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                         if (origSeq != "") {
 177                                 int group;
 178                                 string trashCode = "";
 179
 180                                 if(qFileName != ""){
 181                                         if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 182                                         else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 183                                         if ((!qtrim) && (origSeq.length() != currSeq.getUnaligned().length())) {
 184                                                 success = 0; //if you don't want to trim and the sequence does not meet quality requirements, move to scrap
 185                                         }
 186                                         if(!success)                    {       trashCode += 'q';                                                               }
 187                                 }
 188                                 if(barcodes.size() != 0){
 189
 190                                         success = stripBarcode(currSeq, group);
 191                                         if(!success){   trashCode += 'b';       }
 192                                 }
 193                                 if(numFPrimers != 0){
 194                                         success = stripForward(currSeq);
 195                                         if(!success){   trashCode += 'f';       }
 196                                 }
 197                                 if(numRPrimers != 0){
 198                                         success = stripReverse(currSeq);
 199                                         if(!success){   trashCode += 'r';       }
 200                                 }
 201                                 if(minLength > 0 || maxLength > 0){
 202                                         success = cullLength(currSeq);
 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                         }
 233                         gobble(inFASTA);
 234                 }
 235                 inFASTA.close();
 236                 outFASTA.close();
 237                 scrapFASTA.close();
 238                 outGroups.close();
 239                 if(qFileName != "")     {       qFile.close();  }
 240
 241                 for(int i=0;i<fastaFileNames.size();i++){
 242                         fastaFileNames[i]->close();
 243                         delete fastaFileNames[i];
 244                 }
 245
 246                 for(int i=0;i<fastaFileNames.size();i++){
 247                         string seqName;
 248                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 249                         ofstream outGroups;
 250                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 251
 252                         while(!inFASTA.eof()){
 253                                 if(inFASTA.get() == '>'){
 254                                         inFASTA >> seqName;
 255                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 256                                 }
 257                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 258                         }
 259                         outGroups.close();
 260                         inFASTA.close();
 261                 }
 262
 263
 264                 return 0;
 265         }
 266         catch(exception& e) {
 267                 errorOut(e, "TrimSeqsCommand", "execute");
 268                 exit(1);
 269         }
 270 }
 271
 272 //***************************************************************************************************************
 273
 274 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 275         try {
 276                 ifstream inOligos;
 277                 openInputFile(oligoFile, inOligos);
 278
 279                 ofstream test;
 280
 281                 string type, oligo, group;
 282                 int index=0;
 283
 284                 while(!inOligos.eof()){
 285                         inOligos >> type;
 286
 287                         if(type[0] == '#'){
 288                                 while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 289                         }
 290                         else{
 291                                 inOligos >> oligo;
 292
 293                                 for(int i=0;i<oligo.length();i++){
 294                                         oligo[i] = toupper(oligo[i]);
 295                                         if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 296                                 }
 297
 298                                 if(type == "forward"){
 299                                         forPrimer.push_back(oligo);
 300                                 }
 301                                 else if(type == "reverse"){
 302                                         revPrimer.push_back(oligo);
 303                                 }
 304                                 else if(type == "barcode"){
 305                                         inOligos >> group;
 306                                         barcodes[oligo]=index++;
 307                                         groupVector.push_back(group);
 308
 309                                         if(allFiles){
 310                                                 outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 311                                         }
 312                                 }
 313                         }
 314                 }
 315
 316                 inOligos.close();
 317
 318                 numFPrimers = forPrimer.size();
 319                 numRPrimers = revPrimer.size();
 320
 321         }
 322         catch(exception& e) {
 323                 errorOut(e, "TrimSeqsCommand", "getOligos");
 324                 exit(1);
 325         }
 326
 327 }
 328
 329 //***************************************************************************************************************
 330
 331 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 332         try {
 333                 string rawSequence = seq.getUnaligned();
 334                 bool success = 0;       //guilty until proven innocent
 335
 336                 for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 337                         string oligo = it->first;
 338                         if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 339                                 success = 0;
 340                                 break;
 341                         }
 342
 343                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 344                                 group = it->second;
 345                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 346                                 success = 1;
 347                                 break;
 348                         }
 349                 }
 350                 return success;
 351
 352         }
 353         catch(exception& e) {
 354                 errorOut(e, "TrimSeqsCommand", "stripBarcode");
 355                 exit(1);
 356         }
 357
 358 }
 359
 360 //***************************************************************************************************************
 361
 362 bool TrimSeqsCommand::stripForward(Sequence& seq){
 363         try {
 364                 string rawSequence = seq.getUnaligned();
 365                 bool success = 0;       //guilty until proven innocent
 366
 367                 for(int i=0;i<numFPrimers;i++){
 368                         string oligo = forPrimer[i];
 369
 370                         if(rawSequence.length() < oligo.length()){
 371                                 success = 0;
 372                                 break;
 373                         }
 374
 375                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 376                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 377                                 success = 1;
 378                                 break;
 379                         }
 380                 }
 381
 382                 return success;
 383
 384         }
 385         catch(exception& e) {
 386                 errorOut(e, "TrimSeqsCommand", "stripForward");
 387                 exit(1);
 388         }
 389 }
 390
 391 //***************************************************************************************************************
