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 phyloTreeName = tfile.substr(0,tfile.find_last_of(".")+1) + "tree.train";
21 string probFileName = tfile.substr(0,tfile.find_last_of(".")+1) + tempFile.substr(0,tempFile.find_last_of(".")+1) + char('0'+ kmerSize) + "mer.prob";
22 string probFileName2 = tfile.substr(0,tfile.find_last_of(".")+1) + tempFile.substr(0,tempFile.find_last_of(".")+1) + char('0'+ kmerSize) + "mer.numNonZero";
27 ifstream phyloTreeTest(phyloTreeName.c_str());
28 ifstream probFileTest2(probFileName2.c_str());
29 ifstream probFileTest(probFileName.c_str());
31 int start = time(NULL);
33 if(probFileTest && probFileTest2 && phyloTreeTest){
34 m->mothurOut("Reading template taxonomy... "); cout.flush();
36 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
38 m->mothurOut("DONE."); m->mothurOutEndLine();
40 genusNodes = phyloTree->getGenusNodes();
41 genusTotals = phyloTree->getGenusTotals();
43 m->mothurOut("Reading template probabilities... "); cout.flush();
44 readProbFile(probFileTest, probFileTest2, probFileName, probFileName2);
48 //create search database and names vector
49 generateDatabaseAndNames(tfile, tempFile, method, ksize, 0.0, 0.0, 0.0, 0.0);
51 genusNodes = phyloTree->getGenusNodes();
52 genusTotals = phyloTree->getGenusTotals();
54 m->mothurOut("Calculating template taxonomy tree... "); cout.flush();
56 phyloTree->printTreeNodes(phyloTreeName);
58 m->mothurOut("DONE."); m->mothurOutEndLine();
60 m->mothurOut("Calculating template probabilities... "); cout.flush();
62 numKmers = database->getMaxKmer() + 1;
64 //initialze probabilities
65 wordGenusProb.resize(numKmers);
67 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
72 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
78 openOutputFile(probFileName, out);
80 out << numKmers << endl;
83 openOutputFile(probFileName2, out2);
91 for (int i = 0; i < numKmers; i++) {
92 if (m->control_pressed) { break; }
95 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
106 vector<int> seqsWithWordi = database->getSequencesWithKmer(i);
109 for (int k = 0; k < genusNodes.size(); k++) { count[genusNodes[k]] = 0; }
111 //for each sequence with that word
112 for (int j = 0; j < seqsWithWordi.size(); j++) {
113 int temp = phyloTree->getIndex(names[seqsWithWordi[j]]);
114 count[temp]++; //increment count of seq in this genus who have this word
117 //probabilityInTemplate = (# of seqs with that word in template + 0.05) / (total number of seqs in template + 1);
118 float probabilityInTemplate = (seqsWithWordi.size() + 0.50) / (float) (names.size() + 1);
121 for (int k = 0; k < genusNodes.size(); k++) {
122 //probabilityInThisTaxonomy = (# of seqs with that word in this taxonomy + probabilityInTemplate) / (total number of seqs in this taxonomy + 1);
123 wordGenusProb[i][k] = log((count[genusNodes[k]] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
124 if (count[genusNodes[k]] != 0) {
126 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
130 out << k << '\t' << wordGenusProb[i][k] << '\t';
141 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
146 out2 << probabilityInTemplate << '\t' << numNotZero << endl;
154 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
165 //read in new phylotree with less info. - its faster
166 ifstream phyloTreeTest(phyloTreeName.c_str());
169 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
172 m->mothurOut("DONE."); m->mothurOutEndLine();
173 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
175 catch(exception& e) {
176 m->errorOut(e, "Bayesian", "Bayesian");
180 /**************************************************************************************************/
181 Bayesian::~Bayesian() {
184 if (database != NULL) { delete database; }
186 catch(exception& e) {
187 m->errorOut(e, "Bayesian", "~Bayesian");
192 /**************************************************************************************************/
193 string Bayesian::getTaxonomy(Sequence* seq) {
198 //get words contained in query
199 //getKmerString returns a string where the index in the string is hte kmer number
200 //and the character at that index can be converted to be the number of times that kmer was seen
202 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
204 vector<int> queryKmers;
205 for (int i = 0; i < queryKmerString.length(); i++) {
206 if (queryKmerString[i] != '!') { //this kmer is in the query
207 queryKmers.push_back(i);
211 if (queryKmers.size() == 0) { m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); return "bad seq"; }
213 int index = getMostProbableTaxonomy(queryKmers);
215 if (m->control_pressed) { return tax; }
217 //bootstrap - to set confidenceScore
218 int numToSelect = queryKmers.size() / 8;
219 tax = bootstrapResults(queryKmers, index, numToSelect);
223 catch(exception& e) {
224 m->errorOut(e, "Bayesian", "getTaxonomy");
228 /**************************************************************************************************/
229 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
232 map<int, int> confidenceScores;
234 map<int, int>::iterator itBoot;
235 map<int, int>::iterator itBoot2;
236 map<int, int>::iterator itConvert;
238 for (int i = 0; i < iters; i++) {
239 if (m->control_pressed) { return "control"; }
243 for (int j = 0; j < numToSelect; j++) {
244 int index = int(rand() % kmers.size());
247 temp.push_back(kmers[index]);
251 int newTax = getMostProbableTaxonomy(temp);
252 TaxNode taxonomy = phyloTree->get(newTax);
254 //add to confidence results
255 while (taxonomy.level != 0) { //while you are not at the root
257 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
259 if (itBoot2 == confidenceScores.end()) { //not already in confidence scores
260 confidenceScores[newTax] = 1;
262 confidenceScores[newTax]++;
