5 * Created by westcott on 11/3/09.
6 * Copyright 2009 Schloss Lab. All rights reserved.
12 #include "phylosummary.h"
13 #include "referencedb.h"
14 /**************************************************************************************************/
15 Bayesian::Bayesian(string tfile, string tempFile, string method, int ksize, int cutoff, int i, int tid, bool f, bool sh) :
16 Classify(), kmerSize(ksize), confidenceThreshold(cutoff), iters(i) {
18 ReferenceDB* rdb = ReferenceDB::getInstance();
23 string baseName = tempFile;
25 if (baseName == "saved") { baseName = rdb->getSavedReference(); }
27 string baseTName = tfile;
28 if (baseTName == "saved") { baseTName = rdb->getSavedTaxonomy(); }
30 /************calculate the probablity that each word will be in a specific taxonomy*************/
31 string tfileroot = m->getFullPathName(baseTName.substr(0,baseTName.find_last_of(".")+1));
32 string tempfileroot = m->getRootName(m->getSimpleName(baseName));
33 string phyloTreeName = tfileroot + "tree.train";
34 string phyloTreeSumName = tfileroot + "tree.sum";
35 string probFileName = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.prob";
36 string probFileName2 = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.numNonZero";
41 ifstream phyloTreeTest(phyloTreeName.c_str());
42 ifstream probFileTest2(probFileName2.c_str());
43 ifstream probFileTest(probFileName.c_str());
44 ifstream probFileTest3(phyloTreeSumName.c_str());
46 int start = time(NULL);
48 //if they are there make sure they were created after this release date
49 bool FilesGood = false;
50 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3){
51 FilesGood = checkReleaseDate(probFileTest, probFileTest2, phyloTreeTest, probFileTest3);
54 //if you want to save, but you dont need to calculate then just read
55 if (rdb->save && probFileTest && probFileTest2 && phyloTreeTest && probFileTest3 && FilesGood && (tempFile != "saved")) {
57 m->openInputFile(tempFile, saveIn);
59 while (!saveIn.eof()) {
60 Sequence temp(saveIn);
63 rdb->referenceSeqs.push_back(temp);
68 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3 && FilesGood){
69 if (tempFile == "saved") { m->mothurOutEndLine(); m->mothurOut("Using sequences from " + rdb->getSavedReference() + " that are saved in memory."); m->mothurOutEndLine(); }
71 m->mothurOut("Reading template taxonomy... "); cout.flush();
73 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
75 m->mothurOut("DONE."); m->mothurOutEndLine();
77 genusNodes = phyloTree->getGenusNodes();
78 genusTotals = phyloTree->getGenusTotals();
80 if (tfile == "saved") {
81 m->mothurOutEndLine(); m->mothurOut("Using probabilties from " + rdb->getSavedTaxonomy() + " that are saved in memory... "); cout.flush();;
82 wordGenusProb = rdb->wordGenusProb;
83 WordPairDiffArr = rdb->WordPairDiffArr;
85 m->mothurOut("Reading template probabilities... "); cout.flush();
86 readProbFile(probFileTest, probFileTest2, probFileName, probFileName2);
90 if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
93 //create search database and names vector
94 generateDatabaseAndNames(tfile, tempFile, method, ksize, 0.0, 0.0, 0.0, 0.0);
96 //prevents errors caused by creating shortcut files if you had an error in the sanity check.
