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 = getRootName(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);
80 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
87 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
93 openOutputFile(probFileName, out);
95 //output mothur version
96 out << "#" << m->getVersion() << endl;
98 out << numKmers << endl;
100 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() {
209 if (database != NULL) { delete database; }
211 catch(exception& e) {
212 m->errorOut(e, "Bayesian", "~Bayesian");
217 /**************************************************************************************************/
218 string Bayesian::getTaxonomy(Sequence* seq) {
223 //get words contained in query
224 //getKmerString returns a string where the index in the string is hte kmer number
225 //and the character at that index can be converted to be the number of times that kmer was seen
227 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
229 vector<int> queryKmers;
230 for (int i = 0; i < queryKmerString.length(); i++) {
231 if (queryKmerString[i] != '!') { //this kmer is in the query
232 queryKmers.push_back(i);
236 if (queryKmers.size() == 0) { m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); return "bad seq"; }
238 int index = getMostProbableTaxonomy(queryKmers);
240 if (m->control_pressed) { return tax; }
241 //cout << seq->getName() << '\t' << index << endl;
242 //bootstrap - to set confidenceScore
243 int numToSelect = queryKmers.size() / 8;
244 tax = bootstrapResults(queryKmers, index, numToSelect);
248 catch(exception& e) {
249 m->errorOut(e, "Bayesian", "getTaxonomy");
253 /**************************************************************************************************/
254 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
257 map<int, int> confidenceScores;
259 map<int, int>::iterator itBoot;
260 map<int, int>::iterator itBoot2;
261 map<int, int>::iterator itConvert;
263 for (int i = 0; i < iters; i++) {
264 if (m->control_pressed) { return "control"; }
268 for (int j = 0; j < numToSelect; j++) {
269 int index = int(rand() % kmers.size());
272 temp.push_back(kmers[index]);
276 int newTax = getMostProbableTaxonomy(temp);
277 TaxNode taxonomyTemp = phyloTree->get(newTax);
279 //add to confidence results
280 while (taxonomyTemp.level != 0) { //while you are not at the root
282 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
284 if (itBoot2 == confidenceScores.end()) { //not already in confidence scores
285 confidenceScores[newTax] = 1;
287 confidenceScores[newTax]++;
290 newTax = taxonomyTemp.parent;
291 taxonomyTemp = phyloTree->get(newTax);
296 string confidenceTax = "";
299 int seqTaxIndex = tax;
300 TaxNode seqTax = phyloTree->get(tax);
302 while (seqTax.level != 0) { //while you are not at the root
304 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
307 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
308 confidence = confidenceScores[seqTaxIndex];
311 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
312 confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
313 simpleTax = seqTax.name + ";" + simpleTax;
316 seqTaxIndex = seqTax.parent;
317 seqTax = phyloTree->get(seqTax.parent);
320 if (confidenceTax == "") { confidenceTax = "unclassified;"; simpleTax = "unclassified;"; }
321 return confidenceTax;
324 catch(exception& e) {
325 m->errorOut(e, "Bayesian", "bootstrapResults");
329 /**************************************************************************************************/
330 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
332 int indexofGenus = 0;
334 double maxProbability = -1000000.0;
335 //find taxonomy with highest probability that this sequence is from it
336 for (int k = 0; k < genusNodes.size(); k++) {
337 //for each taxonomy calc its probability
339 for (int i = 0; i < queryKmer.size(); i++) {
340 prob += wordGenusProb[queryKmer[i]][k];
343 //is this the taxonomy with the greatest probability?
