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                         }
 175                         if(barcodes.size() != 0){
 176
 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 ((currSeq.getUnaligned().length() > 300) && (success)) {  cout << "too long " << currSeq.getUnaligned().length() << endl;  }
 191                                 if(!success){   trashCode += 'l'; }
 192                         }
 193                         if(maxHomoP > 0){
 194                                 success = cullHomoP(currSeq);
 195                                 if(!success){   trashCode += 'h';       }
 196                         }
 197                         if(maxAmbig != -1){
 198                                 success = cullAmbigs(currSeq);
 199                                 if(!success){   trashCode += 'n';       }
 200                         }
 201
 202                         if(flip){       currSeq.reverseComplement();    }               // should go last
 203
 204                         if(trashCode.length() == 0){
 205                                 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.
 206                                 currSeq.printSequence(outFASTA);
 207                                 if(barcodes.size() != 0){
 208                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 209
 210                                         if(allFiles){
 211                                                 currSeq.printSequence(*fastaFileNames[group]);
 212                                         }
 213                                 }
 214                         }
 215                         else{
 216                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 217                                 currSeq.setUnaligned(origSeq);
 218                                 currSeq.printSequence(scrapFASTA);
 219                         }
 220                         gobble(inFASTA);
 221                 }
 222                 inFASTA.close();
 223                 outFASTA.close();
 224                 scrapFASTA.close();
 225                 outGroups.close();
 226                 if(qFileName != "")     {       qFile.close();  }
 227
 228                 for(int i=0;i<fastaFileNames.size();i++){
 229                         fastaFileNames[i]->close();
 230                         delete fastaFileNames[i];
 231                 }
 232
 233                 for(int i=0;i<fastaFileNames.size();i++){
 234                         string seqName;
 235                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 236                         ofstream outGroups;
 237                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 238
 239                         while(!inFASTA.eof()){
 240                                 if(inFASTA.get() == '>'){
 241                                         inFASTA >> seqName;
 242                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 243                                 }
 244                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 245                         }
 246                         outGroups.close();
 247                         inFASTA.close();
 248                 }
 249
 250
 251                 return 0;
 252         }
 253         catch(exception& e) {
 254                 errorOut(e, "TrimSeqsCommand", "execute");
 255                 exit(1);
 256         }
 257 }
 258
 259 //***************************************************************************************************************
 260
 261 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 262         try {
 263                 ifstream inOligos;
 264                 openInputFile(oligoFile, inOligos);
 265
 266                 ofstream test;
 267
 268                 string type, oligo, group;
 269                 int index=0;
 270
 271                 while(!inOligos.eof()){
 272                         inOligos >> type;
 273
 274                         if(type[0] == '#'){
 275                                 while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 276                         }
 277                         else{
 278                                 inOligos >> oligo;
 279
 280                                 for(int i=0;i<oligo.length();i++){
 281                                         oligo[i] = toupper(oligo[i]);
 282                                         if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 283                                 }
 284
 285                                 if(type == "forward"){
 286                                         forPrimer.push_back(oligo);
 287                                 }
 288                                 else if(type == "reverse"){
 289                                         revPrimer.push_back(oligo);
 290                                 }
 291                                 else if(type == "barcode"){
 292                                         inOligos >> group;
 293                                         barcodes[oligo]=index++;
 294                                         groupVector.push_back(group);
 295
 296                                         if(allFiles){
 297                                                 outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 298                                         }
 299                                 }
 300                         }
 301                 }
 302
 303                 inOligos.close();
 304
 305                 numFPrimers = forPrimer.size();
 306                 numRPrimers = revPrimer.size();
 307
 308         }
 309         catch(exception& e) {
 310                 errorOut(e, "TrimSeqsCommand", "getOligos");
 311                 exit(1);
 312         }
 313
 314 }
 315
 316 //***************************************************************************************************************
 317
 318 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 319         try {
 320                 string rawSequence = seq.getUnaligned();
 321                 bool success = 0;       //guilty until proven innocent
 322
 323                 for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 324                         string oligo = it->first;
 325                         if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 326                                 success = 0;
 327                                 break;
 328                         }
 329
 330                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 331                                 group = it->second;
 332                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 333                                 success = 1;
 334                                 break;
 335                         }
 336                 }
 337                 return success;
 338
 339         }
 340         catch(exception& e) {
 341                 errorOut(e, "TrimSeqsCommand", "stripBarcode");
 342                 exit(1);
 343         }
 344
 345 }
 346
 347 //***************************************************************************************************************
 348
 349 bool TrimSeqsCommand::stripForward(Sequence& seq){
 350         try {
 351                 string rawSequence = seq.getUnaligned();
 352                 bool success = 0;       //guilty until proven innocent
 353
 354                 for(int i=0;i<numFPrimers;i++){
 355                         string oligo = forPrimer[i];
 356
 357                         if(rawSequence.length() < oligo.length()){
 358                                 success = 0;
 359                                 break;
 360                         }
 361
 362                         if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 363                                 seq.setUnaligned(rawSequence.substr(oligo.length()));
 364                                 success = 1;
 365                                 break;
 366                         }
 367                 }
 368
 369                 return success;
 370
 371         }
 372         catch(exception& e) {
 373                 errorOut(e, "TrimSeqsCommand", "stripForward");
 374                 exit(1);
 375         }
 376 }
 377
 378 //***************************************************************************************************************
