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 #include "phylosummary.h"
  13 #include "referencedb.h"
  14 /**************************************************************************************************/
  15 Bayesian::Bayesian(string tfile, string tempFile, string method, int ksize, int cutoff, int i, int tid, bool f, bool sh) :
  16 Classify(), kmerSize(ksize), confidenceThreshold(cutoff), iters(i) {
  17         try {
  18                 ReferenceDB* rdb = ReferenceDB::getInstance();
  19
  20                 threadID = tid;
  21                 flip = f;
  22         shortcuts = sh;
  23                 string baseName = tempFile;
  24
  25                 if (baseName == "saved") { baseName = rdb->getSavedReference(); }
  26
  27                 string baseTName = tfile;
  28                 if (baseTName == "saved") { baseTName = rdb->getSavedTaxonomy(); }
  29
  30                 /************calculate the probablity that each word will be in a specific taxonomy*************/
  31                 string tfileroot = m->getFullPathName(baseTName.substr(0,baseTName.find_last_of(".")+1));
  32                 string tempfileroot = m->getRootName(m->getSimpleName(baseName));
  33                 string phyloTreeName = tfileroot + "tree.train";
  34                 string phyloTreeSumName = tfileroot + "tree.sum";
  35                 string probFileName = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.prob";
  36                 string probFileName2 = tfileroot + tempfileroot + char('0'+ kmerSize) + "mer.numNonZero";
  37
  38                 ofstream out;
  39                 ofstream out2;
  40
  41                 ifstream phyloTreeTest(phyloTreeName.c_str());
  42                 ifstream probFileTest2(probFileName2.c_str());
  43                 ifstream probFileTest(probFileName.c_str());
  44                 ifstream probFileTest3(phyloTreeSumName.c_str());
  45
  46                 int start = time(NULL);
  47
  48                 //if they are there make sure they were created after this release date
  49                 bool FilesGood = false;
  50                 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3){
  51                         FilesGood = checkReleaseDate(probFileTest, probFileTest2, phyloTreeTest, probFileTest3);
  52                 }
  53
  54                 //if you want to save, but you dont need to calculate then just read
  55                 if (rdb->save && probFileTest && probFileTest2 && phyloTreeTest && probFileTest3 && FilesGood && (tempFile != "saved")) {
  56                         ifstream saveIn;
  57                         m->openInputFile(tempFile, saveIn);
  58
  59                         while (!saveIn.eof()) {
  60                                 Sequence temp(saveIn);
  61                                 m->gobble(saveIn);
  62
  63                                 rdb->referenceSeqs.push_back(temp);
  64                         }
  65                         saveIn.close();
  66                 }
  67
  68                 if(probFileTest && probFileTest2 && phyloTreeTest && probFileTest3 && FilesGood){
  69                         if (tempFile == "saved") { m->mothurOutEndLine();  m->mothurOut("Using sequences from " + rdb->getSavedReference() + " that are saved in memory.");     m->mothurOutEndLine(); }
  70
  71                         m->mothurOut("Reading template taxonomy...     "); cout.flush();
  72
  73                         phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
  74
  75                         m->mothurOut("DONE."); m->mothurOutEndLine();
  76
  77                         genusNodes = phyloTree->getGenusNodes();
  78                         genusTotals = phyloTree->getGenusTotals();
  79
  80                         if (tfile == "saved") {
  81                                 m->mothurOutEndLine();  m->mothurOut("Using probabilties from " + rdb->getSavedTaxonomy() + " that are saved in memory...    ");        cout.flush();;
  82                                 wordGenusProb = rdb->wordGenusProb;
  83                                 WordPairDiffArr = rdb->WordPairDiffArr;
  84                         }else {
  85                                 m->mothurOut("Reading template probabilities...     "); cout.flush();
  86                                 readProbFile(probFileTest, probFileTest2, probFileName, probFileName2);
  87                         }
  88
  89                         //save probabilities
  90                         if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
  91                 }else{
  92
  93                         //create search database and names vector
  94                         generateDatabaseAndNames(tfile, tempFile, method, ksize, 0.0, 0.0, 0.0, 0.0);
  95
  96                         //prevents errors caused by creating shortcut files if you had an error in the sanity check.
