trimseqscommand.cpp

   1 /*
   2  *  trimseqscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Pat Schloss on 6/6/09.
   6  *  Copyright 2009 Patrick D. Schloss. All rights reserved.
   7  *
   8  */
   9
  10 #include "trimseqscommand.h"
  11
  12 //***************************************************************************************************************
  13
  14 TrimSeqsCommand::TrimSeqsCommand(string option){
  15         try {
  16
  17                 abort = false;
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string AlignArray[] =  {"fasta", "flip", "oligos", "maxambig", "maxhomop", "minlength", "maxlength", "qfile", "qthreshold", "qaverage", "allfiles"};
  25
  26                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  27
  28                         OptionParser parser(option);
  29                         map<string,string> parameters = parser.getParameters();
  30
  31                         ValidParameters validParameter;
  32
  33                         //check to make sure all parameters are valid for command
  34                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  35                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  36                         }
  37
  38                         //check for required parameters
  39                         fastaFile = validParameter.validFile(parameters, "fasta", true);
  40                         if (fastaFile == "not found") { cout << "fasta is a required parameter for the screen.seqs command." << endl; abort = true; }
  41                         else if (fastaFile == "not open") { abort = true; }
  42
  43
  44                         //check for optional parameter and set defaults
  45                         // ...at some point should added some additional type checking...
  46                         string temp;
  47                         temp = validParameter.validFile(parameters, "flip", false);
  48                         if (temp == "not found"){       flip = 0;       }
  49                         else if(isTrue(temp))   {       flip = 1;       }
  50
  51                         temp = validParameter.validFile(parameters, "oligos", true);
  52                         if (temp == "not found"){       oligoFile = "";         }
  53                         else if(temp == "not open"){    abort = true;   }
  54                         else                                    {       oligoFile = temp;               }
  55
  56                         temp = validParameter.validFile(parameters, "maxambig", false);         if (temp == "not found") { temp = "-1"; }
  57                         convert(temp, maxAmbig);
  58
  59                         temp = validParameter.validFile(parameters, "maxhomop", false);         if (temp == "not found") { temp = "0"; }
  60                         convert(temp, maxHomoP);
  61
  62                         temp = validParameter.validFile(parameters, "minlength", false);        if (temp == "not found") { temp = "0"; }
  63                         convert(temp, minLength);
  64
  65                         temp = validParameter.validFile(parameters, "maxlength", false);        if (temp == "not found") { temp = "0"; }
  66                         convert(temp, maxLength);
  67
  68                         temp = validParameter.validFile(parameters, "qfile", true);
  69                         if (temp == "not found")        {       qFileName = "";         }
  70                         else if(temp == "not open")     {       abort = 0;              }
  71                         else                                            {       qFileName = temp;       }
  72
  73                         temp = validParameter.validFile(parameters, "qthreshold", false);       if (temp == "not found") { temp = "0"; }
  74                         convert(temp, qThreshold);
  75
  76                         temp = validParameter.validFile(parameters, "qaverage", false);         if (temp == "not found") { temp = "0"; }
  77                         convert(temp, qAverage);
  78
  79                         temp = validParameter.validFile(parameters, "allfiles", false);         if (temp == "not found") { temp = "F"; }
  80                         allFiles = isTrue(temp);
  81
  82                         if(allFiles && oligoFile == ""){
  83                                 cout << "You selected allfiles, but didn't enter an oligos file.  Ignoring the allfiles request." << endl;
  84                         }
  85                         if((qAverage != 0 && qThreshold != 0) && qFileName == ""){
  86                                 cout << "You didn't provide a quality file name, quality criteria will be ignored." << endl;
  87                                 qAverage=0;
  88                                 qThreshold=0;
  89                         }
  90                         if(!flip && oligoFile=="" && !maxLength && !minLength && (maxAmbig==-1) && !maxHomoP && qFileName == ""){
  91                                 cout << "You didn't set any options... quiting command." << endl;
  92                                 abort = true;
  93                         }
  94                 }
  95
  96         }
  97         catch(exception& e) {
