classifyseqscommand.cpp

   1 /*
   2  *  classifyseqscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 11/2/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "classifyseqscommand.h"
  11 #include "sequence.hpp"
  12 #include "bayesian.h"
  13 #include "phylotree.h"
  14 #include "knn.h"
  15
  16 //**********************************************************************************************************************
  17
  18 ClassifySeqsCommand::ClassifySeqsCommand(string option){
  19         try {
  20                 abort = false;
  21
  22                 //allow user to run help
  23                 if(option == "help") { help(); abort = true; }
  24
  25                 else {
  26
  27                         //valid paramters for this command
  28                         string AlignArray[] =  {"template","fasta","search","ksize","method","processors","taxonomy","match","mismatch","gapopen","gapextend","numwanted","cutoff","probs"};
  29                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  30
  31                         OptionParser parser(option);
  32                         map<string, string> parameters = parser.getParameters();
  33
  34                         ValidParameters validParameter;
  35
  36                         //check to make sure all parameters are valid for command
  37                         for (map<string, string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  38                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  39                         }
  40
  41                         //check for required parameters
  42                         templateFileName = validParameter.validFile(parameters, "template", true);
  43                         if (templateFileName == "not found") {
  44                                 mothurOut("template is a required parameter for the classify.seqs command.");
  45                                 mothurOutEndLine();
  46                                 abort = true;
  47                         }
  48                         else if (templateFileName == "not open") { abort = true; }
  49
  50                         fastaFileName = validParameter.validFile(parameters, "fasta", true);
  51                         if (fastaFileName == "not found") {
  52                                 mothurOut("fasta is a required parameter for the classify.seqs command.");
  53                                 mothurOutEndLine();
  54                                 abort = true;
  55                         }
  56                         else if (fastaFileName == "not open") { abort = true; }
  57
  58                         taxonomyFileName = validParameter.validFile(parameters, "taxonomy", true);
  59                         if (taxonomyFileName == "not found") {
  60                                 mothurOut("taxonomy is a required parameter for the classify.seqs command.");
  61                                 mothurOutEndLine();
  62                                 abort = true;
  63                         }
  64                         else if (taxonomyFileName == "not open") { abort = true; }
  65
  66
  67                         //check for optional parameter and set defaults
  68                         // ...at some point should added some additional type checking...
  69                         string temp;
  70                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
  71                         convert(temp, kmerSize);
  72
  73                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
  74                         convert(temp, processors);
  75
  76                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
  77
  78                         method = validParameter.validFile(parameters, "method", false);         if (method == "not found"){     method = "bayesian";    }
  79
  80                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
  81                         convert(temp, match);
  82
  83                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
  84                         convert(temp, misMatch);
  85
  86                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
  87                         convert(temp, gapOpen);
  88
  89                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
  90                         convert(temp, gapExtend);
  91
  92                         temp = validParameter.validFile(parameters, "numwanted", false);        if (temp == "not found"){       temp = "10";                    }
  93                         convert(temp, numWanted);
  94
  95                         temp = validParameter.validFile(parameters, "cutoff", false);           if (temp == "not found"){       temp = "0";                             }
  96                         convert(temp, cutoff);
  97
  98                         temp = validParameter.validFile(parameters, "probs", false);            if (temp == "not found"){       temp = "true";                  }
  99                         probs = isTrue(temp);
 100
 101
 102                         if ((method == "bayesian") && (search != "kmer"))  {
 103                                 mothurOut("The bayesian method requires the kmer search." + search + "will be disregarded." ); mothurOutEndLine();
 104                                 search = "kmer";
 105                         }
 106                 }
 107
 108         }
 109         catch(exception& e) {
 110                 errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
 111                 exit(1);
 112         }
 113 }
 114
 115 //**********************************************************************************************************************
 116
 117 ClassifySeqsCommand::~ClassifySeqsCommand(){
 118
 119         if (abort == false) {
 120                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 121         }
 122 }
 123
 124 //**********************************************************************************************************************
 125
 126 void ClassifySeqsCommand::help(){
 127         try {
 128                 mothurOut("The classify.seqs command reads a fasta file containing sequences and creates a .taxonomy file and a .tax.summary file.\n");