97 if (m->control_pressed) { m->mothurRemove(phyloTreeName); m->mothurRemove(probFileName); m->mothurRemove(probFileName2); }
99 genusNodes = phyloTree->getGenusNodes();
100 genusTotals = phyloTree->getGenusTotals();
102 m->mothurOut("Calculating template taxonomy tree... "); cout.flush();
104 phyloTree->printTreeNodes(phyloTreeName);
106 m->mothurOut("DONE."); m->mothurOutEndLine();
108 m->mothurOut("Calculating template probabilities... "); cout.flush();
110 numKmers = database->getMaxKmer() + 1;
112 //initialze probabilities
113 wordGenusProb.resize(numKmers);
114 WordPairDiffArr.resize(numKmers);
116 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
122 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
129 m->openOutputFile(probFileName, out);
131 //output mothur version
132 out << "#" << m->getVersion() << endl;
134 out << numKmers << endl;
136 m->openOutputFile(probFileName2, out2);
138 //output mothur version
139 out2 << "#" << m->getVersion() << endl;
147 for (int i = 0; i < numKmers; i++) {
148 //m->mothurOut("[DEBUG]: kmer = " + toString(i) + "\n");
150 if (m->control_pressed) { break; }
153 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
158 if (shortcuts) { out << i << '\t'; }
164 vector<int> seqsWithWordi = database->getSequencesWithKmer(i);
166 //for each sequence with that word
167 vector<int> count; count.resize(genusNodes.size(), 0);
168 for (int j = 0; j < seqsWithWordi.size(); j++) {
169 int temp = phyloTree->getGenusIndex(names[seqsWithWordi[j]]);
170 count[temp]++; //increment count of seq in this genus who have this word
173 //probabilityInTemplate = (# of seqs with that word in template + 0.50) / (total number of seqs in template + 1);
174 float probabilityInTemplate = (seqsWithWordi.size() + 0.50) / (float) (names.size() + 1);
175 diffPair tempProb(log(probabilityInTemplate), 0.0);
176 WordPairDiffArr[i] = tempProb;
179 for (int k = 0; k < genusNodes.size(); k++) {
180 //probabilityInThisTaxonomy = (# of seqs with that word in this taxonomy + probabilityInTemplate) / (total number of seqs in this taxonomy + 1);
183 wordGenusProb[i][k] = log((count[k] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
188 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
193 if (shortcuts) { out << k << '\t' << wordGenusProb[i][k] << '\t' ; }
204 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
211 out2 << probabilityInTemplate << '\t' << numNotZero << '\t' << log(probabilityInTemplate) << endl;
220 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
233 //read in new phylotree with less info. - its faster
234 ifstream phyloTreeTest(phyloTreeName.c_str());
237 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
240 if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
244 if (m->debug) { m->mothurOut("[DEBUG]: about to generateWordPairDiffArr\n"); }
245 generateWordPairDiffArr();
246 if (m->debug) { m->mothurOut("[DEBUG]: done generateWordPairDiffArr\n"); }
249 if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
251 m->mothurOut("DONE."); m->mothurOutEndLine();
252 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
254 catch(exception& e) {
255 m->errorOut(e, "Bayesian", "Bayesian");
259 /**************************************************************************************************/
260 Bayesian::~Bayesian() {
262 if (phyloTree != NULL) { delete phyloTree; }
263 if (database != NULL) { delete database; }
265 catch(exception& e) {
266 m->errorOut(e, "Bayesian", "~Bayesian");
271 /**************************************************************************************************/
272 string Bayesian::getTaxonomy(Sequence* seq) {
278 //get words contained in query
279 //getKmerString returns a string where the index in the string is hte kmer number
280 //and the character at that index can be converted to be the number of times that kmer was seen
281 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
283 vector<int> queryKmers;
284 for (int i = 0; i < queryKmerString.length()-1; i++) { // the -1 is to ignore any kmer with an N in it
285 if (queryKmerString[i] != '!') { //this kmer is in the query
286 queryKmers.push_back(i);
290 //if user wants to test reverse compliment and its reversed use that instead
292 if (isReversed(queryKmers)) {
294 seq->reverseComplement();
295 queryKmerString = kmer.getKmerString(seq->getUnaligned());
297 for (int i = 0; i < queryKmerString.length()-1; i++) { // the -1 is to ignore any kmer with an N in it
298 if (queryKmerString[i] != '!') { //this kmer is in the query
299 queryKmers.push_back(i);
305 if (queryKmers.size() == 0) { m->mothurOut(seq->getName() + " is bad. It has no kmers of length " + toString(kmerSize) + "."); m->mothurOutEndLine(); simpleTax = "unknown;"; return "unknown;"; }
308 int index = getMostProbableTaxonomy(queryKmers);
310 if (m->control_pressed) { return tax; }
312 //bootstrap - to set confidenceScore
313 int numToSelect = queryKmers.size() / 8;
315 if (m->debug) { m->mothurOut(seq->getName() + "\t"); }
317 tax = bootstrapResults(queryKmers, index, numToSelect);
319 if (m->debug) { m->mothurOut("\n"); }
323 catch(exception& e) {
324 m->errorOut(e, "Bayesian", "getTaxonomy");
328 /**************************************************************************************************/
329 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
332 map<int, int> confidenceScores;
