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