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 + 0.50) / (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);
145 wordGenusProb[i][k] = log((count[genusNodes[k]] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
147 if (count[genusNodes[k]] != 0) {
150 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
155 out << k << '\t' << wordGenusProb[i][k] << '\t';
166 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
172 out2 << probabilityInTemplate << '\t' << numNotZero << endl;
180 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
192 //read in new phylotree with less info. - its faster
193 ifstream phyloTreeTest(phyloTreeName.c_str());
196 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
200 m->mothurOut("DONE."); m->mothurOutEndLine();
201 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
203 catch(exception& e) {
204 m->errorOut(e, "Bayesian", "Bayesian");
208 /**************************************************************************************************/
209 Bayesian::~Bayesian() {
213 if (database != NULL) { delete database; }
215 catch(exception& e) {
216 m->errorOut(e, "Bayesian", "~Bayesian");
221 /**************************************************************************************************/
222 string Bayesian::getTaxonomy(Sequence* seq) {
227 //get words contained in query
228 //getKmerString returns a string where the index in the string is hte kmer number
229 //and the character at that index can be converted to be the number of times that kmer was seen
231 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
233 vector<int> queryKmers;
234 for (int i = 0; i < queryKmerString.length()-1; i++) { // the -1 is to ignore any kmer with an N in it
235 if (queryKmerString[i] != '!') { //this kmer is in the query
236 queryKmers.push_back(i);
240 if (queryKmers.size() == 0) { m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); return "bad seq"; }
243 int index = getMostProbableTaxonomy(queryKmers);
245 if (m->control_pressed) { return tax; }
247 //bootstrap - to set confidenceScore
248 int numToSelect = queryKmers.size() / 8;
250 // tax = bootstrapResults(queryKmers, index, numToSelect);
254 catch(exception& e) {
255 m->errorOut(e, "Bayesian", "getTaxonomy");
259 /**************************************************************************************************/
260 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
263 map<int, int> confidenceScores;
265 //initialize confidences to 0
267 TaxNode seq = phyloTree->get(tax);
268 confidenceScores[tax] = 0;
270 while (seq.level != 0) { //while you are not at the root
271 seqIndex = seq.parent;
272 confidenceScores[seqIndex] = 0;
273 seq = phyloTree->get(seq.parent);
276 map<int, int>::iterator itBoot;
277 map<int, int>::iterator itBoot2;
278 map<int, int>::iterator itConvert;
280 for (int i = 0; i < iters; i++) {
281 if (m->control_pressed) { return "control"; }
284 for (int j = 0; j < numToSelect; j++) {
285 int index = int(rand() % kmers.size());
288 temp.push_back(kmers[index]);
292 int newTax = getMostProbableTaxonomy(temp);
294 TaxNode taxonomyTemp = phyloTree->get(newTax);
296 //add to confidence results
297 while (taxonomyTemp.level != 0) { //while you are not at the root
298 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
300 if (itBoot2 != confidenceScores.end()) { //this is a classification we need a confidence for
304 newTax = taxonomyTemp.parent;
305 taxonomyTemp = phyloTree->get(newTax);
310 string confidenceTax = "";
313 int seqTaxIndex = tax;
314 TaxNode seqTax = phyloTree->get(tax);
316 while (seqTax.level != 0) { //while you are not at the root
318 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
321 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
322 confidence = itBoot2->second;
325 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
326 confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
327 simpleTax = seqTax.name + ";" + simpleTax;
330 seqTaxIndex = seqTax.parent;
331 seqTax = phyloTree->get(seqTax.parent);
334 if (confidenceTax == "") { confidenceTax = "unclassified;"; simpleTax = "unclassified;"; }
335 return confidenceTax;
338 catch(exception& e) {
339 m->errorOut(e, "Bayesian", "bootstrapResults");
343 /**************************************************************************************************/
344 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
346 int indexofGenus = 0;
348 double maxProbability = -1000000.0;
349 //find taxonomy with highest probability that this sequence is from it
352 cout << genusNodes.size() << endl;
355 for (int k = 0; k < genusNodes.size(); k++) {
356 //for each taxonomy calc its probability
358 double prob = 0.0000;
359 for (int i = 0; i < queryKmer.size(); i++) {
360 prob += wordGenusProb[queryKmer[i]][k];
363 cout << phyloTree->get(genusNodes[k]).name << '\t' << prob << endl;
365 //is this the taxonomy with the greatest probability?