  98                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
  99                 exit(1);
 100         }
 101         catch(...) {
 102                 cout << "An unknown error has occurred in the TrimSeqsCommand class function TrimSeqsCommand. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 103                 exit(1);
 104         }
 105 }
 106 //**********************************************************************************************************************
 107
 108 void TrimSeqsCommand::help(){
 109         try {
 110                 cout << "The trim.seqs command reads a fastaFile and creates ....." << "\n";
 111                 cout << "The trim.seqs command parameters are fasta, flip, oligos, maxambig, maxhomop, minlength and maxlength." << "\n";
 112                 cout << "The fasta parameter is required." << "\n";
 113                 cout << "The flip parameter .... The default is 0." << "\n";
 114                 cout << "The oligos parameter .... The default is ""." << "\n";
 115                 cout << "The maxambig parameter .... The default is -1." << "\n";
 116                 cout << "The maxhomop parameter .... The default is 0." << "\n";
 117                 cout << "The minlength parameter .... The default is 0." << "\n";
 118                 cout << "The maxlength parameter .... The default is 0." << "\n";
 119                 cout << "The trim.seqs command should be in the following format: " << "\n";
 120                 cout << "trim.seqs(fasta=yourFastaFile, flip=yourFlip, oligos=yourOligos, maxambig=yourMaxambig,  " << "\n";
 121                 cout << "maxhomop=yourMaxhomop, minlength=youMinlength, maxlength=yourMaxlength)  " << "\n";
 122                 cout << "Example trim.seqs(fasta=abrecovery.fasta, flip=..., oligos=..., maxambig=..., maxhomop=..., minlength=..., maxlength=...)." << "\n";
 123                 cout << "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta)." << "\n" << "\n";
 124
 125         }
 126         catch(exception& e) {
 127                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 128                 exit(1);
 129         }
 130         catch(...) {
 131                 cout << "An unknown error has occurred in the TrimSeqsCommand class function help. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 132                 exit(1);
 133         }
 134 }
 135
 136
 137 //***************************************************************************************************************
 138
 139 TrimSeqsCommand::~TrimSeqsCommand(){    /*      do nothing      */      }
 140
 141 //***************************************************************************************************************
 142
 143 int TrimSeqsCommand::execute(){
 144         try{
 145
 146                 if (abort == true) { return 0; }
 147
 148                 ifstream inFASTA;
 149                 openInputFile(fastaFile, inFASTA);
 150
 151                 ofstream outFASTA;
 152                 string trimSeqFile = getRootName(fastaFile) + "trim.fasta";
 153                 openOutputFile(trimSeqFile, outFASTA);
 154
 155                 ofstream outGroups;
 156                 vector<ofstream*> fastaFileNames;
 157                 if(oligoFile != ""){
 158                         string groupFile = getRootName(fastaFile) + "groups";
 159                         openOutputFile(groupFile, outGroups);
 160                         getOligos(fastaFileNames);
 161                 }
 162
 163                 ofstream scrapFASTA;
 164                 string scrapSeqFile = getRootName(fastaFile) + "scrap.fasta";
 165                 openOutputFile(scrapSeqFile, scrapFASTA);
 166
 167                 ifstream qFile;
 168                 if(qFileName != "")     {       openInputFile(qFileName, qFile);        }
 169
 170                 bool success;
 171
 172                 while(!inFASTA.eof()){
 173                         Sequence currSeq(inFASTA);
 174                         string origSeq = currSeq.getUnaligned();
 175                         int group;
 176                         string trashCode = "";
 177
 178                         if(qFileName != ""){
 179                                 if(qThreshold != 0)             {       success = stripQualThreshold(currSeq, qFile);   }
 180                                 else if(qAverage != 0)  {       success = cullQualAverage(currSeq, qFile);              }
 181                                 if(!success)                    {       trashCode += 'q';                                                               }
 182                                 qFile.close();
 183                         }
 184                         if(barcodes.size() != 0){
 185                                 success = stripBarcode(currSeq, group);
 186                                 if(!success){   trashCode += 'b';       }
 187                         }
 188                         if(numFPrimers != 0){
 189                                 success = stripForward(currSeq);
 190                                 if(!success){   trashCode += 'f';       }
 191                         }
 192                         if(numRPrimers != 0){
 193                                 success = stripReverse(currSeq);
 194                                 if(!success){   trashCode += 'r';       }