 392
 393 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 394         try {
 395                 string rawSequence = seq.getUnaligned();
 396                 bool success = 0;       //guilty until proven innocent
 397
 398                 for(int i=0;i<numRPrimers;i++){
 399                         string oligo = revPrimer[i];
 400
 401                         if(rawSequence.length() < oligo.length()){
 402                                 success = 0;
 403                                 break;
 404                         }
 405
 406                         if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 407                                 seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 408                                 success = 1;
 409                                 break;
 410                         }
 411                 }
 412                 return success;
 413
 414         }
 415         catch(exception& e) {
 416                 errorOut(e, "TrimSeqsCommand", "stripReverse");
 417                 exit(1);
 418         }
 419 }
 420
 421 //***************************************************************************************************************
 422
 423 bool TrimSeqsCommand::cullLength(Sequence& seq){
 424         try {
 425
 426                 int length = seq.getNumBases();
 427                 bool success = 0;       //guilty until proven innocent
 428
 429                 if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 430                 else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 431                 else                                                                                            {       success = 0;    }
 432
 433                 return success;
 434
 435         }
 436         catch(exception& e) {
 437                 errorOut(e, "TrimSeqsCommand", "cullLength");
 438                 exit(1);
 439         }
 440
 441 }
 442
 443 //***************************************************************************************************************
 444
 445 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 446         try {
 447                 int longHomoP = seq.getLongHomoPolymer();
 448                 bool success = 0;       //guilty until proven innocent
 449
 450                 if(longHomoP <= maxHomoP){      success = 1;    }
 451                 else                                    {       success = 0;    }
 452
 453                 return success;
 454         }
 455         catch(exception& e) {
 456                 errorOut(e, "TrimSeqsCommand", "cullHomoP");
 457                 exit(1);
 458         }
 459
 460 }
 461
 462 //***************************************************************************************************************
 463
 464 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 465         try {
 466                 int numNs = seq.getAmbigBases();
 467                 bool success = 0;       //guilty until proven innocent
 468
 469                 if(numNs <= maxAmbig)   {       success = 1;    }
 470                 else                                    {       success = 0;    }
 471
 472                 return success;
 473         }
 474         catch(exception& e) {
 475                 errorOut(e, "TrimSeqsCommand", "cullAmbigs");
 476                 exit(1);
 477         }
 478
 479 }
 480
 481 //***************************************************************************************************************
 482
 483 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 484         try {
 485                 bool success = 1;
 486                 int length = oligo.length();
 487
 488                 for(int i=0;i<length;i++){
 489
 490                         if(oligo[i] != seq[i]){
 491                                 if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 492                                 else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 493                                 else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 494                                 else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 495                                 else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 496                                 else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 497                                 else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 498                                 else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 499                                 else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 500                                 else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 501                                 else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 502                                 else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 503
 504                                 if(success == 0)        {       break;  }
 505                         }
 506                         else{
 507                                 success = 1;
 508                         }
 509                 }
 510
 511                 return success;
 512         }
 513         catch(exception& e) {
 514                 errorOut(e, "TrimSeqsCommand", "compareDNASeq");
 515                 exit(1);
 516         }
 517
 518 }
 519
 520 //***************************************************************************************************************
 521
 522 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 523         try {
 524                 string rawSequence = seq.getUnaligned();
 525                 int seqLength = rawSequence.length();
 526                 string name;
 527
 528                 qFile >> name;
 529                 if (name.length() != 0) {  if(name.substr(1) != seq.getName())  {       mothurOut("sequence name mismatch btwn fasta and qual file"); mothurOutEndLine();       }  }
 530                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 531
 532                 int score;
 533                 int end = seqLength;
 534
 535                 for(int i=0;i<seqLength;i++){
 536                         qFile >> score;
 537
 538                         if(score <= qThreshold){
 539                                 end = i;
 540                                 break;
 541                         }
 542                 }
 543                 for(int i=end+1;i<seqLength;i++){
 544                         qFile >> score;
 545                 }
 546
 547                 seq.setUnaligned(rawSequence.substr(0,end));
 548
 549                 return 1;
 550         }
 551         catch(exception& e) {
 552                 errorOut(e, "TrimSeqsCommand", "stripQualThreshold");
 553                 exit(1);
 554         }
 555 }
 556
 557 //***************************************************************************************************************
 558
 559 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 560         try {
 561                 string rawSequence = seq.getUnaligned();
 562                 int seqLength = seq.getNumBases();
 563                 bool success = 0;       //guilty until proven innocent
 564                 string name;
 565
 566                 qFile >> name;
 567                 if (name[0] == '>') {  if(name.substr(1) != seq.getName())      {       mothurOut("sequence name mismatch btwn fasta: " + seq.getName() + " and qual file: " + name); mothurOutEndLine();       } }
 568
 569                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 570
 571                 float score;
 572                 float average = 0;
 573
 574                 for(int i=0;i<seqLength;i++){
 575                         qFile >> score;
 576                         average += score;
 577                 }
 578                 average /= seqLength;
 579
 580                 if(average >= qAverage) {       success = 1;    }
 581                 else                                    {       success = 0;    }
 582
 583                 return success;
 584         }
 585         catch(exception& e) {
 586                 errorOut(e, "TrimSeqsCommand", "cullQualAverage");
 587                 exit(1);
 588         }
 589 }
 590
 591 //***************************************************************************************************************