  97                         if (m->control_pressed) {  m->mothurRemove(phyloTreeName);  m->mothurRemove(probFileName); m->mothurRemove(probFileName2); }
  98                         else{
  99                                 genusNodes = phyloTree->getGenusNodes();
 100                                 genusTotals = phyloTree->getGenusTotals();
 101
 102                                 m->mothurOut("Calculating template taxonomy tree...     "); cout.flush();
 103
 104                                 phyloTree->printTreeNodes(phyloTreeName);
 105
 106                                 m->mothurOut("DONE."); m->mothurOutEndLine();
 107
 108                                 m->mothurOut("Calculating template probabilities...     "); cout.flush();
 109
 110                                 numKmers = database->getMaxKmer() + 1;
 111
 112                                 //initialze probabilities
 113                                 wordGenusProb.resize(numKmers);
 114                                 WordPairDiffArr.resize(numKmers);
 115
 116                                 for (int j = 0; j < wordGenusProb.size(); j++) {        wordGenusProb[j].resize(genusNodes.size());             }
 117                 ofstream out;
 118                                 ofstream out2;
 119
 120                                 #ifdef USE_MPI
 121                                         int pid;
 122                                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 123
 124                                         if (pid == 0) {
 125                                 #endif
 126
 127
 128                 if (shortcuts) {
 129                     m->openOutputFile(probFileName, out);
 130
 131                     //output mothur version
 132                     out << "#" << m->getVersion() << endl;
 133
 134                     out << numKmers << endl;
 135
 136                     m->openOutputFile(probFileName2, out2);
 137
 138                     //output mothur version
 139                     out2 << "#" << m->getVersion() << endl;
 140                 }
 141
 142                                 #ifdef USE_MPI
 143                                         }
 144                                 #endif
 145
 146                                 //for each word
 147                                 for (int i = 0; i < numKmers; i++) {
 148                                         if (m->control_pressed) {  break; }
 149
 150                                         #ifdef USE_MPI
 151                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 152
 153                                                 if (pid == 0) {
 154                                         #endif
 155
 156                     if (shortcuts) {  out << i << '\t'; }
 157
 158                                         #ifdef USE_MPI
 159                                                 }
 160                                         #endif
 161
 162                                         vector<int> seqsWithWordi = database->getSequencesWithKmer(i);
 163
 164                                         //for each sequence with that word
 165                     vector<int> count; count.resize(genusNodes.size(), 0);
 166                                         for (int j = 0; j < seqsWithWordi.size(); j++) {
 167                                                 int temp = phyloTree->getGenusIndex(names[seqsWithWordi[j]]);
 168                                                 count[temp]++;  //increment count of seq in this genus who have this word
 169                                         }
 170
 171                                         //probabilityInTemplate = (# of seqs with that word in template + 0.50) / (total number of seqs in template + 1);
 172                                         float probabilityInTemplate = (seqsWithWordi.size() + 0.50) / (float) (names.size() + 1);
 173                                         diffPair tempProb(log(probabilityInTemplate), 0.0);
 174                                         WordPairDiffArr[i] = tempProb;
 175
 176                                         int numNotZero = 0;
 177                                         for (int k = 0; k < genusNodes.size(); k++) {
 178                                                 //probabilityInThisTaxonomy = (# of seqs with that word in this taxonomy + probabilityInTemplate) / (total number of seqs in this taxonomy + 1);
 179
 180
 181                                                 wordGenusProb[i][k] = log((count[k] + probabilityInTemplate) / (float) (genusTotals[k] + 1));
 182
 183                                                 if (count[k] != 0) {
 184                                                         #ifdef USE_MPI
 185                                                                 int pid;
 186                                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 187
 188                                                                 if (pid == 0) {
 189                                                         #endif
 190
 191                             if (shortcuts) { out << k << '\t' << wordGenusProb[i][k] << '\t' ; }
 192
 193                                                         #ifdef USE_MPI
 194                                                                 }
 195                                                         #endif