 195                         }
 196                         if(minLength > 0 || maxLength > 0){
 197                                 success = cullLength(currSeq);
 198                                 if(!success){   trashCode += 'l';       }
 199                         }
 200                         if(maxHomoP > 0){
 201                                 success = cullHomoP(currSeq);
 202                                 if(!success){   trashCode += 'h';       }
 203                         }
 204                         if(maxAmbig != -1){
 205                                 success = cullAmbigs(currSeq);
 206                                 if(!success){   trashCode += 'n';       }
 207                         }
 208
 209                         if(flip){       currSeq.reverseComplement();    }               // should go last
 210
 211                         if(trashCode.length() == 0){
 212                                 currSeq.printSequence(outFASTA);
 213                                 if(barcodes.size() != 0){
 214                                         outGroups << currSeq.getName() << '\t' << groupVector[group] << endl;
 215
 216                                         if(allFiles){
 217                                                 currSeq.printSequence(*fastaFileNames[group]);
 218                                         }
 219                                 }
 220                         }
 221                         else{
 222                                 currSeq.setName(currSeq.getName() + '|' + trashCode);
 223                                 currSeq.setUnaligned(origSeq);
 224                                 currSeq.printSequence(scrapFASTA);
 225                         }
 226                         gobble(inFASTA);
 227                 }
 228                 inFASTA.close();
 229                 outFASTA.close();
 230                 scrapFASTA.close();
 231                 outGroups.close();
 232
 233                 for(int i=0;i<fastaFileNames.size();i++){
 234                         fastaFileNames[i]->close();
 235                         delete fastaFileNames[i];
 236                 }
 237
 238                 for(int i=0;i<fastaFileNames.size();i++){
 239                         string seqName;
 240                         openInputFile(getRootName(fastaFile) + groupVector[i] + ".fasta", inFASTA);
 241                         ofstream outGroups;
 242                         openOutputFile(getRootName(fastaFile) + groupVector[i] + ".groups", outGroups);
 243
 244                         while(!inFASTA.eof()){
 245                                 if(inFASTA.get() == '>'){
 246                                         inFASTA >> seqName;
 247                                         outGroups << seqName << '\t' << groupVector[i] << endl;
 248                                 }
 249                                 while (!inFASTA.eof())  {       char c = inFASTA.get(); if (c == 10 || c == 13){        break;  }       }
 250                         }
 251                         outGroups.close();
 252                         inFASTA.close();
 253                 }
 254
 255
 256                 return 0;
 257         }
 258         catch(exception& e) {
 259                 cout << "Standard Error: " << e.what() << " has occurred in the TrimSeqsCommand class Function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 260                 exit(1);
 261         }
 262         catch(...) {
 263                 cout << "An unknown error has occurred in the TrimSeqsCommand class function execute. Please contact Pat Schloss at pschloss@microbio.umass.edu." << "\n";
 264                 exit(1);
 265         }
 266 }
 267
 268 //***************************************************************************************************************
 269
 270 void TrimSeqsCommand::getOligos(vector<ofstream*>& outFASTAVec){
 271
 272         ifstream inOligos;
 273         openInputFile(oligoFile, inOligos);
 274
 275         ofstream test;
 276
 277         string type, oligo, group;
 278         int index=0;
 279
 280         while(!inOligos.eof()){
 281                 inOligos >> type;
 282
 283                 if(type[0] == '#'){
 284                         while (!inOligos.eof()) {       char c = inOligos.get(); if (c == 10 || c == 13){       break;  }       } // get rest of line if there's any crap there
 285                 }
 286                 else{
 287                         inOligos >> oligo;
 288
 289                         for(int i=0;i<oligo.length();i++){
 290                                 oligo[i] = toupper(oligo[i]);
 291                                 if(oligo[i] == 'U')     {       oligo[i] = 'T'; }
 292                         }
 293
 294                         if(type == "forward"){
 295                                 forPrimer.push_back(oligo);
 296                         }
 297                         else if(type == "reverse"){
 298                                 revPrimer.push_back(oligo);
 299                         }
 300                         else if(type == "barcode"){
 301                                 inOligos >> group;
 302                                 barcodes[oligo]=index++;
