trimseqscommand.cpp

   1 /*
   2  *  trimseqscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Pat Schloss on 6/6/09.
   6  *  Copyright 2009 Patrick D. Schloss. All rights reserved.
   7  *
   8  */
   9
  10 #include "trimseqscommand.h"
  11
  12 //***************************************************************************************************************
  13
  14 TrimSeqsCommand::TrimSeqsCommand(string option){
  15         try {
  16
  17                 abort = false;
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string AlignArray[] =  {"fasta", "flip", "oligos", "maxambig", "maxhomop", "minlength", "maxlength", "qfile", "qthreshold", "qaverage", "allfiles"};
  25
  26                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  27
  28                         OptionParser parser(option);
  29                         map<string,string> parameters = parser.getParameters();
  30
  31                         ValidParameters validParameter;
  32
  33                         //check to make sure all parameters are valid for command
  34                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  35                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  36                         }
  37
  38                         //check for required parameters
  39                         fastaFile = validParameter.validFile(parameters, "fasta", true);
  40                         if (fastaFile == "not found") { cout << "fasta is a required parameter for the screen.seqs command." << endl; abort = true; }
  41                         else if (fastaFile == "not open") { abort = true; }
  42
  43
  44                         //check for optional parameter and set defaults
  45                         // ...at some point should added some additional type checking...
  46                         string temp;
  47                         temp = validParameter.validFile(parameters, "flip", false);
  48                         if (temp == "not found"){       flip = 0;       }
  49                         else if(isTrue(temp))   {       flip = 1;       }
  50
  51                         temp = validParameter.validFile(parameters, "oligos", true);
  52                         if (temp == "not found"){       oligoFile = "";         }
  53                         else if(temp == "not open"){    abort = true;   }
  54                         else                                    {       oligoFile = temp;               }
  55
  56                         temp = validParameter.validFile(parameters, "maxambig", false);         if (temp == "not found") { temp = "-1"; }
  57                         convert(temp, maxAmbig);
  58
  59                         temp = validParameter.validFile(parameters, "maxhomop", false);         if (temp == "not found") { temp = "0"; }
  60                         convert(temp, maxHomoP);
  61
  62                         temp = validParameter.validFile(parameters, "minlength", false);        if (temp == "not found") { temp = "0"; }
  63                         convert(temp, minLength);
  64
  65                         temp = validParameter.validFile(parameters, "maxlength", false);        if (temp == "not found") { temp = "0"; }
  66                         convert(temp, maxLength);
  67
  68                         temp = validParameter.validFile(parameters, "qfile", true);
  69                         if (temp == "not found")        {       qFileName = "";         }
  70                         else if(temp == "not open")     {       abort = 0;              }
  71                         else                                            {       qFileName = temp;       }
  72
  73                         temp = validParameter.validFile(parameters, "qthreshold", false);       if (temp == "not found") { temp = "0"; }
  74                         convert(temp, qThreshold);
  75
  76                         temp = validParameter.validFile(parameters, "qaverage", false);         if (temp == "not found") { temp = "0"; }
  77                         convert(temp, qAverage);
  78
  79                         temp = validParameter.validFile(parameters, "allfiles", false);         if (temp == "not found") { temp = "F"; }
  80                         allFiles = isTrue(temp);
  81
  82                         if(allFiles && oligoFile == ""){
  83                                 cout << "You selected allfiles, but didn't enter an oligos file.  Ignoring the allfiles request." << endl;
  84                         }
  85
  86                         if(!flip && oligoFile=="" && !maxLength && !minLength && (maxAmbig==-1) && !maxHomoP && qFileName == ""){
  87                                 cout << "You didn't set any options... quiting command." << endl;
  88                                 abort = true;
  89                         }
  90
  91                 }
  92
  93         }
  94         catch(exception& e) {
  95                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
  96                 exit(1);
  97         }
  98         catch(...) {
  99                 cout << "An unknown error has occurred in the TrimSeqsCommand class function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 100                 exit(1);
 101         }
 102 }
 103 //**********************************************************************************************************************
 104
 105 void TrimSeqsCommand::help(){
 106         try {
 107                 cout << "The trim.seqs command reads a fastaFile and creates ....." << "\n";
 108                 cout << "The trim.seqs command parameters are fasta, flip, oligos, maxambig, maxhomop, minlength and maxlength." << "\n";
 109                 cout << "The fasta parameter is required." << "\n";