 379
 380 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 381         try {
 382                 string rawSequence = seq.getUnaligned();
 383                 bool success = 0;       //guilty until proven innocent
 384
 385                 for(int i=0;i<numRPrimers;i++){
 386                         string oligo = revPrimer[i];
 387
 388                         if(rawSequence.length() < oligo.length()){
 389                                 success = 0;
 390                                 break;
 391                         }
 392
 393                         if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 394                                 seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 395                                 success = 1;
 396                                 break;
 397                         }
 398                 }
 399                 return success;
 400
 401         }
 402         catch(exception& e) {
 403                 errorOut(e, "TrimSeqsCommand", "stripReverse");
 404                 exit(1);
 405         }
 406 }
 407
 408 //***************************************************************************************************************
 409
 410 bool TrimSeqsCommand::cullLength(Sequence& seq){
 411         try {
 412
 413                 int length = seq.getNumBases();
 414                 bool success = 0;       //guilty until proven innocent
 415
 416                 if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 417                 else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 418                 else                                                                                            {       success = 0;    }
 419
 420                 return success;
 421
 422         }
 423         catch(exception& e) {
 424                 errorOut(e, "TrimSeqsCommand", "cullLength");
 425                 exit(1);
 426         }
 427
 428 }
 429
 430 //***************************************************************************************************************
 431
 432 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 433         try {
 434                 int longHomoP = seq.getLongHomoPolymer();
 435                 bool success = 0;       //guilty until proven innocent
 436
 437                 if(longHomoP <= maxHomoP){      success = 1;    }
 438                 else                                    {       success = 0;    }
 439
 440                 return success;
 441         }
 442         catch(exception& e) {
 443                 errorOut(e, "TrimSeqsCommand", "cullHomoP");
 444                 exit(1);
 445         }
 446
 447 }
 448
 449 //***************************************************************************************************************
 450
 451 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 452         try {
 453                 int numNs = seq.getAmbigBases();
 454                 bool success = 0;       //guilty until proven innocent
 455
 456                 if(numNs <= maxAmbig)   {       success = 1;    }
 457                 else                                    {       success = 0;    }
 458
 459                 return success;
 460         }
 461         catch(exception& e) {
 462                 errorOut(e, "TrimSeqsCommand", "cullAmbigs");
 463                 exit(1);
 464         }
 465
 466 }
 467
 468 //***************************************************************************************************************
 469
 470 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 471         try {
 472                 bool success = 1;
 473                 int length = oligo.length();
 474
 475                 for(int i=0;i<length;i++){
 476
 477                         if(oligo[i] != seq[i]){
 478                                 if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 479                                 else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 480                                 else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 481                                 else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 482                                 else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 483                                 else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 484                                 else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 485                                 else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 486                                 else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 487                                 else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 488                                 else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 489                                 else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 490
 491                                 if(success == 0)        {       break;  }
 492                         }
 493                         else{
 494                                 success = 1;
 495                         }
 496                 }
 497
 498                 return success;
 499         }
 500         catch(exception& e) {
 501                 errorOut(e, "TrimSeqsCommand", "compareDNASeq");
 502                 exit(1);
 503         }
 504
 505 }
 506
 507 //***************************************************************************************************************
 508
 509 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 510         try {
 511                 string rawSequence = seq.getUnaligned();
 512                 int seqLength = rawSequence.length();
 513                 string name;
 514
 515                 qFile >> name;
 516                 if (name.length() != 0) {  if(name.substr(1) != seq.getName())  {       mothurOut("sequence name mismatch btwn fasta and qual file"); mothurOutEndLine();       }  }
 517                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 518
 519                 int score;
 520                 int end = seqLength;
 521
 522                 for(int i=0;i<seqLength;i++){
 523                         qFile >> score;
 524
 525                         if(score <= qThreshold){
 526                                 end = i;
 527                                 break;
 528                         }
 529                 }
 530                 for(int i=end+1;i<seqLength;i++){
 531                         qFile >> score;
 532                 }
 533
 534                 seq.setUnaligned(rawSequence.substr(0,end));
 535
 536                 return 1;
 537         }
 538         catch(exception& e) {
 539                 errorOut(e, "TrimSeqsCommand", "stripQualThreshold");
 540                 exit(1);
 541         }
 542 }
 543
 544 //***************************************************************************************************************
 545
 546 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 547         try {
 548                 string rawSequence = seq.getUnaligned();
 549                 int seqLength = seq.getNumBases();
 550                 bool success = 0;       //guilty until proven innocent
 551                 string name;
 552
 553                 qFile >> name;
 554                 if (name[0] == '>') {  if(name.substr(1) != seq.getName())      {       mothurOut("sequence name mismatch btwn fasta: " + seq.getName() + " and qual file: " + name); mothurOutEndLine();       } }
 555
 556                 while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 557
 558                 float score;
 559                 float average = 0;
 560
 561                 for(int i=0;i<seqLength;i++){
 562                         qFile >> score;
 563                         average += score;
 564                 }
 565                 average /= seqLength;
 566
 567                 if(average >= qAverage) {       success = 1;    }
 568                 else                                    {       success = 0;    }
 569
 570                 return success;
 571         }
 572         catch(exception& e) {
 573                 errorOut(e, "TrimSeqsCommand", "cullQualAverage");
 574                 exit(1);
 575         }
 576 }
 577
 578 //***************************************************************************************************************