 196
 197                                                         numNotZero++;
 198                                                 }
 199                                         }
 200
 201                                         #ifdef USE_MPI
 202                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 203
 204                                                 if (pid == 0) {
 205                                         #endif
 206
 207                             if (shortcuts) {
 208                                 out << endl;
 209                                 out2 << probabilityInTemplate << '\t' << numNotZero << '\t' << log(probabilityInTemplate) << endl;
 210                             }
 211
 212                                         #ifdef USE_MPI
 213                                                 }
 214                                         #endif
 215                                 }
 216
 217                                 #ifdef USE_MPI
 218                                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 219
 220                                         if (pid == 0) {
 221                                 #endif
 222
 223                         if (shortcuts) {
 224                             out.close();
 225                             out2.close();
 226                         }
 227                                 #ifdef USE_MPI
 228                                         }
 229                                 #endif
 230
 231                                 //read in new phylotree with less info. - its faster
 232                                 ifstream phyloTreeTest(phyloTreeName.c_str());
 233                                 delete phyloTree;
 234
 235                                 phyloTree = new PhyloTree(phyloTreeTest, phyloTreeName);
 236
 237                                 //save probabilities
 238                                 if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
 239                         }
 240                 }
 241
 242                 generateWordPairDiffArr();
 243
 244                 //save probabilities
 245                 if (rdb->save) { rdb->wordGenusProb = wordGenusProb; rdb->WordPairDiffArr = WordPairDiffArr; }
 246
 247                 m->mothurOut("DONE."); m->mothurOutEndLine();
 248                 m->mothurOut("It took " + toString(time(NULL) - start) + " seconds get probabilities. "); m->mothurOutEndLine();
 249         }
 250         catch(exception& e) {
 251                 m->errorOut(e, "Bayesian", "Bayesian");
 252                 exit(1);
 253         }
 254 }
 255 /**************************************************************************************************/
 256 Bayesian::~Bayesian() {
 257         try {
 258         if (phyloTree != NULL) { delete phyloTree; }
 259         if (database != NULL) {  delete database; }
 260         }
 261         catch(exception& e) {
 262                 m->errorOut(e, "Bayesian", "~Bayesian");
 263                 exit(1);
 264         }
 265 }
 266
 267 /**************************************************************************************************/
 268 string Bayesian::getTaxonomy(Sequence* seq) {
 269         try {
 270                 string tax = "";
 271                 Kmer kmer(kmerSize);
 272                 flipped = false;
 273
 274                 //get words contained in query
 275                 //getKmerString returns a string where the index in the string is hte kmer number
 276                 //and the character at that index can be converted to be the number of times that kmer was seen
 277                 string queryKmerString = kmer.getKmerString(seq->getUnaligned());
 278
 279                 vector<int> queryKmers;
 280                 for (int i = 0; i < queryKmerString.length()-1; i++) {  // the -1 is to ignore any kmer with an N in it
 281                         if (queryKmerString[i] != '!') { //this kmer is in the query
 282                                 queryKmers.push_back(i);
 283                         }
 284                 }
 285
 286                 //if user wants to test reverse compliment and its reversed use that instead
 287                 if (flip) {
 288                         if (isReversed(queryKmers)) {
 289                                 flipped = true;
 290                                 seq->reverseComplement();
 291                                 queryKmerString = kmer.getKmerString(seq->getUnaligned());
 292                                 queryKmers.clear();
 293                                 for (int i = 0; i < queryKmerString.length()-1; i++) {  // the -1 is to ignore any kmer with an N in it
 294                                         if (queryKmerString[i] != '!') { //this kmer is in the query
 295                                                 queryKmers.push_back(i);
 296                                         }
 297                                 }
 298                         }
 299                 }
 300
 301                 if (queryKmers.size() == 0) {  m->mothurOut(seq->getName() + "is bad."); m->mothurOutEndLine(); simpleTax = "unknown;";  return "unknown;"; }
 302
 303
 304                 int index = getMostProbableTaxonomy(queryKmers);
 305
 306                 if (m->control_pressed) { return tax; }
 307