 110                 cout << "The flip parameter .... The default is 0." << "\n";
 111                 cout << "The oligos parameter .... The default is ""." << "\n";
 112                 cout << "The maxambig parameter .... The default is -1." << "\n";
 113                 cout << "The maxhomop parameter .... The default is 0." << "\n";
 114                 cout << "The minlength parameter .... The default is 0." << "\n";
 115                 cout << "The maxlength parameter .... The default is 0." << "\n";
 116                 cout << "The trim.seqs command should be in the following format: " << "\n";
 117                 cout << "trim.seqs(fasta=yourFastaFile, flip=yourFlip, oligos=yourOligos, maxambig=yourMaxambig,  " << "\n";
 118                 cout << "maxhomop=yourMaxhomop, minlength=youMinlength, maxlength=yourMaxlength)  " << "\n";
 119                 cout << "Example trim.seqs(fasta=abrecovery.fasta, flip=..., oligos=..., maxambig=..., maxhomop=..., minlength=..., maxlength=...)." << "\n";
 120                 cout << "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta)." << "\n" << "\n";
 121
 122         }
 123         catch(exception& e) {
 124                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 125                 exit(1);
 126         }
 127         catch(...) {
 128                 cout << "An unknown error has occurred in the TrimSeqsCommand class function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 129                 exit(1);
 130         }
 131 }
 132
 133
 134 //***************************************************************************************************************
 135
 136 TrimSeqsCommand::~TrimSeqsCommand(){    /*      do nothing      */      }
 137
 138 //***************************************************************************************************************
 139
 140 int TrimSeqsCommand::execute(){
 141         try{
 142
 143                 if (abort == true) { return 0; }
 144
 145                 vector<ofstream*> groupFileNames;
 146                 vector<ofstream*> fastaFileNames;
 147                 if(oligoFile != "")     {       getOligos(fastaFileNames, groupFileNames);      }
 148
 149                 ifstream inFASTA;
 150                 openInputFile(fastaFile, inFASTA);
 151
 152                 ofstream outFASTA;
 153                 string trimSeqFile = getRootName(fastaFile) + "trim.fasta";
 154                 openOutputFile(trimSeqFile, outFASTA);
 155
 156                 ofstream outGroups;
 157                 string groupFile = getRootName(fastaFile) + "groups";
 158                 openOutputFile(groupFile, outGroups);
 159
 160                 ofstream scrapFASTA;
 161                 string scrapSeqFile = getRootName(fastaFile) + "scrap.fasta";
 162                 openOutputFile(scrapSeqFile, scrapFASTA);
 163
 164                 ifstream qFile;
 165                 if(qFileName != "")     {       openInputFile(qFileName, qFile);        }
 166
 167                 bool success;
 168
 169                 while(!inFASTA.eof()){
 170                         Sequence currSeq(inFASTA);
 171                         string origSeq = currSeq.getUnaligned();
 172                         int group;
 173                         string trashCode = "";
 174
 175                         if(qFileName != ""){
 176                                 if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 177                                 else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 178                                 if(!success)                    {       trashCode += 'q';                                                               }
 179                         }
 180                         if(barcodes.size() != 0){
 181                                 success = stripBarcode(currSeq, group);
 182                                 if(!success){   trashCode += 'b';       }
 183                         }
 184                         if(numFPrimers != 0){
 185                                 success = stripForward(currSeq);
 186                                 if(!success){   trashCode += 'f';       }
 187                         }
 188                         if(numRPrimers != 0){
 189                                 success = stripReverse(currSeq);
 190                                 if(!success){   trashCode += 'r';       }
 191                         }
 192                         if(minLength > 0 || maxLength > 0){
 193                                 success = cullLength(currSeq);
 194                                 if(!success){   trashCode += 'l';       }
 195                         }
 196                         if(maxHomoP > 0){
 197                                 success = cullHomoP(currSeq);
 198                                 if(!success){   trashCode += 'h';       }
 199                         }
 200                         if(maxAmbig != -1){
 201                                 success = cullAmbigs(currSeq);
 202                                 if(!success){   trashCode += 'n';       }
 203                         }
 204
 205                         if(flip){       currSeq.reverseComplement();    }               // should go last
 206
 207                         if(trashCode.length() == 0){
 208                                 currSeq.printSequence(outFASTA);
 209                                 if(barcodes.size() != 0){
 210                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 211
 212                                         if(allFiles){
 213                                                 *groupFileNames[group] << currSeq.getName() << '\t' << groupVector[group] << endl;
 214                                                 currSeq.printSequence(*fastaFileNames[group]);