 129                 mothurOut("The classify.seqs command parameters are template, fasta, search, ksize, method, taxonomy, processors, match, mismatch, gapopen, gapextend, numwanted and probs.\n");
 130                 mothurOut("The template, fasta and taxonomy parameters are required.\n");
 131                 mothurOut("The search parameter allows you to specify the method to find most similar template.  Your options are: suffix, kmer and blast. The default is kmer.\n");
 132                 mothurOut("The method parameter allows you to specify classification method to use.  Your options are: bayesian and knn. The default is bayesian.\n");
 133                 mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 134                 mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
 135                 mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 136                 mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 137                 mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -1.0.\n");
 138                 mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -2.0.\n");
 139                 mothurOut("The numwanted parameter allows you to specify the number of sequence matches you want with the knn method.  The default is 10.\n");
 140                 mothurOut("The cutoff parameter allows you to specify a bootstrap confidence threshold for your taxonomy.  The default is 0.\n");
 141                 mothurOut("The probs parameter shut off the bootstrapping results for the bayesian method. The default is true, meaning you want the bootstrapping to be run.\n");
 142                 mothurOut("The classify.seqs command should be in the following format: \n");
 143                 mothurOut("classify.seqs(template=yourTemplateFile, fasta=yourFastaFile, method=yourClassificationMethod, search=yourSearchmethod, ksize=yourKmerSize, taxonomy=yourTaxonomyFile, processors=yourProcessors) \n");
 144                 mothurOut("Example classify.seqs(fasta=amazon.fasta, template=core.filtered, method=knn, search=gotoh, ksize=8, processors=2)\n");
 145                 mothurOut("The .taxonomy file consists of 2 columns: 1 = your sequence name, 2 = the taxonomy for your sequence. \n");
 146                 mothurOut("The .tax.summary is a summary of the different taxonomies represented in your fasta file. \n");
 147                 mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFastaFile).\n\n");
 148         }
 149         catch(exception& e) {
 150                 errorOut(e, "ClassifySeqsCommand", "help");
 151                 exit(1);
 152         }
 153 }
 154
 155
 156 //**********************************************************************************************************************
 157
 158 int ClassifySeqsCommand::execute(){
 159         try {
 160                 if (abort == true) {    return 0;       }
 161
 162                 if(method == "bayesian")                        {       classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, probs);           }
 163                 else if(method == "knn")                        {       classify = new Knn(taxonomyFileName, templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, numWanted);                               }
 164                 else {
 165                         mothurOut(search + " is not a valid method option. I will run the command using bayesian.");
 166                         mothurOutEndLine();
 167                         classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, probs);
 168                 }
 169
 170                 int numFastaSeqs = 0;
 171
 172                 string newTaxonomyFile = getRootName(fastaFileName) + "taxonomy";
 173                 string tempTaxonomyFile = getRootName(fastaFileName) + "taxonomy.temp";
 174                 string taxSummary = getRootName(fastaFileName) + "tax.summary";
 175
 176                 int start = time(NULL);
 177 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 178                 if(processors == 1){
 179                         ifstream inFASTA;
 180                         openInputFile(fastaFileName, inFASTA);
 181                         numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 182                         inFASTA.close();
 183
 184                         lines.push_back(new linePair(0, numFastaSeqs));
 185
 186                         driver(lines[0], newTaxonomyFile, tempTaxonomyFile);
 187                 }
 188                 else{
 189                         vector<int> positions;
 190                         processIDS.resize(0);
 191
 192                         ifstream inFASTA;
 193                         openInputFile(fastaFileName, inFASTA);
 194
 195                         string input;
 196                         while(!inFASTA.eof()){
 197                                 input = getline(inFASTA);
 198                                 if (input.length() != 0) {
 199                                         if(input[0] == '>'){    int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1);       }
 200                                 }
 201                         }
 202                         inFASTA.close();
 203
 204                         numFastaSeqs = positions.size();
 205
 206                         int numSeqsPerProcessor = numFastaSeqs / processors;
 207
 208                         for (int i = 0; i < processors; i++) {
 209                                 int startPos = positions[ i * numSeqsPerProcessor ];
 210                                 if(i == processors - 1){
 211                                         numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
 212                                 }
 213                                 lines.push_back(new linePair(startPos, numSeqsPerProcessor));
 214                         }
 215                         createProcesses(newTaxonomyFile, tempTaxonomyFile);
 216
 217                         rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 218                         rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 219
 220                         for(int i=1;i<processors;i++){
 221                                 appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 222                                 appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 223                                 remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 224                                 remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 225                         }
