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", "qtrim"};
  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, "qtrim", false);    if (temp == "not found") { temp = "F"; }
  77                         qtrim = isTrue(temp);
  78
  79                         temp = validParameter.validFile(parameters, "qaverage", false);         if (temp == "not found") { temp = "0"; }
  80                         convert(temp, qAverage);
  81
  82                         temp = validParameter.validFile(parameters, "allfiles", false);         if (temp == "not found") { temp = "F"; }
  83                         allFiles = isTrue(temp);
  84
  85                         if(allFiles && oligoFile == ""){
  86                                 mothurOut("You selected allfiles, but didn't enter an oligos file.  Ignoring the allfiles request."); mothurOutEndLine();
  87                         }
  88                         if((qAverage != 0 && qThreshold != 0) && qFileName == ""){
  89                                 mothurOut("You didn't provide a quality file name, quality criteria will be ignored."); mothurOutEndLine();
  90                                 qAverage=0;
  91                                 qThreshold=0;
  92                         }
  93                         if(!flip && oligoFile=="" && !maxLength && !minLength && (maxAmbig==-1) && !maxHomoP && qFileName == ""){
  94                                 mothurOut("You didn't set any options... quiting command."); mothurOutEndLine();
  95                                 abort = true;
  96                         }
  97                 }
  98
  99         }
 100         catch(exception& e) {
 101                 errorOut(e, "TrimSeqsCommand", "TrimSeqsCommand");
 102                 exit(1);
 103         }
 104 }
 105 //**********************************************************************************************************************
 106
 107 void TrimSeqsCommand::help(){
 108         try {
 109                 mothurOut("The trim.seqs command reads a fastaFile and creates .....\n");
 110                 mothurOut("The trim.seqs command parameters are fasta, flip, oligos, maxambig, maxhomop, minlength, maxlength, qfile, qthreshold, qaverage, qtrim and allfiles.\n");
 111                 mothurOut("The fasta parameter is required.\n");
 112                 mothurOut("The flip parameter .... The default is 0.\n");
 113                 mothurOut("The oligos parameter .... The default is "".\n");
 114                 mothurOut("The maxambig parameter .... The default is -1.\n");
 115                 mothurOut("The maxhomop parameter .... The default is 0.\n");
 116                 mothurOut("The minlength parameter .... The default is 0.\n");
 117                 mothurOut("The maxlength parameter .... The default is 0.\n");
 118                 mothurOut("The qfile parameter .....\n");
 119                 mothurOut("The qthreshold parameter .... The default is 0.\n");
 120                 mothurOut("The qaverage parameter .... The default is 0.\n");
 121                 mothurOut("The allfiles parameter .... The default is F.\n");
 122                 mothurOut("The qtrim parameter .... The default is F.\n");
 123                 mothurOut("The trim.seqs command should be in the following format: \n");
 124                 mothurOut("trim.seqs(fasta=yourFastaFile, flip=yourFlip, oligos=yourOligos, maxambig=yourMaxambig,  \n");
 125                 mothurOut("maxhomop=yourMaxhomop, minlength=youMinlength, maxlength=yourMaxlength)  \n");
 126                 mothurOut("Example trim.seqs(fasta=abrecovery.fasta, flip=..., oligos=..., maxambig=..., maxhomop=..., minlength=..., maxlength=...).\n");
 127                 mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n");
 128                 mothurOut("For more details please check out the wiki http://www.mothur.org/wiki/Trim.seqs .\n\n");
 129
 130         }
 131         catch(exception& e) {
 132                 errorOut(e, "TrimSeqsCommand", "help");
 133                 exit(1);
 134         }
 135 }
 136
 137
 138 //***************************************************************************************************************
 139
 140 TrimSeqsCommand::~TrimSeqsCommand(){    /*      do nothing      */      }
 141
 142 //***************************************************************************************************************
 143
 144 int TrimSeqsCommand::execute(){
 145         try{
 146
 147                 if (abort == true) { return 0; }
 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                 vector<ofstream*> fastaFileNames;
 158                 if(oligoFile != ""){
 159                         string groupFile = getRootName(fastaFile) + "groups";
 160                         openOutputFile(groupFile, outGroups);
 161                         getOligos(fastaFileNames);
 162                 }
 163
 164                 ofstream scrapFASTA;
 165                 string scrapSeqFile = getRootName(fastaFile) + "scrap.fasta";
 166                 openOutputFile(scrapSeqFile, scrapFASTA);
 167
 168                 ifstream qFile;
 169                 if(qFileName != "")     {       openInputFile(qFileName, qFile);        }
 170
 171                 bool success;
 172
 173                 while(!inFASTA.eof()){
 174                         Sequence currSeq(inFASTA);
 175                         string origSeq = currSeq.getUnaligned();
 176                         int group;
 177                         string trashCode = "";
 178
 179                         if(qFileName != ""){
 180                                 if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 181                                 else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 182                                 if ((!qtrim) && (origSeq.length() != currSeq.getUnaligned().length())) {