 215                                         }
 216                                 }
 217                         }
 218                         else{
 219                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 220                                 currSeq.setUnaligned(origSeq);
 221                                 currSeq.printSequence(scrapFASTA);
 222                         }
 223                         gobble(inFASTA);
 224                 }
 225                 inFASTA.close();
 226                 outFASTA.close();
 227                 scrapFASTA.close();
 228                 outGroups.close();
 229
 230                 for(int i=0;i<groupFileNames.size();i++){
 231                         groupFileNames[i]->close();
 232                         delete groupFileNames[i];
 233
 234                         fastaFileNames[i]->close();
 235                         delete fastaFileNames[i];
 236                 }
 237
 238                 return 0;
 239         }
 240         catch(exception& e) {
 241                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 242                 exit(1);
 243         }
 244         catch(...) {
 245                 cout << "An unknown error has occurred in the TrimSeqsCommand class function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 246                 exit(1);
 247         }
 248 }
 249
 250 //***************************************************************************************************************
 251
 252 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec, vector<ofstream*>& outGroupsVec){
 253
 254         ifstream inOligos;
 255         openInputFile(oligoFile, inOligos);
 256
 257         ofstream test;
 258
 259         string type, oligo, group;
 260         int index=0;
 261
 262         while(!inOligos.eof()){
 263                 inOligos >> type;
 264
 265                 if(type[0] == '#'){
 266                         while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 267                 }
 268                 else{
 269                         inOligos >> oligo;
 270
 271                         for(int i=0;i<oligo.length();i++){
 272                                 oligo[i] = toupper(oligo[i]);
 273                                 if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 274                         }
 275
 276                         if(type == "forward"){
 277                                 forPrimer.push_back(oligo);
 278                         }
 279                         else if(type == "reverse"){
 280                                 revPrimer.push_back(oligo);
 281                         }
 282                         else if(type == "barcode"){
 283                                 inOligos >> group;
 284                                 barcodes[oligo]=index++;
 285                                 groupVector.push_back(group);
 286
 287                                 if(allFiles){
 288                                         outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 289                                         outGroupsVec.push_back(new ofstream((getRootName(fastaFile) + group + ".groups").c_str(), ios::ate));
 290                                 }
 291                         }
 292                 }
 293         }
 294
 295         numFPrimers = forPrimer.size();
 296         numRPrimers = revPrimer.size();
 297 }
 298
 299 //***************************************************************************************************************
 300
 301 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 302
 303         string rawSequence = seq.getUnaligned();
 304         bool success = 0;       //guilty until proven innocent
 305
 306         for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 307                 string oligo = it->first;
 308                 if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 309                         success = 0;
 310                         break;
 311                 }
 312
 313                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 314                         group = it->second;
 315                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 316                         success = 1;
 317                         break;
 318                 }
 319         }
 320         return success;
 321
 322 }
 323
 324 //***************************************************************************************************************
 325
 326 bool TrimSeqsCommand::stripForward(Sequence& seq){
 327
 328         string rawSequence = seq.getUnaligned();
 329         bool success = 0;       //guilty until proven innocent
 330
 331         for(int i=0;i<numFPrimers;i++){
 332                 string oligo = forPrimer[i];
 333
 334                 if(rawSequence.length() < oligo.length()){
 335                         success = 0;
 336                         break;
 337                 }
 338
 339                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 340                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 341                         success = 1;
 342                         break;
 343                 }
 344         }
 345
 346         return success;
 347
 348 }
 349
 350 //***************************************************************************************************************
 351
 352 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 353
 354         string rawSequence = seq.getUnaligned();
 355         bool success = 0;       //guilty until proven innocent
 356
 357         for(int i=0;i<numRPrimers;i++){
 358                 string oligo = revPrimer[i];
 359
 360                 if(rawSequence.length() < oligo.length()){
