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 = 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 = 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                 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
 110                 vector<int> queryKmers;
 111                 for (int i = 0; i < queryKmerString.length(); i++) {
 112                         if (queryKmerString[i] != '!') { //this kmer is in the query
 113                                 queryKmers.push_back(i);
 114                         }
 115                 }
 116
 117                 int index = getMostProbableTaxonomy(queryKmers);
 118
 119                 if (m->control_pressed) { return tax; }
 120
 121                 //bootstrap - to set confidenceScore
 122                 int numToSelect = queryKmers.size() / 8;
 123                 tax = bootstrapResults(queryKmers, index, numToSelect);
 124
 125                 return tax;
 126         }
 127         catch(exception& e) {
 128                 m->errorOut(e, "Bayesian", "getTaxonomy");
 129                 exit(1);
 130         }
 131 }
 132 /**************************************************************************************************/
 133 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
 134         try {
 135
 136                 //taxConfidenceScore.clear(); //clear out previous seqs scores
 137
 138                 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
 139                                                                                 //map of classification to confidence for all areas seen
 140                                                                            //ie. Bacteria;Alphaproteobacteria;Rhizobiales;Azorhizobium_et_rel.;Methylobacterium_et_rel.;Bosea;
 141                                                                            //ie. Bacteria -> 100, Alphaproteobacteria -> 100, Rhizobiales -> 87, Azorhizobium_et_rel. -> 78, Methylobacterium_et_rel. -> 70, Bosea -> 50
 142                 confidenceScores.resize(100);  //if you have more than 100 levels of classification...
 143
 144                 map<string, int>::iterator itBoot;
 145                 map<string, int>::iterator itBoot2;
 146                 map<int, int>::iterator itConvert;
 147
 148                 for (int i = 0; i < iters; i++) {
 149                         if (m->control_pressed) { return "control"; }
 150
 151                         vector<int> temp;
 152
 153                         for (int j = 0; j < numToSelect; j++) {
 154                                 int index = int(rand() % kmers.size());
 155
 156                                 //add word to temp
 157                                 temp.push_back(kmers[index]);
 158                         }
 159
 160                         //get taxonomy
 161                         int newTax = getMostProbableTaxonomy(temp);
 162                         TaxNode taxonomy = phyloTree->get(newTax);
 163
 164                         //add to confidence results
 165                         while (taxonomy.level != 0) { //while you are not at the root
 166
 167                                 itBoot2 = confidenceScores[taxonomy.level].find(taxonomy.name); //is this a classification we already have a count on
 168
 169                                 if (itBoot2 == confidenceScores[taxonomy.level].end()) { //not already in confidence scores
 170                                         confidenceScores[taxonomy.level][taxonomy.name] = 1;
 171                                 }else{
 172                                         confidenceScores[taxonomy.level][taxonomy.name]++;
 173                                 }
 174
 175                                 taxonomy = phyloTree->get(taxonomy.parent);
 176                         }
 177                 }
 178
 179                 string confidenceTax = "";
 180                 simpleTax = "";
 181                 TaxNode seqTax = phyloTree->get(tax);
 182
 183                 while (seqTax.level != 0) { //while you are not at the root
 184
 185                                 itBoot2 = confidenceScores[seqTax.level].find(seqTax.name); //is this a classification we already have a count on
 186
 187                                 int confidence = 0;
 188                                 if (itBoot2 != confidenceScores[seqTax.level].end()) { //not already in confidence scores
 189                                         confidence = confidenceScores[seqTax.level][seqTax.name];
 190                                 }
 191
 192                                 if (confidence >= confidenceThreshold) {
 193                                         confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
 194                                         simpleTax = seqTax.name + ";" + simpleTax;
 195                                 }
 196
 197                                 seqTax = phyloTree->get(seqTax.parent);
 198                 }
 199
 200                 return confidenceTax;
 201
 202         }
 203         catch(exception& e) {
 204                 m->errorOut(e, "Bayesian", "bootstrapResults");
 205                 exit(1);
 206         }
 207 }
 208 /**************************************************************************************************/
 209 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
 210         try {
 211                 int indexofGenus;
 212
 213                 double maxProbability = -1000000.0;
 214                 //find taxonomy with highest probability that this sequence is from it
 215                 for (int k = 0; k < genusNodes.size(); k++) {
 216
 217                         //for each taxonomy calc its probability
 218                         double prob = 1.0;
 219                         for (int i = 0; i < queryKmer.size(); i++) {
 220                                 prob += wordGenusProb[queryKmer[i]][k];
 221                         }
 222
 223                         //is this the taxonomy with the greatest probability?
 224                         if (prob > maxProbability) {
 225                                 indexofGenus = genusNodes[k];
 226                                 maxProbability = prob;
 227                         }
 228                 }
 229
 230                 return indexofGenus;
 231         }
 232         catch(exception& e) {
 233                 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
 234                 exit(1);
 235         }
 236 }
 237 /*************************************************************************************************
 238 map<string, int> Bayesian::parseTaxMap(string newTax) {
 239         try{
 240
 241                 map<string, int> parsed;
 242
 243                 newTax = newTax.substr(0, newTax.length()-1);  //get rid of last ';'
 244
 245                 //parse taxonomy
 246                 string individual;
 247                 while (newTax.find_first_of(';') != -1) {
 248                         individual = newTax.substr(0,newTax.find_first_of(';'));
 249                         newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
 250                         parsed[individual] = 1;
 251                 }
 252
 253                 //get last one
 254                 parsed[newTax] = 1;
 255
 256                 return parsed;
 257
 258         }
 259         catch(exception& e) {
 260                 m->errorOut(e, "Bayesian", "parseTax");
 261                 exit(1);
 262         }
 263 }
 264 /**************************************************************************************************/
 265 void Bayesian::readProbFile(ifstream& in, ifstream& inNum) {
 266         try{
 267
 268                 int kmer, name, count;  count = 0;
 269                 vector<int> num; num.resize(numKmers);
 270                 float prob;
 271                 vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
 272
 273                 while (inNum) {
 274                         inNum >> zeroCountProb[count] >> num[count];
 275                         count++;
 276                         gobble(inNum);
 277                 }
 278                 inNum.close();
 279
 280                 while(in) {
 281                         in >> kmer;
 282
 283                         //set them all to zero value
 284                         for (int i = 0; i < genusNodes.size(); i++) {
 285                                 wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
 286                         }
 287
 288                         //get probs for nonzero values
 289                         for (int i = 0; i < num[kmer]; i++) {
 290                                 in >> name >> prob;
 291                                 wordGenusProb[kmer][name] = prob;
 292                         }
 293
 294                         gobble(in);
 295                 }
 296                 in.close();
 297         }
 298         catch(exception& e) {
 299                 m->errorOut(e, "Bayesian", "readProbFile");
 300                 exit(1);
 301         }
 302 }
 303 /**************************************************************************************************/
 304
 305
 306
 307
 308
 309