 183                                         success = 0; //if you don't want to trim and the sequence does not meet quality requirements, move to scrap
 184                                 }
 185                                 if(!success)                    {       trashCode += 'q';                                                               }
 186                         }
 187                         if(barcodes.size() != 0){
 188
 189                                 success = stripBarcode(currSeq, group);
 190                                 if(!success){   trashCode += 'b';       }
 191                         }
 192                         if(numFPrimers != 0){
 193                                 success = stripForward(currSeq);
 194                                 if(!success){   trashCode += 'f';       }
 195                         }
 196                         if(numRPrimers != 0){
 197                                 success = stripReverse(currSeq);
 198                                 if(!success){   trashCode += 'r';       }
 199                         }
 200                         if(minLength > 0 || maxLength > 0){
 201                                 success = cullLength(currSeq);
 202                         if ((currSeq.getUnaligned().length() > 300) && (success)) {  cout << "too long " << currSeq.getUnaligned().length() << endl;  }
 203                                 if(!success){   trashCode += 'l'; }
 204                         }
 205                         if(maxHomoP > 0){
 206                                 success = cullHomoP(currSeq);
 207                                 if(!success){   trashCode += 'h';       }
 208                         }
 209                         if(maxAmbig != -1){
 210                                 success = cullAmbigs(currSeq);
 211                                 if(!success){   trashCode += 'n';       }
 212                         }
 213
 214                         if(flip){       currSeq.reverseComplement();    }               // should go last
 215
 216                         if(trashCode.length() == 0){
 217                                 currSeq.setAligned(currSeq.getUnaligned());  //this is because of a modification we made to the sequence class to fix a bug.  all seqs have an aligned version, which is the version that gets printed.
 218                                 currSeq.printSequence(outFASTA);
 219                                 if(barcodes.size() != 0){
 220                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 221
 222                                         if(allFiles){
 223                                                 currSeq.printSequence(*fastaFileNames[group]);
 224                                         }
 225                                 }
 226                         }
 227                         else{
 228                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 229                                 currSeq.setUnaligned(origSeq);
 230                                 currSeq.printSequence(scrapFASTA);
 231                         }
 232                         gobble(inFASTA);
 233                 }
 234                 inFASTA.close();
 235                 outFASTA.close();
 236                 scrapFASTA.close();
 237                 outGroups.close();
 238                 if(qFileName != "")     {       qFile.close();  }
 239
 240                 for(int i=0;i<fastaFileNames.size();i++){
 241                         fastaFileNames[i]->close();
 242                         delete fastaFileNames[i];
 243                 }
 244
 245                 for(int i=0;i<fastaFileNames.size();i++){
 246                         string seqName;
 247                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 248                         ofstream outGroups;
 249                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 250
 251                         while(!inFASTA.eof()){
 252                                 if(inFASTA.get() == '>'){
 253                                         inFASTA >> seqName;
 254                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 255                                 }
 256                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 257                         }
 258                         outGroups.close();
 259                         inFASTA.close();
 260                 }
 261
 262
 263                 return 0;
 264         }
 265         catch(exception& e) {
 266                 errorOut(e, "TrimSeqsCommand", "execute");
 267                 exit(1);
 268         }
 269 }
 270
 271 //***************************************************************************************************************
 272
 273 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 274         try {
 275                 ifstream inOligos;
 276                 openInputFile(oligoFile, inOligos);
 277
 278                 ofstream test;
 279
 280                 string type, oligo, group;
 281                 int index=0;
 282
 283                 while(!inOligos.eof()){
 284                         inOligos >> type;
 285
 286                         if(type[0] == '#'){
 287                                 while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 288                         }
 289                         else{
 290                                 inOligos >> oligo;
 291
 292                                 for(int i=0;i<oligo.length();i++){
 293                                         oligo[i] = toupper(oligo[i]);
 294                                         if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 295                                 }
 296
 297                                 if(type == "forward"){
 298                                         forPrimer.push_back(oligo);
 299                                 }
 300                                 else if(type == "reverse"){
 301                                         revPrimer.push_back(oligo);