 308                 //bootstrap - to set confidenceScore
 309                 int numToSelect = queryKmers.size() / 8;
 310
 311                 tax = bootstrapResults(queryKmers, index, numToSelect);
 312
 313                 return tax;
 314         }
 315         catch(exception& e) {
 316                 m->errorOut(e, "Bayesian", "getTaxonomy");
 317                 exit(1);
 318         }
 319 }
 320 /**************************************************************************************************/
 321 string Bayesian::bootstrapResults(vector<int> kmers, int tax, int numToSelect) {
 322         try {
 323
 324                 map<int, int> confidenceScores;
 325
 326                 //initialize confidences to 0
 327                 int seqIndex = tax;
 328                 TaxNode seq = phyloTree->get(tax);
 329                 confidenceScores[tax] = 0;
 330
 331                 while (seq.level != 0) { //while you are not at the root
 332                         seqIndex = seq.parent;
 333                         confidenceScores[seqIndex] = 0;
 334                         seq = phyloTree->get(seq.parent);
 335                 }
 336
 337                 map<int, int>::iterator itBoot;
 338                 map<int, int>::iterator itBoot2;
 339                 map<int, int>::iterator itConvert;
 340
 341                 for (int i = 0; i < iters; i++) {
 342                         if (m->control_pressed) { return "control"; }
 343
 344                         vector<int> temp;
 345                         for (int j = 0; j < numToSelect; j++) {
 346                                 int index = int(rand() % kmers.size());
 347
 348                                 //add word to temp
 349                                 temp.push_back(kmers[index]);
 350                         }
 351
 352                         //get taxonomy
 353                         int newTax = getMostProbableTaxonomy(temp);
 354                         //int newTax = 1;
 355                         TaxNode taxonomyTemp = phyloTree->get(newTax);
 356
 357                         //add to confidence results
 358                         while (taxonomyTemp.level != 0) { //while you are not at the root
 359                                 itBoot2 = confidenceScores.find(newTax); //is this a classification we already have a count on
 360
 361                                 if (itBoot2 != confidenceScores.end()) { //this is a classification we need a confidence for
 362                                         (itBoot2->second)++;
 363                                 }
 364
 365                                 newTax = taxonomyTemp.parent;
 366                                 taxonomyTemp = phyloTree->get(newTax);
 367                         }
 368
 369                 }
 370
 371                 string confidenceTax = "";
 372                 simpleTax = "";
 373
 374                 int seqTaxIndex = tax;
 375                 TaxNode seqTax = phyloTree->get(tax);
 376
 377                 while (seqTax.level != 0) { //while you are not at the root
 378
 379                                 itBoot2 = confidenceScores.find(seqTaxIndex); //is this a classification we already have a count on
 380
 381                                 int confidence = 0;
 382                                 if (itBoot2 != confidenceScores.end()) { //already in confidence scores
 383                                         confidence = itBoot2->second;
 384                                 }
 385
 386                                 if (((confidence/(float)iters) * 100) >= confidenceThreshold) {
 387                                         confidenceTax = seqTax.name + "(" + toString(((confidence/(float)iters) * 100)) + ");" + confidenceTax;
 388                                         simpleTax = seqTax.name + ";" + simpleTax;
 389                                 }
 390
 391                                 seqTaxIndex = seqTax.parent;
 392                                 seqTax = phyloTree->get(seqTax.parent);
 393                 }
 394
 395                 if (confidenceTax == "") { confidenceTax = "unknown;"; simpleTax = "unknown;";  }
 396
 397                 return confidenceTax;
 398
 399         }
 400         catch(exception& e) {
 401                 m->errorOut(e, "Bayesian", "bootstrapResults");
 402                 exit(1);
 403         }
 404 }
 405 /**************************************************************************************************/
 406 int Bayesian::getMostProbableTaxonomy(vector<int> queryKmer) {
 407         try {
 408                 int indexofGenus = 0;
 409
 410                 double maxProbability = -1000000.0;
 411                 //find taxonomy with highest probability that this sequence is from it
 412
 413
 414 //              cout << genusNodes.size() << endl;
 415
 416
 417                 for (int k = 0; k < genusNodes.size(); k++) {
 418                         //for each taxonomy calc its probability
 419
 420                         double prob = 0.0000;
 421                         for (int i = 0; i < queryKmer.size(); i++) {
 422                                 prob += wordGenusProb[queryKmer[i]][k];
 423                         }
 424
 425 //                      cout << phyloTree->get(genusNodes[k]).name << '\t' << prob << endl;
 426
 427                         //is this the taxonomy with the greatest probability?
