bayesian.cpp

   1 /*
   2  *  bayesian.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 11/3/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "bayesian.h"
  11 #include "kmer.hpp"
  12
  13 /**************************************************************************************************/
  14 Bayesian::Bayesian(string tfile, string tempFile, string method, int ksize, int cutoff, int i) :
  15 Classify(tfile, tempFile, method, ksize, 0.0, 0.0, 0.0, 0.0), kmerSize(ksize), confidenceThreshold(cutoff), iters(i)  {
  16         try {
  17
  18                 numKmers = database->getMaxKmer() + 1;
  19
  20                 //initialze probabilities
  21                 wordGenusProb.resize(numKmers);
  22
  23                 genusNodes = phyloTree->getGenusNodes();
  24
  25                 for (int j = 0; j < wordGenusProb.size(); j++) {        wordGenusProb[j].resize(genusNodes.size());             }
  26
  27                 //reset counts because we are on a new word
  28                 for (int j = 0; j < genusNodes.size(); j++) {
  29                         TaxNode temp = phyloTree->get(genusNodes[j]);
  30                         genusTotals.push_back(temp.accessions.size());
  31                 }
  32
  33
  34                 /************calculate the probablity that each word will be in a specific taxonomy*************/
  35                 ofstream out;
  36                 string probFileName = tfile.substr(0,tfile.find_last_of(".")+1) + tempFile.substr(0,tempFile.find_last_of(".")+1) + char('0'+ kmerSize) + "mer.prob";
  37                 ifstream probFileTest(probFileName.c_str());
  38
  39                 ofstream out2;
  40                 string probFileName2 = tfile.substr(0,tfile.find_last_of(".")+1) + tempFile.substr(0,tempFile.find_last_of(".")+1) + char('0'+ kmerSize) + "mer.numNonZero";
  41                 ifstream probFileTest2(probFileName2.c_str());
  42
  43                 int start = time(NULL);
  44
  45                 if(probFileTest && probFileTest2){
  46                         m->mothurOut("Reading template probabilities...     "); cout.flush();
  47                         readProbFile(probFileTest, probFileTest2);
  48                 }else{
  49                         m->mothurOut("Calculating template probabilities...     "); cout.flush();
  50
  51                         ofstream out;
  52                         openOutputFile(probFileName, out);
  53
  54                         ofstream out2;
  55                         openOutputFile(probFileName2, out2);
  56
  57                         //for each word
  58                         for (int i = 0; i < numKmers; i++) {
  59                                 if (m->control_pressed) { break; }
  60
  61                                 out << i << '\t';
  62
  63                                 vector<int> seqsWithWordi = database->getSequencesWithKmer(i);
  64
  65                                 map<int, int> count;
  66                                 for (int k = 0; k < genusNodes.size(); k++) {  count[genusNodes[k]] = 0;  }
  67
  68                                 //for each sequence with that word
  69                                 for (int j = 0; j < seqsWithWordi.size(); j++) {
  70                                         int temp = phyloTree->getIndex(names[seqsWithWordi[j]]);
  71                                         count[temp]++;  //increment count of seq in this genus who have this word
  72                                 }
  73
  74                                 //probabilityInTemplate = (# of seqs with that word in template + 0.05) / (total number of seqs in template + 1);
  75                                 float probabilityInTemplate = (seqsWithWordi.size() + 0.50) / (float) (names.size() + 1);
  76
  77                                 int numNotZero = 0;
  78                                 for (int k = 0; k < genusNodes.size(); k++) {
  79                                         //probabilityInThisTaxonomy = (# of seqs with that word in this taxonomy + probabilityInTemplate) / (total number of seqs in this taxonomy + 1);
  80                                         wordGenusProb[i][k] = log((count[genusNodes[k]] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
  81                                         if (count[genusNodes[k]] != 0) {  out << k << '\t' << wordGenusProb[i][k] << '\t';  numNotZero++;  }
  82                                 }
  83                                 out << endl;
  84                                 out2 << probabilityInTemplate << '\t' << numNotZero << endl;
  85                         }
  86
  87                         out.close();
  88                         out2.close();
  89                 }
  90
  91
  92                 m->mothurOut("DONE."); m->mothurOutEndLine();
  93                 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
  94         }
  95         catch(exception& e) {
  96                 m->errorOut(e, "Bayesian", "Bayesian");