 302                                 }
 303                                 else if(type == "barcode"){
 304                                         inOligos >> group;
 305                                         barcodes[oligo]=index++;
 306                                         groupVector.push_back(group);
 307
 308                                         if(allFiles){
 309                                                 outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 310                                         }
 311                                 }
 312                         }
 313                 }
 314
 315                 inOligos.close();
 316
 317                 numFPrimers = forPrimer.size();
 318                 numRPrimers = revPrimer.size();
 319
 320         }
 321         catch(exception& e) {
 322                 errorOut(e, "TrimSeqsCommand", "getOligos");
 323                 exit(1);
 324         }
 325
 326 }
 327
 328 //***************************************************************************************************************
 329
 330 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 331         try {
 332                 string rawSequence = seq.getUnaligned();
 333                 bool success = 0;       //guilty until proven innocent
 334
 335                 for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 336                         string oligo = it->first;
 337                         if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 338                                 success = 0;
 339                                 break;
 340                         }
 341
 342                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 343                                 group = it->second;
 344                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 345                                 success = 1;
 346                                 break;
 347                         }
 348                 }
 349                 return success;
 350
 351         }
 352         catch(exception& e) {
 353                 errorOut(e, "TrimSeqsCommand", "stripBarcode");
 354                 exit(1);
 355         }
 356
 357 }
 358
 359 //***************************************************************************************************************
 360
 361 bool TrimSeqsCommand::stripForward(Sequence& seq){
 362         try {
 363                 string rawSequence = seq.getUnaligned();
 364                 bool success = 0;       //guilty until proven innocent
 365
 366                 for(int i=0;i<numFPrimers;i++){
 367                         string oligo = forPrimer[i];
 368
 369                         if(rawSequence.length() < oligo.length()){
 370                                 success = 0;
 371                                 break;
 372                         }
 373
 374                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 375                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 376                                 success = 1;
 377                                 break;
 378                         }
 379                 }
 380
 381                 return success;
 382
 383         }
 384         catch(exception& e) {
 385                 errorOut(e, "TrimSeqsCommand", "stripForward");
 386                 exit(1);
 387         }
 388 }
 389
 390 //***************************************************************************************************************
 391
 392 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 393         try {
 394                 string rawSequence = seq.getUnaligned();
 395                 bool success = 0;       //guilty until proven innocent
 396
 397                 for(int i=0;i<numRPrimers;i++){
 398                         string oligo = revPrimer[i];
 399
 400                         if(rawSequence.length() < oligo.length()){
 401                                 success = 0;
 402                                 break;
 403                         }
 404
 405                         if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 406                                 seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 407                                 success = 1;
 408                                 break;
 409                         }
 410                 }
 411                 return success;
 412
 413         }
 414         catch(exception& e) {
 415                 errorOut(e, "TrimSeqsCommand", "stripReverse");
 416                 exit(1);
 417         }
 418 }
 419
 420 //***************************************************************************************************************
 421
 422 bool TrimSeqsCommand::cullLength(Sequence& seq){
 423         try {
 424
 425                 int length = seq.getNumBases();
 426                 bool success = 0;       //guilty until proven innocent
 427
 428                 if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 429                 else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 430                 else                                                                                            {       success = 0;    }
 431
 432                 return success;
 433
 434         }
 435         catch(exception& e) {
 436                 errorOut(e, "TrimSeqsCommand", "cullLength");
 437                 exit(1);
 438         }
 439
 440 }
 441
 442 //***************************************************************************************************************
 443
 444 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 445         try {
 446                 int longHomoP = seq.getLongHomoPolymer();
 447                 bool success = 0;       //guilty until proven innocent
 448
 449                 if(longHomoP <= maxHomoP){      success = 1;    }
 450                 else                                    {       success = 0;    }