 428                         if (prob > maxProbability) {
 429                                 indexofGenus = genusNodes[k];
 430                                 maxProbability = prob;
 431                         }
 432                 }
 433
 434
 435                 return indexofGenus;
 436         }
 437         catch(exception& e) {
 438                 m->errorOut(e, "Bayesian", "getMostProbableTaxonomy");
 439                 exit(1);
 440         }
 441 }
 442 //********************************************************************************************************************
 443 //if it is more probable that the reverse compliment kmers are in the template, then we assume the sequence is reversed.
 444 bool Bayesian::isReversed(vector<int>& queryKmers){
 445         try{
 446                 bool reversed = false;
 447                 float prob = 0;
 448                 float reverseProb = 0;
 449
 450         for (int i = 0; i < queryKmers.size(); i++){
 451             int kmer = queryKmers[i];
 452             if (kmer >= 0){
 453                 prob += WordPairDiffArr[kmer].prob;
 454                                 reverseProb += WordPairDiffArr[kmer].reverseProb;
 455             }
 456         }
 457
 458         if (reverseProb > prob){ reversed = true; }
 459
 460                 return reversed;
 461         }
 462         catch(exception& e) {
 463                 m->errorOut(e, "Bayesian", "isReversed");
 464                 exit(1);
 465         }
 466 }
 467 //********************************************************************************************************************
 468 int Bayesian::generateWordPairDiffArr(){
 469         try{
 470                 Kmer kmer(kmerSize);
 471                 for (int i = 0; i < WordPairDiffArr.size(); i++) {
 472                         int reversedWord = kmer.getReverseKmerNumber(i);
 473                         WordPairDiffArr[i].reverseProb = WordPairDiffArr[reversedWord].prob;
 474                 }
 475
 476                 return 0;
 477         }catch(exception& e) {
 478                 m->errorOut(e, "Bayesian", "generateWordPairDiffArr");
 479                 exit(1);
 480         }
 481 }
 482 /*************************************************************************************************
 483 map<string, int> Bayesian::parseTaxMap(string newTax) {
 484         try{
 485
 486                 map<string, int> parsed;
 487
 488                 newTax = newTax.substr(0, newTax.length()-1);  //get rid of last ';'
 489
 490                 //parse taxonomy
 491                 string individual;
 492                 while (newTax.find_first_of(';') != -1) {
 493                         individual = newTax.substr(0,newTax.find_first_of(';'));
 494                         newTax = newTax.substr(newTax.find_first_of(';')+1, newTax.length());
 495                         parsed[individual] = 1;
 496                 }
 497
 498                 //get last one
 499                 parsed[newTax] = 1;
 500
 501                 return parsed;
 502
 503         }
 504         catch(exception& e) {
 505                 m->errorOut(e, "Bayesian", "parseTax");
 506                 exit(1);
 507         }
 508 }
 509 **************************************************************************************************/
 510 void Bayesian::readProbFile(ifstream& in, ifstream& inNum, string inName, string inNumName) {
 511         try{
 512
 513                 #ifdef USE_MPI
 514
 515                         int pid, num, num2, processors;
 516                         vector<unsigned long long> positions;
 517                         vector<unsigned long long> positions2;
 518
 519                         MPI_Status status;
 520                         MPI_File inMPI;
 521                         MPI_File inMPI2;
 522                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 523                         MPI_Comm_size(MPI_COMM_WORLD, &processors);
 524                         int tag = 2001;
 525
 526                         char inFileName[1024];
 527                         strcpy(inFileName, inNumName.c_str());
 528
 529                         char inFileName2[1024];
 530                         strcpy(inFileName2, inName.c_str());
 531
 532                         MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 533                         MPI_File_open(MPI_COMM_WORLD, inFileName2, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI2);  //comm, filename, mode, info, filepointer
 534
 535                         if (pid == 0) {
 536                                 positions = m->setFilePosEachLine(inNumName, num);
 537                                 positions2 = m->setFilePosEachLine(inName, num2);
 538
 539                                 for(int i = 1; i < processors; i++) {
 540                                         MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 541                                         MPI_Send(&positions[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 542
 543                                         MPI_Send(&num2, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 544                                         MPI_Send(&positions2[0], (num2+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 545                                 }
 546
 547                         }else{
 548                                 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 549                                 positions.resize(num+1);