265 newTax = taxonomy.parent;
266 taxonomy = phyloTree->get(taxonomy.parent);
271 string confidenceTax = "";
274 int seqTaxIndex = tax;
275 TaxNode seqTax = phyloTree->get(tax);
277 while (seqTax.level != 0) { //while you are not at the root
279 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
282 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
283 confidence = confidenceScores[seqTaxIndex];
286 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
287 confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
288 simpleTax = seqTax.name + ";" + simpleTax;
291 seqTaxIndex = seqTax.parent;
292 seqTax = phyloTree->get(seqTax.parent);
295 if (confidenceTax == "") { confidenceTax = "unclassified;"; simpleTax = "unclassified;"; }
296 return confidenceTax;
299 catch(exception& e) {
300 m->errorOut(e, "Bayesian", "bootstrapResults");
304 /**************************************************************************************************/
305 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
307 int indexofGenus = 0;
309 double maxProbability = -1000000.0;
310 //find taxonomy with highest probability that this sequence is from it
311 for (int k = 0; k < genusNodes.size(); k++) {
312 //for each taxonomy calc its probability
314 for (int i = 0; i < queryKmer.size(); i++) {
315 prob += wordGenusProb[queryKmer[i]][k];
318 //is this the taxonomy with the greatest probability?
319 if (prob > maxProbability) {
320 indexofGenus = genusNodes[k];
321 maxProbability = prob;
327 catch(exception& e) {
328 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
332 /*************************************************************************************************
333 map<string, int> Bayesian::parseTaxMap(string newTax) {
336 map<string, int> parsed;
338 newTax = newTax.substr(0, newTax.length()-1); //get rid of last ';'
342 while (newTax.find_first_of(';') != -1) {
343 individual = newTax.substr(0,newTax.find_first_of(';'));
344 newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
345 parsed[individual] = 1;
354 catch(exception& e) {
355 m->errorOut(e, "Bayesian", "parseTax");
359 /**************************************************************************************************/
360 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
365 int pid, num, num2, processors;
366 vector<long> positions;
367 vector<long> positions2;
372 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
373 MPI_Comm_size(MPI_COMM_WORLD, &processors);
376 char inFileName[1024];
377 strcpy(inFileName, inNumName.c_str());
379 char inFileName2[1024];
380 strcpy(inFileName2, inName.c_str());
382 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
383 MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2); //comm, filename, mode, info, filepointer
386 positions = setFilePosEachLine(inNumName, num);
387 positions2 = setFilePosEachLine(inName, num2);
389 for(int i = 1; i < processors; i++) {
390 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
391 MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
393 MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
394 MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
398 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
399 positions.resize(num+1);
400 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
402 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
403 positions2.resize(num2+1);
404 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
408 int length = positions2[1] - positions2[0];
409 char* buf = new char[length];
411 MPI_File_read_at(inMPI2, positions2[0], buf, length, MPI_CHAR, &status);
413 string tempBuf = buf;
414 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
417 istringstream iss (tempBuf,istringstream::in);
420 //initialze probabilities
421 wordGenusProb.resize(numKmers);
423 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
426 vector<int> numbers; numbers.resize(numKmers);
428 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
431 for(int i=0;i<num;i++){
433 length = positions[i+1] - positions[i];
434 char* buf4 = new char[length];
436 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
439 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
442 istringstream iss (tempBuf,istringstream::in);
443 iss >> zeroCountProb[i] >> numbers[i];
446 MPI_File_close(&inMPI);
448 for(int i=1;i<num2;i++){
450 length = positions2[i+1] - positions2[i];
451 char* buf4 = new char[length];
453 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
456 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
459 istringstream iss (tempBuf,istringstream::in);
463 //set them all to zero value
464 for (int i = 0; i < genusNodes.size(); i++) {
465 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
468 //get probs for nonzero values
469 for (int i = 0; i < numbers[kmer]; i++) {
471 wordGenusProb[kmer][name] = prob;
475 MPI_File_close(&inMPI2);
476 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
479 in >> numKmers; gobble(in);
481 //initialze probabilities
482 wordGenusProb.resize(numKmers);
484 for (int j = 0; j < wordGenusProb.size(); j++) { wordGenusProb[j].resize(genusNodes.size()); }
486 int kmer, name, count; count = 0;
487 vector<int> num; num.resize(numKmers);
489 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
492 inNum >> zeroCountProb[count] >> num[count];
501 //set them all to zero value
502 for (int i = 0; i < genusNodes.size(); i++) {
503 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
506 //get probs for nonzero values
507 for (int i = 0; i < num[kmer]; i++) {
509 wordGenusProb[kmer][name] = prob;
518 catch(exception& e) {
519 m->errorOut(e, "Bayesian", "readProbFile");
523 /**************************************************************************************************/