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") { mothurOut("fasta is a required parameter for the screen.seqs command."); mothurOutEndLine(); 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                                 mothurOut("You selected allfiles, but didn't enter an oligos file.  Ignoring the allfiles request."); mothurOutEndLine();
  84                         }
  85                         if((qAverage != 0 && qThreshold != 0) && qFileName == ""){
  86                                 mothurOut("You didn't provide a quality file name, quality criteria will be ignored."); mothurOutEndLine();
  87                                 qAverage=0;
  88                                 qThreshold=0;
  89                         }
  90                         if(!flip && oligoFile=="" && !maxLength && !minLength && (maxAmbig==-1) && !maxHomoP && qFileName == ""){
  91                                 mothurOut("You didn't set any options... quiting command."); mothurOutEndLine();
  92                                 abort = true;
  93                         }
  94                 }
  95
  96         }
  97         catch(exception& e) {
  98                 errorOut(e, "TrimSeqsCommand", "TrimSeqsCommand");
  99                 exit(1);
 100         }
 101 }
 102 //**********************************************************************************************************************
 103
 104 void TrimSeqsCommand::help(){
 105         try {
 106                 mothurOut("The trim.seqs command reads a fastaFile and creates .....\n");
 107                 mothurOut("The trim.seqs command parameters are fasta, flip, oligos, maxambig, maxhomop, minlength and maxlength.\n");
 108                 mothurOut("The fasta parameter is required.\n");
 109                 mothurOut("The flip parameter .... The default is 0.\n");
 110                 mothurOut("The oligos parameter .... The default is "".\n");
 111                 mothurOut("The maxambig parameter .... The default is -1.\n");
 112                 mothurOut("The maxhomop parameter .... The default is 0.\n");
 113                 mothurOut("The minlength parameter .... The default is 0.\n");
 114                 mothurOut("The maxlength parameter .... The default is 0.\n");
 115                 mothurOut("The trim.seqs command should be in the following format: \n");
 116                 mothurOut("trim.seqs(fasta=yourFastaFile, flip=yourFlip, oligos=yourOligos, maxambig=yourMaxambig,  \n");
 117                 mothurOut("maxhomop=yourMaxhomop, minlength=youMinlength, maxlength=yourMaxlength)  \n");
 118                 mothurOut("Example trim.seqs(fasta=abrecovery.fasta, flip=..., oligos=..., maxambig=..., maxhomop=..., minlength=..., maxlength=...).\n");
 119                 mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
 120
 121         }
 122         catch(exception& e) {
 123                 errorOut(e, "TrimSeqsCommand", "help");
 124                 exit(1);
 125         }
 126 }
 127
 128
 129 //***************************************************************************************************************
 130
 131 TrimSeqsCommand::~TrimSeqsCommand(){    /*      do nothing      */      }
 132
 133 //***************************************************************************************************************
 134
 135 int TrimSeqsCommand::execute(){
 136         try{
 137
 138                 if (abort == true) { return 0; }
 139
 140                 ifstream inFASTA;
 141                 openInputFile(fastaFile, inFASTA);
 142
 143                 ofstream outFASTA;
 144                 string trimSeqFile = getRootName(fastaFile) + "trim.fasta";
 145                 openOutputFile(trimSeqFile, outFASTA);
 146
 147                 ofstream outGroups;
 148                 vector<ofstream*> fastaFileNames;
 149                 if(oligoFile != ""){
 150                         string groupFile = getRootName(fastaFile) + "groups";
 151                         openOutputFile(groupFile, outGroups);
 152                         getOligos(fastaFileNames);
 153                 }
 154
 155                 ofstream scrapFASTA;
 156                 string scrapSeqFile = getRootName(fastaFile) + "scrap.fasta";
 157                 openOutputFile(scrapSeqFile, scrapFASTA);
 158
 159                 ifstream qFile;
 160                 if(qFileName != "")     {       openInputFile(qFileName, qFile);        }
 161
 162                 bool success;
 163
 164                 while(!inFASTA.eof()){
 165                         Sequence currSeq(inFASTA);
 166                         string origSeq = currSeq.getUnaligned();
 167                         int group;
 168                         string trashCode = "";
 169
 170                         if(qFileName != ""){
 171                                 if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 172                                 else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 173                                 if(!success)                    {       trashCode += 'q';                                                               }
 174                                 qFile.close();
 175                         }
 176                         if(barcodes.size() != 0){
 177                                 success = stripBarcode(currSeq, group);
 178                                 if(!success){   trashCode += 'b';       }
 179                         }
 180                         if(numFPrimers != 0){
 181                                 success = stripForward(currSeq);
 182                                 if(!success){   trashCode += 'f';       }
 183                         }
 184                         if(numRPrimers != 0){
 185                                 success = stripReverse(currSeq);
 186                                 if(!success){   trashCode += 'r';       }
 187                         }
 188                         if(minLength > 0 || maxLength > 0){
 189                                 success = cullLength(currSeq);
 190                                 if(!success){   trashCode += 'l';       }
 191                         }
 192                         if(maxHomoP > 0){
 193                                 success = cullHomoP(currSeq);
 194                                 if(!success){   trashCode += 'h';       }
 195                         }