334 //initialize confidences to 0
336 TaxNode seq = phyloTree->get(tax);
337 confidenceScores[tax] = 0;
339 while (seq.level != 0) { //while you are not at the root
340 seqIndex = seq.parent;
341 confidenceScores[seqIndex] = 0;
342 seq = phyloTree->get(seq.parent);
345 map<int, int>::iterator itBoot;
346 map<int, int>::iterator itBoot2;
347 map<int, int>::iterator itConvert;
349 for (int i = 0; i < iters; i++) {
350 if (m->control_pressed) { return "control"; }
353 for (int j = 0; j < numToSelect; j++) {
354 int index = int(rand() % kmers.size());
357 temp.push_back(kmers[index]);
361 int newTax = getMostProbableTaxonomy(temp);
363 TaxNode taxonomyTemp = phyloTree->get(newTax);
365 //add to confidence results
366 while (taxonomyTemp.level != 0) { //while you are not at the root
367 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
369 if (itBoot2 != confidenceScores.end()) { //this is a classification we need a confidence for
373 newTax = taxonomyTemp.parent;
374 taxonomyTemp = phyloTree->get(newTax);
379 string confidenceTax = "";
382 int seqTaxIndex = tax;
383 TaxNode seqTax = phyloTree->get(tax);
386 while (seqTax.level != 0) { //while you are not at the root
388 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
391 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
392 confidence = itBoot2->second;
395 if (m->debug) { m->mothurOut(seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");"); }
397 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
398 confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
399 simpleTax = seqTax.name + ";" + simpleTax;
402 seqTaxIndex = seqTax.parent;
403 seqTax = phyloTree->get(seqTax.parent);
406 if (confidenceTax == "") { confidenceTax = "unknown;"; simpleTax = "unknown;"; }
408 return confidenceTax;
411 catch(exception& e) {
412 m->errorOut(e, "Bayesian", "bootstrapResults");
416 /**************************************************************************************************/
417 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
419 int indexofGenus = 0;
421 double maxProbability = -1000000.0;
422 //find taxonomy with highest probability that this sequence is from it
425 // cout << genusNodes.size() << endl;
428 for (int k = 0; k < genusNodes.size(); k++) {
429 //for each taxonomy calc its probability
431 double prob = 0.0000;
432 for (int i = 0; i < queryKmer.size(); i++) {
433 prob += wordGenusProb[queryKmer[i]][k];
436 // cout << phyloTree->get(genusNodes[k]).name << '\t' << prob << endl;
438 //is this the taxonomy with the greatest probability?
439 if (prob > maxProbability) {
440 indexofGenus = genusNodes[k];
441 maxProbability = prob;
448 catch(exception& e) {
449 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
453 //********************************************************************************************************************
454 //if it is more probable that the reverse compliment kmers are in the template, then we assume the sequence is reversed.
455 bool Bayesian::isReversed(vector<int>& queryKmers){
457 bool reversed = false;
459 float reverseProb = 0;
461 for (int i = 0; i < queryKmers.size(); i++){
462 int kmer = queryKmers[i];
464 prob += WordPairDiffArr[kmer].prob;
465 reverseProb += WordPairDiffArr[kmer].reverseProb;
469 if (reverseProb > prob){ reversed = true; }
473 catch(exception& e) {
474 m->errorOut(e, "Bayesian", "isReversed");
478 //********************************************************************************************************************
479 int Bayesian::generateWordPairDiffArr(){
482 for (int i = 0; i < WordPairDiffArr.size(); i++) {
483 int reversedWord = kmer.getReverseKmerNumber(i);
484 WordPairDiffArr[i].reverseProb = WordPairDiffArr[reversedWord].prob;
488 }catch(exception& e) {
489 m->errorOut(e, "Bayesian", "generateWordPairDiffArr");
493 /*************************************************************************************************
494 map<string, int> Bayesian::parseTaxMap(string newTax) {
497 map<string, int> parsed;
499 newTax = newTax.substr(0, newTax.length()-1); //get rid of last ';'
503 while (newTax.find_first_of(';') != -1) {
504 individual = newTax.substr(0,newTax.find_first_of(';'));
505 newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
506 parsed[individual] = 1;
515 catch(exception& e) {
516 m->errorOut(e, "Bayesian", "parseTax");
520 **************************************************************************************************/
521 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
526 int pid, num, num2, processors;
527 vector<unsigned long long> positions;
528 vector<unsigned long long> positions2;
533 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
534 MPI_Comm_size(MPI_COMM_WORLD, &processors);
537 char inFileName[1024];
538 strcpy(inFileName, inNumName.c_str());
540 char inFileName2[1024];
541 strcpy(inFileName2, inName.c_str());
543 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
544 MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2); //comm, filename, mode, info, filepointer
547 positions = m->setFilePosEachLine(inNumName, num);
548 positions2 = m->setFilePosEachLine(inName, num2);
550 for(int i = 1; i < processors; i++) {
551 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
552 MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
554 MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
