num;i++){
483 length = positions[i+1] - positions[i];
484 char* buf4 = new char[length];
486 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
489 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
492 istringstream iss (tempBuf,istringstream::in);
493 iss >> zeroCountProb[i] >> numbers[i];
496 MPI_File_close(&inMPI);
498 for(int i=2;i<num2;i++){
500 length = positions2[i+1] - positions2[i];
501 char* buf4 = new char[length];
503 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
506 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
509 istringstream iss (tempBuf,istringstream::in);
513 //set them all to zero value
514 for (int i = 0; i < genusNodes.size(); i++) {
515 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
518 //get probs for nonzero values
519 for (int i = 0; i < numbers[kmer]; i++) {
521 wordGenusProb[kmer][name] = prob;
525 MPI_File_close(&inMPI2);
526 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
529 string line = m->getline(in); m->gobble(in);
531 in >> numKmers; m->gobble(in);
533 //initialze probabilities
534 wordGenusProb.resize(numKmers);
536 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
538 int kmer, name, count; count = 0;
539 vector<int> num; num.resize(numKmers);
541 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
544 string line2 = m->getline(inNum); m->gobble(inNum);
547 inNum >> zeroCountProb[count] >> num[count];
556 //set them all to zero value
557 for (int i = 0; i < genusNodes.size(); i++) {
558 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
561 //get probs for nonzero values
562 for (int i = 0; i < num[kmer]; i++) {
564 wordGenusProb[kmer][name] = prob;
573 catch(exception& e) {
574 m->errorOut(e, "Bayesian", "readProbFile");
578 /**************************************************************************************************/
579 bool Bayesian::checkReleaseDate(ifstream& file1, ifstream& file2, ifstream& file3, ifstream& file4) {
584 vector<string> lines;
585 lines.push_back(m->getline(file1));
586 lines.push_back(m->getline(file2));
587 lines.push_back(m->getline(file3));
588 lines.push_back(m->getline(file4));
590 //before we added this check
591 if ((lines[0][0] != '#') || (lines[1][0] != '#') || (lines[2][0] != '#') || (lines[3][0] != '#')) { good = false; }
594 for (int i = 0; i < lines.size(); i++) { lines[i] = lines[i].substr(1); }
596 //get mothurs current version
597 string version = m->getVersion();
599 vector<string> versionVector;
600 m->splitAtChar(version, versionVector, '.');
602 //check each files version
603 for (int i = 0; i < lines.size(); i++) {
604 vector<string> linesVector;
605 m->splitAtChar(lines[i], linesVector, '.');
607 if (versionVector.size() != linesVector.size()) { good = false; break; }
609 for (int j = 0; j < versionVector.size(); j++) {
611 convert(versionVector[j], num1);
612 convert(linesVector[j], num2);
614 //if mothurs version is newer than this files version, then we want to remake it
615 if (num1 > num2) { good = false; break; }
619 if (!good) { break; }
623 if (!good) { file1.close(); file2.close(); file3.close(); file4.close(); }
624 else { file1.seekg(0); file2.seekg(0); file3.seekg(0); file4.seekg(0); }
628 catch(exception& e) {
629 m->errorOut(e, "Bayesian", "checkReleaseDate");
633 /**************************************************************************************************/