366 if (prob > maxProbability) {
367 indexofGenus = genusNodes[k];
368 maxProbability = prob;
375 catch(exception& e) {
376 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
380 /*************************************************************************************************
381 map<string, int> Bayesian::parseTaxMap(string newTax) {
384 map<string, int> parsed;
386 newTax = newTax.substr(0, newTax.length()-1); //get rid of last ';'
390 while (newTax.find_first_of(';') != -1) {
391 individual = newTax.substr(0,newTax.find_first_of(';'));
392 newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
393 parsed[individual] = 1;
402 catch(exception& e) {
403 m->errorOut(e, "Bayesian", "parseTax");
407 /**************************************************************************************************/
408 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
413 int pid, num, num2, processors;
414 vector<unsigned long int> positions;
415 vector<unsigned long int> positions2;
420 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
421 MPI_Comm_size(MPI_COMM_WORLD, &processors);
424 char inFileName[1024];
425 strcpy(inFileName, inNumName.c_str());
427 char inFileName2[1024];
428 strcpy(inFileName2, inName.c_str());
430 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
431 MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2); //comm, filename, mode, info, filepointer
434 positions = m->setFilePosEachLine(inNumName, num);
435 positions2 = m->setFilePosEachLine(inName, num2);
437 for(int i = 1; i < processors; i++) {
438 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
439 MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
441 MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
442 MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
446 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
447 positions.resize(num+1);
448 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
450 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
451 positions2.resize(num2+1);
452 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
456 int length = positions2[1] - positions2[0];
457 char* buf5 = new char[length];
459 MPI_File_read_at(inMPI2, positions2[0], buf5, length, MPI_CHAR, &status);
463 length = positions2[2] - positions2[1];
464 char* buf = new char[length];
466 MPI_File_read_at(inMPI2, positions2[1], buf, length, MPI_CHAR, &status);
468 string tempBuf = buf;
469 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
472 istringstream iss (tempBuf,istringstream::in);
475 //initialze probabilities
476 wordGenusProb.resize(numKmers);
478 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
481 vector<int> numbers; numbers.resize(numKmers);
483 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
486 length = positions[1] - positions[0];
487 char* buf6 = new char[length];
489 MPI_File_read_at(inMPI2, positions[0], buf6, length, MPI_CHAR, &status);
493 for(int i=1;i<num;i++){
495 length = positions[i+1] - positions[i];
496 char* buf4 = new char[length];
498 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
501 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
504 istringstream iss (tempBuf,istringstream::in);
505 iss >> zeroCountProb[i] >> numbers[i];
508 MPI_File_close(&inMPI);
510 for(int i=2;i<num2;i++){
512 length = positions2[i+1] - positions2[i];
513 char* buf4 = new char[length];
515 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
518 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
521 istringstream iss (tempBuf,istringstream::in);
525 //set them all to zero value
526 for (int i = 0; i < genusNodes.size(); i++) {
527 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
530 //get probs for nonzero values
531 for (int i = 0; i < numbers[kmer]; i++) {
533 wordGenusProb[kmer][name] = prob;
537 MPI_File_close(&inMPI2);
538 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
541 string line = m->getline(in); m->gobble(in);
543 in >> numKmers; m->gobble(in);
545 //initialze probabilities
546 wordGenusProb.resize(numKmers);
548 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
550 int kmer, name, count; count = 0;
551 vector<int> num; num.resize(numKmers);
553 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
556 string line2 = m->getline(inNum); m->gobble(inNum);
559 inNum >> zeroCountProb[count] >> num[count];
568 //set them all to zero value
569 for (int i = 0; i < genusNodes.size(); i++) {
570 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
573 //get probs for nonzero values
574 for (int i = 0; i < num[kmer]; i++) {
576 wordGenusProb[kmer][name] = prob;
585 catch(exception& e) {
586 m->errorOut(e, "Bayesian", "readProbFile");
590 /**************************************************************************************************/
591 bool Bayesian::checkReleaseDate(ifstream& file1, ifstream& file2, ifstream& file3, ifstream& file4) {
596 vector<string> lines;
597 lines.push_back(m->getline(file1));
598 lines.push_back(m->getline(file2));
599 lines.push_back(m->getline(file3));
600 lines.push_back(m->getline(file4));
602 //before we added this check
603 if ((lines[0][0] != '#') || (lines[1][0] != '#') || (lines[2][0] != '#') || (lines[3][0] != '#')) { good = false; }
606 for (int i = 0; i < lines.size(); i++) { lines[i] = lines[i].substr(1); }
608 //get mothurs current version
609 string version = m->getVersion();
611 vector<string> versionVector;
612 m->splitAtChar(version, versionVector, '.');
614 //check each files version
615 for (int i = 0; i < lines.size(); i++) {
616 vector<string> linesVector;
617 m->splitAtChar(lines[i], linesVector, '.');
619 if (versionVector.size() != linesVector.size()) { good = false; break; }
621 for (int j = 0; j < versionVector.size(); j++) {
623 convert(versionVector[j], num1);
624 convert(linesVector[j], num2);
626 //if mothurs version is newer than this files version, then we want to remake it
627 if (num1 > num2) { good = false; break; }
631 if (!good) { break; }
635 if (!good) { file1.close(); file2.close(); file3.close(); file4.close(); }
636 else { file1.seekg(0); file2.seekg(0); file3.seekg(0); file4.seekg(0); }
640 catch(exception& e) {
641 m->errorOut(e, "Bayesian", "checkReleaseDate");
645 /**************************************************************************************************/