555 MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
559 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
560 positions.resize(num+1);
561 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
563 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
564 positions2.resize(num2+1);
565 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
569 int length = positions2[1] - positions2[0];
570 char* buf5 = new char[length];
572 MPI_File_read_at(inMPI2, positions2[0], buf5, length, MPI_CHAR, &status);
576 length = positions2[2] - positions2[1];
577 char* buf = new char[length];
579 MPI_File_read_at(inMPI2, positions2[1], buf, length, MPI_CHAR, &status);
581 string tempBuf = buf;
582 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
585 istringstream iss (tempBuf,istringstream::in);
588 //initialze probabilities
589 wordGenusProb.resize(numKmers);
591 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
594 vector<int> numbers; numbers.resize(numKmers);
596 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
597 WordPairDiffArr.resize(numKmers);
600 length = positions[1] - positions[0];
601 char* buf6 = new char[length];
603 MPI_File_read_at(inMPI2, positions[0], buf6, length, MPI_CHAR, &status);
607 for(int i=1;i<num;i++){
609 length = positions[i+1] - positions[i];
610 char* buf4 = new char[length];
612 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
615 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
618 istringstream iss (tempBuf,istringstream::in);
620 iss >> zeroCountProb[i] >> numbers[i] >> probTemp;
621 WordPairDiffArr[i].prob = probTemp;
625 MPI_File_close(&inMPI);
627 for(int i=2;i<num2;i++){
629 length = positions2[i+1] - positions2[i];
630 char* buf4 = new char[length];
632 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
635 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
638 istringstream iss (tempBuf,istringstream::in);
642 //set them all to zero value
643 for (int i = 0; i < genusNodes.size(); i++) {
644 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
647 //get probs for nonzero values
648 for (int i = 0; i < numbers[kmer]; i++) {
650 wordGenusProb[kmer][name] = prob;
654 MPI_File_close(&inMPI2);
655 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
658 string line = m->getline(in); m->gobble(in);
660 in >> numKmers; m->gobble(in);
661 //cout << threadID << '\t' << line << '\t' << numKmers << &in << '\t' << &inNum << '\t' << genusNodes.size() << endl;
662 //initialze probabilities
663 wordGenusProb.resize(numKmers);
665 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
667 int kmer, name, count; count = 0;
668 vector<int> num; num.resize(numKmers);
670 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
671 WordPairDiffArr.resize(numKmers);
674 string line2 = m->getline(inNum); m->gobble(inNum);
676 //cout << threadID << '\t' << line2 << '\t' << this << endl;
678 inNum >> zeroCountProb[count] >> num[count] >> probTemp;
679 WordPairDiffArr[count].prob = probTemp;
682 //cout << threadID << '\t' << count << endl;
685 //cout << threadID << '\t' << "here1 " << &wordGenusProb << '\t' << &num << endl; //
686 //cout << threadID << '\t' << &genusTotals << '\t' << endl;
687 //cout << threadID << '\t' << genusNodes.size() << endl;
690 //cout << threadID << '\t' << kmer << endl;
691 //set them all to zero value
692 for (int i = 0; i < genusNodes.size(); i++) {
693 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
695 //cout << threadID << '\t' << num[kmer] << "here" << endl;
696 //get probs for nonzero values
697 for (int i = 0; i < num[kmer]; i++) {
699 wordGenusProb[kmer][name] = prob;
705 //cout << threadID << '\t' << "here" << endl;
708 catch(exception& e) {
709 m->errorOut(e, "Bayesian", "readProbFile");
713 /**************************************************************************************************/
714 bool Bayesian::checkReleaseDate(ifstream& file1, ifstream& file2, ifstream& file3, ifstream& file4) {
719 vector<string> lines;
720 lines.push_back(m->getline(file1));
721 lines.push_back(m->getline(file2));
722 lines.push_back(m->getline(file3));
723 lines.push_back(m->getline(file4));
725 //before we added this check
726 if ((lines[0][0] != '#') || (lines[1][0] != '#') || (lines[2][0] != '#') || (lines[3][0] != '#')) { good = false; }
729 for (int i = 0; i < lines.size(); i++) { lines[i] = lines[i].substr(1); }
731 //get mothurs current version
732 string version = m->getVersion();
734 vector<string> versionVector;
735 m->splitAtChar(version, versionVector, '.');
737 //check each files version
738 for (int i = 0; i < lines.size(); i++) {
739 vector<string> linesVector;
740 m->splitAtChar(lines[i], linesVector, '.');
742 if (versionVector.size() != linesVector.size()) { good = false; break; }
744 for (int j = 0; j < versionVector.size(); j++) {
746 convert(versionVector[j], num1);
747 convert(linesVector[j], num2);
749 //if mothurs version is newer than this files version, then we want to remake it
750 if (num1 > num2) { good = false; break; }
754 if (!good) { break; }
758 if (!good) { file1.close(); file2.close(); file3.close(); file4.close(); }
759 else { file1.seekg(0); file2.seekg(0); file3.seekg(0); file4.seekg(0); }
763 catch(exception& e) {
764 m->errorOut(e, "Bayesian", "checkReleaseDate");
768 /**************************************************************************************************/