 196                         if(maxAmbig != -1){
 197                                 success = cullAmbigs(currSeq);
 198                                 if(!success){   trashCode += 'n';       }
 199                         }
 200
 201                         if(flip){       currSeq.reverseComplement();    }               // should go last
 202
 203                         if(trashCode.length() == 0){
 204                                 currSeq.printSequence(outFASTA);
 205                                 if(barcodes.size() != 0){
 206                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 207
 208                                         if(allFiles){
 209                                                 currSeq.printSequence(*fastaFileNames[group]);
 210                                         }
 211                                 }
 212                         }
 213                         else{
 214                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 215                                 currSeq.setUnaligned(origSeq);
 216                                 currSeq.printSequence(scrapFASTA);
 217                         }
 218                         gobble(inFASTA);
 219                 }
 220                 inFASTA.close();
 221                 outFASTA.close();
 222                 scrapFASTA.close();
 223                 outGroups.close();
 224
 225                 for(int i=0;i<fastaFileNames.size();i++){
 226                         fastaFileNames[i]->close();
 227                         delete fastaFileNames[i];
 228                 }
 229
 230                 for(int i=0;i<fastaFileNames.size();i++){
 231                         string seqName;
 232                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 233                         ofstream outGroups;
 234                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 235
 236                         while(!inFASTA.eof()){
 237                                 if(inFASTA.get() == '>'){
 238                                         inFASTA >> seqName;
 239                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 240                                 }
 241                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 242                         }
 243                         outGroups.close();
 244                         inFASTA.close();
 245                 }
 246
 247
 248                 return 0;
 249         }
 250         catch(exception& e) {
 251                 errorOut(e, "TrimSeqsCommand", "execute");
 252                 exit(1);
 253         }
 254 }
 255
 256 //***************************************************************************************************************
 257
 258 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 259
 260         ifstream inOligos;
 261         openInputFile(oligoFile, inOligos);
 262
 263         ofstream test;
 264
 265         string type, oligo, group;
 266         int index=0;
 267
 268         while(!inOligos.eof()){
 269                 inOligos >> type;
 270
 271                 if(type[0] == '#'){
 272                         while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 273                 }
 274                 else{
 275                         inOligos >> oligo;
 276
 277                         for(int i=0;i<oligo.length();i++){
 278                                 oligo[i] = toupper(oligo[i]);
 279                                 if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 280                         }
 281
 282                         if(type == "forward"){
 283                                 forPrimer.push_back(oligo);
 284                         }
 285                         else if(type == "reverse"){
 286                                 revPrimer.push_back(oligo);
 287                         }
 288                         else if(type == "barcode"){
 289                                 inOligos >> group;
 290                                 barcodes[oligo]=index++;
 291                                 groupVector.push_back(group);
 292
 293                                 if(allFiles){
 294                                         outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 295                                 }
 296                         }
 297                 }
 298         }
 299
 300         inOligos.close();
 301
 302         numFPrimers = forPrimer.size();
 303         numRPrimers = revPrimer.size();
 304 }
 305
 306 //***************************************************************************************************************
 307
 308 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 309
 310         string rawSequence = seq.getUnaligned();
 311         bool success = 0;       //guilty until proven innocent
 312
 313         for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 314                 string oligo = it->first;
 315                 if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 316                         success = 0;
 317                         break;
 318                 }
 319
 320                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 321                         group = it->second;
 322                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 323                         success = 1;
 324                         break;
 325                 }
 326         }
 327         return success;
 328
 329 }
 330
 331 //***************************************************************************************************************
 332
 333 bool TrimSeqsCommand::stripForward(Sequence& seq){
 334
 335         string rawSequence = seq.getUnaligned();
 336         bool success = 0;       //guilty until proven innocent
 337
 338         for(int i=0;i<numFPrimers;i++){
 339                 string oligo = forPrimer[i];
 340
 341                 if(rawSequence.length() < oligo.length()){
 342                         success = 0;
 343                         break;
 344                 }
 345
 346                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 347                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 348                         success = 1;
 349                         break;
 350                 }
 351         }
 352
 353         return success;
 354
 355 }
 356