344 if (prob > maxProbability) {
345 indexofGenus = genusNodes[k];
346 maxProbability = prob;
352 catch(exception& e) {
353 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
357 /*************************************************************************************************
358 map<string, int> Bayesian::parseTaxMap(string newTax) {
361 map<string, int> parsed;
363 newTax = newTax.substr(0, newTax.length()-1); //get rid of last ';'
367 while (newTax.find_first_of(';') != -1) {
368 individual = newTax.substr(0,newTax.find_first_of(';'));
369 newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
370 parsed[individual] = 1;
379 catch(exception& e) {
380 m->errorOut(e, "Bayesian", "parseTax");
384 /**************************************************************************************************/
385 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
390 int pid, num, num2, processors;
391 vector<unsigned long int> positions;
392 vector<unsigned long int> positions2;
397 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
398 MPI_Comm_size(MPI_COMM_WORLD, &processors);
401 char inFileName[1024];
402 strcpy(inFileName, inNumName.c_str());
404 char inFileName2[1024];
405 strcpy(inFileName2, inName.c_str());
407 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
408 MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2); //comm, filename, mode, info, filepointer
411 positions = setFilePosEachLine(inNumName, num);
412 positions2 = setFilePosEachLine(inName, num2);
414 for(int i = 1; i < processors; i++) {
415 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
416 MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
418 MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
419 MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
423 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
424 positions.resize(num+1);
425 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
427 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
428 positions2.resize(num2+1);
429 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
433 int length = positions2[1] - positions2[0];
434 char* buf5 = new char[length];
436 MPI_File_read_at(inMPI2, positions2[0], buf5, length, MPI_CHAR, &status);
440 length = positions2[2] - positions2[1];
441 char* buf = new char[length];
443 MPI_File_read_at(inMPI2, positions2[1], buf, length, MPI_CHAR, &status);
445 string tempBuf = buf;
446 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
449 istringstream iss (tempBuf,istringstream::in);
452 //initialze probabilities
453 wordGenusProb.resize(numKmers);
455 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
458 vector<int> numbers; numbers.resize(numKmers);
460 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
463 length = positions[1] - positions[0];
464 char* buf6 = new char[length];
466 MPI_File_read_at(inMPI2, positions[0], buf6, length, MPI_CHAR, &status);
470 for(int i=1;i<num;i++){
472 length = positions[i+1] - positions[i];
473 char* buf4 = new char[length];
475 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
478 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
481 istringstream iss (tempBuf,istringstream::in);
482 iss >> zeroCountProb[i] >> numbers[i];
485 MPI_File_close(&inMPI);
487 for(int i=2;i<num2;i++){
489 length = positions2[i+1] - positions2[i];
490 char* buf4 = new char[length];
492 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
495 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
498 istringstream iss (tempBuf,istringstream::in);
502 //set them all to zero value
503 for (int i = 0; i < genusNodes.size(); i++) {
504 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
507 //get probs for nonzero values
508 for (int i = 0; i < numbers[kmer]; i++) {
510 wordGenusProb[kmer][name] = prob;
514 MPI_File_close(&inMPI2);
515 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
518 string line = getline(in); gobble(in);
520 in >> numKmers; gobble(in);
522 //initialze probabilities
523 wordGenusProb.resize(numKmers);
525 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
527 int kmer, name, count; count = 0;
528 vector<int> num; num.resize(numKmers);
530 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
533 string line2 = getline(inNum); gobble(inNum);
536 inNum >> zeroCountProb[count] >> num[count];
545 //set them all to zero value
546 for (int i = 0; i < genusNodes.size(); i++) {
547 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
550 //get probs for nonzero values
551 for (int i = 0; i < num[kmer]; i++) {
553 wordGenusProb[kmer][name] = prob;
562 catch(exception& e) {
563 m->errorOut(e, "Bayesian", "readProbFile");
567 /**************************************************************************************************/
568 bool Bayesian::checkReleaseDate(ifstream& file1, ifstream& file2, ifstream& file3, ifstream& file4) {
573 vector<string> lines;
574 lines.push_back(getline(file1));
575 lines.push_back(getline(file2));
576 lines.push_back(getline(file3));
577 lines.push_back(getline(file4));
579 //before we added this check
580 if ((lines[0][0] != '#') || (lines[1][0] != '#') || (lines[2][0] != '#') || (lines[3][0] != '#')) { good = false; }
583 for (int i = 0; i < lines.size(); i++) { lines[i] = lines[i].substr(1); }
585 //get mothurs current version
586 string version = m->getVersion();
588 vector<string> versionVector;
589 splitAtChar(version, versionVector, '.');
591 //check each files version
592 for (int i = 0; i < lines.size(); i++) {
593 vector<string> linesVector;
594 splitAtChar(lines[i], linesVector, '.');
596 if (versionVector.size() != linesVector.size()) { good = false; break; }
598 for (int j = 0; j < versionVector.size(); j++) {
600 convert(versionVector[j], num1);
601 convert(linesVector[j], num2);
603 //if mothurs version is newer than this files version, then we want to remake it
604 if (num1 > num2) { good = false; break; }
608 if (!good) { break; }
612 if (!good) { file1.close(); file2.close(); file3.close(); file4.close(); }
613 else { file1.seekg(0); file2.seekg(0); file3.seekg(0); file4.seekg(0); }
617 catch(exception& e) {
618 m->errorOut(e, "Bayesian", "checkReleaseDate");
622 /**************************************************************************************************/