 550                                 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 551
 552                                 MPI_Recv(&num2, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 553                                 positions2.resize(num2+1);
 554                                 MPI_Recv(&positions2[0], (num2+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 555                         }
 556
 557                         //read version
 558                         int length = positions2[1] - positions2[0];
 559                         char* buf5 = new char[length];
 560
 561                         MPI_File_read_at(inMPI2, positions2[0], buf5, length, MPI_CHAR, &status);
 562                         delete buf5;
 563
 564                         //read numKmers
 565                         length = positions2[2] - positions2[1];
 566                         char* buf = new char[length];
 567
 568                         MPI_File_read_at(inMPI2, positions2[1], buf, length, MPI_CHAR, &status);
 569
 570                         string tempBuf = buf;
 571                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
 572                         delete buf;
 573
 574                         istringstream iss (tempBuf,istringstream::in);
 575                         iss >> numKmers;
 576
 577                         //initialze probabilities
 578                         wordGenusProb.resize(numKmers);
 579
 580                         for (int j = 0; j < wordGenusProb.size(); j++) {        wordGenusProb[j].resize(genusNodes.size());             }
 581
 582                         int kmer, name;
 583                         vector<int> numbers; numbers.resize(numKmers);
 584                         float prob;
 585                         vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
 586                         WordPairDiffArr.resize(numKmers);
 587
 588                         //read version
 589                         length = positions[1] - positions[0];
 590                         char* buf6 = new char[length];
 591
 592                         MPI_File_read_at(inMPI2, positions[0], buf6, length, MPI_CHAR, &status);
 593                         delete buf6;
 594
 595                         //read file
 596                         for(int i=1;i<num;i++){
 597                                 //read next sequence
 598                                 length = positions[i+1] - positions[i];
 599                                 char* buf4 = new char[length];
 600
 601                                 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
 602
 603                                 tempBuf = buf4;
 604                                 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
 605                                 delete buf4;
 606
 607                                 istringstream iss (tempBuf,istringstream::in);
 608                                 float probTemp;
 609                                 iss >> zeroCountProb[i] >> numbers[i] >> probTemp;
 610                                 WordPairDiffArr[i].prob = probTemp;
 611
 612                         }
 613
 614                         MPI_File_close(&inMPI);
 615
 616                         for(int i=2;i<num2;i++){
 617                                 //read next sequence
 618                                 length = positions2[i+1] - positions2[i];
 619                                 char* buf4 = new char[length];
 620
 621                                 MPI_File_read_at(inMPI2, positions2[i], buf4, length, MPI_CHAR, &status);
 622
 623                                 tempBuf = buf4;
 624                                 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
 625                                 delete buf4;
 626
 627                                 istringstream iss (tempBuf,istringstream::in);
 628
 629                                 iss >> kmer;
 630
 631                                 //set them all to zero value
 632                                 for (int i = 0; i < genusNodes.size(); i++) {
 633                                         wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
 634                                 }
 635
 636                                 //get probs for nonzero values
 637                                 for (int i = 0; i < numbers[kmer]; i++) {
 638                                         iss >> name >> prob;
 639                                         wordGenusProb[kmer][name] = prob;
 640                                 }
 641
 642                         }
 643                         MPI_File_close(&inMPI2);
 644                         MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 645                 #else
 646                         //read version
 647                         string line = m->getline(in); m->gobble(in);
 648
 649                         in >> numKmers; m->gobble(in);
 650                         //cout << threadID << '\t' << line << '\t' << numKmers << &in << '\t' << &inNum << '\t' << genusNodes.size() << endl;
 651                         //initialze probabilities
 652                         wordGenusProb.resize(numKmers);
 653
 654                         for (int j = 0; j < wordGenusProb.size(); j++) {        wordGenusProb[j].resize(genusNodes.size());             }
 655
 656                         int kmer, name, count;  count = 0;