 357 //***************************************************************************************************************
 358
 359 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 360
 361         string rawSequence = seq.getUnaligned();
 362         bool success = 0;       //guilty until proven innocent
 363
 364         for(int i=0;i<numRPrimers;i++){
 365                 string oligo = revPrimer[i];
 366
 367                 if(rawSequence.length() < oligo.length()){
 368                         success = 0;
 369                         break;
 370                 }
 371
 372                 if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 373                         seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 374                         success = 1;
 375                         break;
 376                 }
 377         }
 378         return success;
 379
 380 }
 381
 382 //***************************************************************************************************************
 383
 384 bool TrimSeqsCommand::cullLength(Sequence& seq){
 385
 386         int length = seq.getNumBases();
 387         bool success = 0;       //guilty until proven innocent
 388
 389         if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 390         else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 391         else                                                                                            {       success = 0;    }
 392
 393         return success;
 394
 395 }
 396
 397 //***************************************************************************************************************
 398
 399 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 400
 401         int longHomoP = seq.getLongHomoPolymer();
 402         bool success = 0;       //guilty until proven innocent
 403
 404         if(longHomoP <= maxHomoP){      success = 1;    }
 405         else                                    {       success = 0;    }
 406
 407         return success;
 408
 409 }
 410
 411 //***************************************************************************************************************
 412
 413 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 414
 415         int numNs = seq.getAmbigBases();
 416         bool success = 0;       //guilty until proven innocent
 417
 418         if(numNs <= maxAmbig)   {       success = 1;    }
 419         else                                    {       success = 0;    }
 420
 421         return success;
 422
 423 }
 424
 425 //***************************************************************************************************************
 426
 427 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 428
 429         bool success = 1;
 430         int length = oligo.length();
 431
 432         for(int i=0;i<length;i++){
 433
 434                 if(oligo[i] != seq[i]){
 435                         if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 436                         else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 437                         else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 438                         else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 439                         else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 440                         else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 441                         else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 442                         else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 443                         else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 444                         else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 445                         else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 446                         else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 447
 448                         if(success == 0)        {       break;  }
 449                 }
 450                 else{
 451                         success = 1;
 452                 }
 453         }
 454
 455         return success;
 456 }
 457
 458 //***************************************************************************************************************
 459
 460 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 461
 462         string rawSequence = seq.getUnaligned();
 463         int seqLength = rawSequence.length();
 464         string name;
 465
 466         qFile >> name;
 467         if(name.substr(1) != seq.getName())     {       mothurOut("sequence name mismatch btwn fasta and qual file"); mothurOutEndLine();       }
 468         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 469
 470         int score;
 471         int end = seqLength;
 472
 473         for(int i=0;i<seqLength;i++){
 474                 qFile >> score;
 475
 476                 if(score <= qThreshold){
 477                         end = i;
 478                         break;
 479                 }
 480         }
 481         for(int i=end+1;i<seqLength;i++){
 482                 qFile >> score;
 483         }
 484
 485         seq.setUnaligned(rawSequence.substr(0,end));
 486
 487         return 1;
 488 }
 489
 490 //***************************************************************************************************************
 491
 492 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 493
 494         string rawSequence = seq.getUnaligned();
 495         int seqLength = seq.getNumBases();
 496         bool success = 0;       //guilty until proven innocent
 497         string name;
 498
 499         qFile >> name;
 500         if(name.substr(1) != seq.getName())     {       mothurOut("sequence name mismatch btwn fasta and qual file"); mothurOutEndLine();       }
 501         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 502
 503         float score;
 504         float average = 0;
 505
 506         for(int i=0;i<seqLength;i++){
 507                 qFile >> score;
 508                 average += score;
 509         }
 510         average /= seqLength;
 511
 512         if(average >= qAverage) {       success = 1;    }
 513         else                                    {       success = 0;    }
 514
 515         return success;
 516 }
 517
 518 //***************************************************************************************************************
 519
 520