  97                 exit(1);
  98         }
  99 }
 100 /**************************************************************************************************/
 101 string Bayesian::getTaxonomy(Sequence* seq) {
 102         try {
 103                 string tax = "";
 104                 Kmer kmer(kmerSize);
 105
 106                 //get words contained in query
 107                 //getKmerString returns a string where the index in the string is hte kmer number
 108                 //and the character at that index can be converted to be the number of times that kmer was seen
 109
 110                 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
 111
 112                 vector<int> queryKmers;
 113                 for (int i = 0; i < queryKmerString.length(); i++) {
 114                         if (queryKmerString[i] != '!') { //this kmer is in the query
 115                                 queryKmers.push_back(i);
 116
 117                         }
 118                 }
 119
 120                 if (queryKmers.size() == 0) {  m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); return "bad seq"; }
 121
 122                 int index = getMostProbableTaxonomy(queryKmers);
 123
 124
 125                 if (m->control_pressed) { return tax; }
 126
 127                 //bootstrap - to set confidenceScore
 128                 int numToSelect = queryKmers.size() / 8;
 129                 tax = bootstrapResults(queryKmers, index, numToSelect);
 130
 131                 return tax;
 132         }
 133         catch(exception& e) {
 134                 m->errorOut(e, "Bayesian", "getTaxonomy");
 135                 exit(1);
 136         }
 137 }
 138 /**************************************************************************************************/
 139 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
 140         try {
 141
 142                 //taxConfidenceScore.clear(); //clear out previous seqs scores
 143
 144                 vector< map<string, int> > confidenceScores; //you need the added vector level of confusion to account for the level that name is seen since they can be the same
 145                                                                                 //map of classification to confidence for all areas seen
 146                                                                            //ie. Bacteria;Alphaproteobacteria;Rhizobiales;Azorhizobium_et_rel.;Methylobacterium_et_rel.;Bosea;
 147                                                                            //ie. Bacteria -> 100, Alphaproteobacteria -> 100, Rhizobiales -> 87, Azorhizobium_et_rel. -> 78, Methylobacterium_et_rel. -> 70, Bosea -> 50
 148                 confidenceScores.resize(100);  //if you have more than 100 levels of classification...
 149
 150                 map<string, int>::iterator itBoot;
 151                 map<string, int>::iterator itBoot2;
 152                 map<int, int>::iterator itConvert;
 153
 154                 for (int i = 0; i < iters; i++) {
 155                         if (m->control_pressed) { return "control"; }
 156
 157                         vector<int> temp;
 158
 159                         for (int j = 0; j < numToSelect; j++) {
 160                                 int index = int(rand() % kmers.size());
 161
 162                                 //add word to temp
 163                                 temp.push_back(kmers[index]);
 164                         }
 165
 166                         //get taxonomy
 167                         int newTax = getMostProbableTaxonomy(temp);
 168                         TaxNode taxonomy = phyloTree->get(newTax);
 169
 170                         //add to confidence results
 171                         while (taxonomy.level != 0) { //while you are not at the root
 172
 173                                 itBoot2 = confidenceScores[taxonomy.level].find(taxonomy.name); //is this a classification we already have a count on
 174
 175                                 if (itBoot2 == confidenceScores[taxonomy.level].end()) { //not already in confidence scores
 176                                         confidenceScores[taxonomy.level][taxonomy.name] = 1;
 177                                 }else{
 178                                         confidenceScores[taxonomy.level][taxonomy.name]++;
 179                                 }
 180
 181                                 taxonomy = phyloTree->get(taxonomy.parent);
 182                         }
 183
 184                 }
 185
 186                 string confidenceTax = "";
 187                 simpleTax = "";
 188                 TaxNode seqTax = phyloTree->get(tax);
 189
 190                 while (seqTax.level != 0) { //while you are not at the root
 191
 192                                 itBoot2 = confidenceScores[seqTax.level].find(seqTax.name); //is this a classification we already have a count on
 193
 194                                 int confidence = 0;