 226
 227                 }
 228 #else
 229                 ifstream inFASTA;
 230                 openInputFile(fastaFileName, inFASTA);
 231                 numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 232                 inFASTA.close();
 233
 234                 lines.push_back(new linePair(0, numFastaSeqs));
 235
 236                 driver(lines[0], newTaxonomyFile, tempTaxonomyFile);
 237 #endif
 238                 delete classify;
 239
 240                 //make taxonomy tree from new taxonomy file
 241                 ifstream inTaxonomy;
 242                 openInputFile(tempTaxonomyFile, inTaxonomy);
 243
 244                 string accession, taxaList;
 245                 PhyloTree taxaBrowser;
 246
 247                 //read in users taxonomy file and add sequences to tree
 248                 while(!inTaxonomy.eof()){
 249                         inTaxonomy >> accession >> taxaList;
 250
 251                         taxaBrowser.addSeqToTree(accession, taxaList);
 252
 253                         gobble(inTaxonomy);
 254                 }
 255                 inTaxonomy.close();
 256                 remove(tempTaxonomyFile.c_str());
 257
 258                 taxaBrowser.assignHeirarchyIDs(0);
 259
 260                 ofstream outTaxTree;
 261                 openOutputFile(taxSummary, outTaxTree);
 262
 263                 taxaBrowser.print(outTaxTree);
 264
 265                 mothurOutEndLine();
 266                 mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences.");
 267                 mothurOutEndLine();
 268                 mothurOutEndLine();
 269
 270                 return 0;
 271         }
 272         catch(exception& e) {
 273                 errorOut(e, "ClassifySeqsCommand", "execute");
 274                 exit(1);
 275         }
 276 }
 277 /**************************************************************************************************/
 278
 279 void ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile) {
 280         try {
 281 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 282                 int process = 0;
 283                 //              processIDS.resize(0);
 284
 285                 //loop through and create all the processes you want
 286                 while (process != processors) {
 287                         int pid = fork();
 288
 289                         if (pid > 0) {
 290                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 291                                 process++;
 292                         }else if (pid == 0){
 293                                 driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp");
 294                                 exit(0);
 295                         }else { mothurOut("unable to spawn the necessary processes."); mothurOutEndLine(); exit(0); }
 296                 }
 297
 298                 //force parent to wait until all the processes are done
 299                 for (int i=0;i<processors;i++) {
 300                         int temp = processIDS[i];
 301                         wait(&temp);
 302                 }
 303 #endif
 304         }
 305         catch(exception& e) {
 306                 errorOut(e, "ClassifySeqsCommand", "createProcesses");
 307                 exit(1);
 308         }
 309 }
 310 /**************************************************************************************************/
 311
 312 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 313         try{
 314
 315                 ofstream output;
 316                 ifstream input;
 317                 openOutputFileAppend(filename, output);
 318                 openInputFile(temp, input);
 319
 320                 while(char c = input.get()){
 321                         if(input.eof())         {       break;                  }
 322                         else                            {       output << c;    }
 323                 }
 324
 325                 input.close();
 326                 output.close();
 327         }
 328         catch(exception& e) {
 329                 errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 330                 exit(1);
 331         }
 332 }
 333
 334 //**********************************************************************************************************************
 335
 336 int ClassifySeqsCommand::driver(linePair* line, string taxFName, string tempTFName){
 337         try {
 338                 ofstream outTax;
 339                 openOutputFile(taxFName, outTax);
 340
 341                 ofstream outTaxSimple;
 342                 openOutputFile(tempTFName, outTaxSimple);
 343
 344                 ifstream inFASTA;
 345                 openInputFile(fastaFileName, inFASTA);
 346
 347                 inFASTA.seekg(line->start);
 348
 349                 string taxonomy;
 350
 351                 for(int i=0;i<line->numSeqs;i++){
 352
 353                         Sequence* candidateSeq = new Sequence(inFASTA);
 354
 355                         if (candidateSeq->getName() != "") {
 356                                 taxonomy = classify->getTaxonomy(candidateSeq);
 357
 358                                 if (taxonomy != "bad seq") {
 359                                         outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 360                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 361                                 }
 362                         }
 363                         delete candidateSeq;
 364
 365                         if((i+1) % 100 == 0){
 366                                 mothurOut("Classifying sequence " + toString(i+1)); mothurOutEndLine();
 367                         }
 368                 }
 369
 370                 inFASTA.close();
 371                 outTax.close();
 372                 outTaxSimple.close();
 373
 374                 return 1;
 375         }
 376         catch(exception& e) {
 377                 errorOut(e, "ClassifySeqsCommand", "driver");
 378                 exit(1);
 379         }
 380 }
 381
 382 /**************************************************************************************************/