 361                         success = 0;
 362                         break;
 363                 }
 364
 365                 if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 366                         seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 367                         success = 1;
 368                         break;
 369                 }
 370         }
 371         return success;
 372
 373 }
 374
 375 //***************************************************************************************************************
 376
 377 bool TrimSeqsCommand::cullLength(Sequence& seq){
 378
 379         int length = seq.getNumBases();
 380         bool success = 0;       //guilty until proven innocent
 381
 382         if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 383         else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 384         else                                                                                            {       success = 0;    }
 385
 386         return success;
 387
 388 }
 389
 390 //***************************************************************************************************************
 391
 392 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 393
 394         int longHomoP = seq.getLongHomoPolymer();
 395         bool success = 0;       //guilty until proven innocent
 396
 397         if(longHomoP <= maxHomoP){      success = 1;    }
 398         else                                    {       success = 0;    }
 399
 400         return success;
 401
 402 }
 403
 404 //***************************************************************************************************************
 405
 406 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 407
 408         int numNs = seq.getAmbigBases();
 409         bool success = 0;       //guilty until proven innocent
 410
 411         if(numNs <= maxAmbig)   {       success = 1;    }
 412         else                                    {       success = 0;    }
 413
 414         return success;
 415
 416 }
 417
 418 //***************************************************************************************************************
 419
 420 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 421
 422         bool success = 1;
 423         int length = oligo.length();
 424
 425         for(int i=0;i<length;i++){
 426
 427                 if(oligo[i] != seq[i]){
 428                         if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 429                         else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 430                         else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 431                         else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 432                         else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 433                         else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 434                         else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 435                         else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 436                         else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 437                         else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 438                         else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 439                         else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 440
 441                         if(success == 0)        {       break;  }
 442                 }
 443                 else{
 444                         success = 1;
 445                 }
 446         }
 447
 448         return success;
 449 }
 450
 451 //***************************************************************************************************************
 452
 453 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 454
 455         string rawSequence = seq.getUnaligned();
 456         int seqLength = rawSequence.length();
 457         string name;
 458
 459         qFile >> name;
 460         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 461         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 462
 463         int score;
 464         int end = seqLength;
 465
 466         for(int i=0;i<seqLength;i++){
 467                 qFile >> score;
 468
 469                 if(score <= qThreshold){
 470                         end = i;
 471                         break;
 472                 }
 473         }
 474         for(int i=end+1;i<seqLength;i++){
 475                 qFile >> score;
 476         }
 477
 478         seq.setUnaligned(rawSequence.substr(0,end));
 479
 480         return 1;
 481 }
 482
 483 //***************************************************************************************************************
 484
 485 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 486
 487         string rawSequence = seq.getUnaligned();
 488         int seqLength = seq.getNumBases();
 489         bool success = 0;       //guilty until proven innocent
 490         string name;
 491
 492         qFile >> name;
 493         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 494         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 495
 496         float score;
 497         float average = 0;
 498
 499         for(int i=0;i<seqLength;i++){
 500                 qFile >> score;
 501                 average += score;
 502         }
 503         average /= seqLength;
 504
 505         if(average >= qAverage) {       success = 1;    }
 506         else                                    {       success = 0;    }
 507
 508         return success;
 509 }
 510
 511 //***************************************************************************************************************
 512
 513