 657                         vector<int> num; num.resize(numKmers);
 658                         float prob;
 659                         vector<float> zeroCountProb; zeroCountProb.resize(numKmers);
 660                         WordPairDiffArr.resize(numKmers);
 661
 662                         //read version
 663                         string line2 = m->getline(inNum); m->gobble(inNum);
 664                         float probTemp;
 665                 //cout << threadID << '\t' << line2 << '\t' << this << endl;
 666                         while (inNum) {
 667                                 inNum >> zeroCountProb[count] >> num[count] >> probTemp;
 668                                 WordPairDiffArr[count].prob = probTemp;
 669                                 count++;
 670                                 m->gobble(inNum);
 671                                 //cout << threadID << '\t' << count << endl;
 672                         }
 673                         inNum.close();
 674                 //cout << threadID << '\t' << "here1 " << &wordGenusProb << '\t' << &num << endl; //
 675                 //cout << threadID << '\t' << &genusTotals << '\t' << endl;
 676                 //cout << threadID << '\t' << genusNodes.size() << endl;
 677                         while(in) {
 678                                 in >> kmer;
 679                         //cout << threadID << '\t' << kmer << endl;
 680                                 //set them all to zero value
 681                                 for (int i = 0; i < genusNodes.size(); i++) {
 682                                         wordGenusProb[kmer][i] = log(zeroCountProb[kmer] / (float) (genusTotals[i]+1));
 683                                 }
 684                         //cout << threadID << '\t' << num[kmer] << "here" << endl;
 685                                 //get probs for nonzero values
 686                                 for (int i = 0; i < num[kmer]; i++) {
 687                                         in >> name >> prob;
 688                                         wordGenusProb[kmer][name] = prob;
 689                                 }
 690
 691                                 m->gobble(in);
 692                         }
 693                         in.close();
 694                 //cout << threadID << '\t' << "here" << endl;
 695                 #endif
 696         }
 697         catch(exception& e) {
 698                 m->errorOut(e, "Bayesian", "readProbFile");
 699                 exit(1);
 700         }
 701 }
 702 /**************************************************************************************************/
 703 bool Bayesian::checkReleaseDate(ifstream& file1, ifstream& file2, ifstream& file3, ifstream& file4) {
 704         try {
 705
 706                 bool good = true;
 707
 708                 vector<string> lines;
 709                 lines.push_back(m->getline(file1));
 710                 lines.push_back(m->getline(file2));
 711                 lines.push_back(m->getline(file3));
 712                 lines.push_back(m->getline(file4));
 713
 714                 //before we added this check
 715                 if ((lines[0][0] != '#') || (lines[1][0] != '#') || (lines[2][0] != '#') || (lines[3][0] != '#')) {  good = false;  }
 716                 else {
 717                         //rip off #
 718                         for (int i = 0; i < lines.size(); i++) { lines[i] = lines[i].substr(1);  }
 719
 720                         //get mothurs current version
 721                         string version = m->getVersion();
 722
 723                         vector<string> versionVector;
 724                         m->splitAtChar(version, versionVector, '.');
 725
 726                         //check each files version
 727                         for (int i = 0; i < lines.size(); i++) {
 728                                 vector<string> linesVector;
 729                                 m->splitAtChar(lines[i], linesVector, '.');
 730
 731                                 if (versionVector.size() != linesVector.size()) { good = false; break; }
 732                                 else {
 733                                         for (int j = 0; j < versionVector.size(); j++) {
 734                                                 int num1, num2;
 735                                                 convert(versionVector[j], num1);
 736                                                 convert(linesVector[j], num2);
 737
 738                                                 //if mothurs version is newer than this files version, then we want to remake it
 739                                                 if (num1 > num2) {  good = false; break;  }
 740                                         }
 741                                 }
 742
 743                                 if (!good) { break; }
 744                         }
 745                 }
 746
 747                 if (!good) {  file1.close(); file2.close(); file3.close(); file4.close();  }
 748                 else { file1.seekg(0);  file2.seekg(0);  file3.seekg(0);  file4.seekg(0);  }
 749
 750                 return good;
 751         }
 752         catch(exception& e) {
 753                 m->errorOut(e, "Bayesian", "checkReleaseDate");
 754                 exit(1);
 755         }
 756 }
 757 /**************************************************************************************************/
 758
 759
 760
 761
 762
 763