 451
 452                 return success;
 453         }
 454         catch(exception& e) {
 455                 errorOut(e, "TrimSeqsCommand", "cullHomoP");
 456                 exit(1);
 457         }
 458
 459 }
 460
 461 //***************************************************************************************************************
 462
 463 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 464         try {
 465                 int numNs = seq.getAmbigBases();
 466                 bool success = 0;       //guilty until proven innocent
 467
 468                 if(numNs <= maxAmbig)   {       success = 1;    }
 469                 else                                    {       success = 0;    }
 470
 471                 return success;
 472         }
 473         catch(exception& e) {
 474                 errorOut(e, "TrimSeqsCommand", "cullAmbigs");
 475                 exit(1);
 476         }
 477
 478 }
 479
 480 //***************************************************************************************************************
 481
 482 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 483         try {
 484                 bool success = 1;
 485                 int length = oligo.length();
 486
 487                 for(int i=0;i<length;i++){
 488
 489                         if(oligo[i] != seq[i]){
 490                                 if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 491                                 else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 492                                 else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 493                                 else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 494                                 else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 495                                 else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 496                                 else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 497                                 else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 498                                 else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 499                                 else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 500                                 else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 501                                 else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 502
 503                                 if(success == 0)        {       break;  }
 504                         }
 505                         else{
 506                                 success = 1;
 507                         }
 508                 }
 509
 510                 return success;
 511         }
 512         catch(exception& e) {
 513                 errorOut(e, "TrimSeqsCommand", "compareDNASeq");
 514                 exit(1);
 515         }
 516
 517 }
 518
 519 //***************************************************************************************************************
 520
 521 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 522         try {
 523                 string rawSequence = seq.getUnaligned();
 524                 int seqLength = rawSequence.length();
 525                 string name;
 526
 527                 qFile >> name;
 528                 if (name.length() != 0) {  if(name.substr(1) != seq.getName())  {       mothurOut("sequence name mismatch btwn fasta and qual file"); mothurOutEndLine();       }  }
 529                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 530
 531                 int score;
 532                 int end = seqLength;
 533
 534                 for(int i=0;i<seqLength;i++){
 535                         qFile >> score;
 536
 537                         if(score <= qThreshold){
 538                                 end = i;
 539                                 break;
 540                         }
 541                 }
 542                 for(int i=end+1;i<seqLength;i++){
 543                         qFile >> score;
 544                 }
 545
 546                 seq.setUnaligned(rawSequence.substr(0,end));
 547
 548                 return 1;
 549         }
 550         catch(exception& e) {
 551                 errorOut(e, "TrimSeqsCommand", "stripQualThreshold");
 552                 exit(1);
 553         }
 554 }
 555
 556 //***************************************************************************************************************
 557
 558 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 559         try {
 560                 string rawSequence = seq.getUnaligned();
 561                 int seqLength = seq.getNumBases();
 562                 bool success = 0;       //guilty until proven innocent
 563                 string name;
 564
 565                 qFile >> name;
 566                 if (name[0] == '>') {  if(name.substr(1) != seq.getName())      {       mothurOut("sequence name mismatch btwn fasta: " + seq.getName() + " and qual file: " + name); mothurOutEndLine();       } }
 567
 568                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 569
 570                 float score;
 571                 float average = 0;
 572
 573                 for(int i=0;i<seqLength;i++){
 574                         qFile >> score;
 575                         average += score;
 576                 }
 577                 average /= seqLength;
 578
 579                 if(average >= qAverage) {       success = 1;    }
 580                 else                                    {       success = 0;    }
 581
 582                 return success;
 583         }
 584         catch(exception& e) {
 585                 errorOut(e, "TrimSeqsCommand", "cullQualAverage");
 586                 exit(1);
 587         }
 588 }
 589
 590 //***************************************************************************************************************