 195                                 if (itBoot2 != confidenceScores[seqTax.level].end()) { //not already in confidence scores
 196                                         confidence = confidenceScores[seqTax.level][seqTax.name];
 197                                 }
 198
 199                                 if (confidence >= confidenceThreshold) {
 200                                         confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
 201                                         simpleTax = seqTax.name + ";" + simpleTax;
 202                                 }
 203
 204                                 seqTax = phyloTree->get(seqTax.parent);
 205                 }
 206
 207                 if (confidenceTax == "") { confidenceTax = "unclassified;"; simpleTax = "unclassified;"; }
 208                 return confidenceTax;
 209
 210         }
 211         catch(exception& e) {
 212                 m->errorOut(e, "Bayesian", "bootstrapResults");
 213                 exit(1);
 214         }
 215 }
 216 /**************************************************************************************************/
 217 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
 218         try {
 219                 int indexofGenus = 0;
 220
 221                 double maxProbability = -1000000.0;
 222                 //find taxonomy with highest probability that this sequence is from it
 223                 for (int k = 0; k < genusNodes.size(); k++) {
 224                         //for each taxonomy calc its probability
 225                         double prob = 1.0;
 226                         for (int i = 0; i < queryKmer.size(); i++) {
 227                                 prob += wordGenusProb[queryKmer[i]][k];
 228                         }
 229
 230                         //is this the taxonomy with the greatest probability?
 231                         if (prob > maxProbability) {
 232                                 indexofGenus = genusNodes[k];
 233                                 maxProbability = prob;
 234                         }
 235                 }
 236
 237                 return indexofGenus;
 238         }
 239         catch(exception& e) {
 240                 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
 241                 exit(1);
 242         }
 243 }
 244 /*************************************************************************************************
 245 map<string, int> Bayesian::parseTaxMap(string newTax) {
 246         try{
 247
 248                 map<string, int> parsed;
 249
 250                 newTax = newTax.substr(0, newTax.length()-1);  //get rid of last ';'
 251
 252                 //parse taxonomy
 253                 string individual;
 254                 while (newTax.find_first_of(';') != -1) {
 255                         individual = newTax.substr(0,newTax.find_first_of(';'));
 256                         newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
 257                         parsed[individual] = 1;
 258                 }
 259
 260                 //get last one
 261                 parsed[newTax] = 1;
 262
 263                 return parsed;
 264
 265         }
 266         catch(exception& e) {
 267                 m->errorOut(e, "Bayesian", "parseTax");
 268                 exit(1);
 269         }
 270 }
 271 /**************************************************************************************************/
 272 void Bayesian::readProbFile(ifstream& in, ifstream& inNum) {
 273         try{
 274
 275                 int kmer, name, count;  count = 0;
 276                 vector<int> num; num.resize(numKmers);
 277                 float prob;
 278                 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
 279
 280                 while (inNum) {
 281                         inNum >> zeroCountProb[count] >> num[count];
 282                         count++;
 283                         gobble(inNum);
 284                 }
 285                 inNum.close();
 286
 287                 while(in) {
 288                         in >> kmer;
 289
 290                         //set them all to zero value
 291                         for (int i = 0; i < genusNodes.size(); i++) {
 292                                 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
 293                         }
 294
 295                         //get probs for nonzero values
 296                         for (int i = 0; i < num[kmer]; i++) {
 297                                 in >> name >> prob;
 298                                 wordGenusProb[kmer][name] = prob;
 299                         }
 300
 301                         gobble(in);
 302                 }
 303                 in.close();
 304         }
 305         catch(exception& e) {
 306                 m->errorOut(e, "Bayesian", "readProbFile");
 307                 exit(1);
 308         }
 309 }
 310 /**************************************************************************************************/
 311
 312
 313
 314
 315
 316