 303                                 groupVector.push_back(group);
 304
 305                                 if(allFiles){
 306                                         outFASTAVec.push_back(new ofstream((getRootName(fastaFile) + group + ".fasta").c_str(), ios::ate));
 307                                 }
 308                         }
 309                 }
 310         }
 311
 312         inOligos.close();
 313
 314         numFPrimers = forPrimer.size();
 315         numRPrimers = revPrimer.size();
 316 }
 317
 318 //***************************************************************************************************************
 319
 320 bool TrimSeqsCommand::stripBarcode(Sequence& seq, int& group){
 321
 322         string rawSequence = seq.getUnaligned();
 323         bool success = 0;       //guilty until proven innocent
 324
 325         for(map<string,int>::iterator it=barcodes.begin();it!=barcodes.end();it++){
 326                 string oligo = it->first;
 327                 if(rawSequence.length() < oligo.length()){      //let's just assume that the barcodes are the same length
 328                         success = 0;
 329                         break;
 330                 }
 331
 332                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 333                         group = it->second;
 334                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 335                         success = 1;
 336                         break;
 337                 }
 338         }
 339         return success;
 340
 341 }
 342
 343 //***************************************************************************************************************
 344
 345 bool TrimSeqsCommand::stripForward(Sequence& seq){
 346
 347         string rawSequence = seq.getUnaligned();
 348         bool success = 0;       //guilty until proven innocent
 349
 350         for(int i=0;i<numFPrimers;i++){
 351                 string oligo = forPrimer[i];
 352
 353                 if(rawSequence.length() < oligo.length()){
 354                         success = 0;
 355                         break;
 356                 }
 357
 358                 if(compareDNASeq(oligo, rawSequence.substr(0,oligo.length()))){
 359                         seq.setUnaligned(rawSequence.substr(oligo.length()));
 360                         success = 1;
 361                         break;
 362                 }
 363         }
 364
 365         return success;
 366
 367 }
 368
 369 //***************************************************************************************************************
 370
 371 bool TrimSeqsCommand::stripReverse(Sequence& seq){
 372
 373         string rawSequence = seq.getUnaligned();
 374         bool success = 0;       //guilty until proven innocent
 375
 376         for(int i=0;i<numRPrimers;i++){
 377                 string oligo = revPrimer[i];
 378
 379                 if(rawSequence.length() < oligo.length()){
 380                         success = 0;
 381                         break;
 382                 }
 383
 384                 if(compareDNASeq(oligo, rawSequence.substr(rawSequence.length()-oligo.length(),oligo.length()))){
 385                         seq.setUnaligned(rawSequence.substr(rawSequence.length()-oligo.length()));
 386                         success = 1;
 387                         break;
 388                 }
 389         }
 390         return success;
 391
 392 }
 393
 394 //***************************************************************************************************************
 395
 396 bool TrimSeqsCommand::cullLength(Sequence& seq){
 397
 398         int length = seq.getNumBases();
 399         bool success = 0;       //guilty until proven innocent
 400
 401         if(length >= minLength && maxLength == 0)                       {       success = 1;    }
 402         else if(length >= minLength && length <= maxLength)     {       success = 1;    }
 403         else                                                                                            {       success = 0;    }
 404
 405         return success;
 406
 407 }
 408
 409 //***************************************************************************************************************
 410
 411 bool TrimSeqsCommand::cullHomoP(Sequence& seq){
 412
 413         int longHomoP = seq.getLongHomoPolymer();
 414         bool success = 0;       //guilty until proven innocent
 415
 416         if(longHomoP <= maxHomoP){      success = 1;    }
 417         else                                    {       success = 0;    }
 418
 419         return success;
 420
 421 }
 422
 423 //***************************************************************************************************************
 424
 425 bool TrimSeqsCommand::cullAmbigs(Sequence& seq){
 426
 427         int numNs = seq.getAmbigBases();
 428         bool success = 0;       //guilty until proven innocent
 429
 430         if(numNs <= maxAmbig)   {       success = 1;    }
 431         else                                    {       success = 0;    }
 432
 433         return success;
 434
 435 }
 436
 437 //***************************************************************************************************************
 438
 439 bool TrimSeqsCommand::compareDNASeq(string oligo, string seq){
 440
 441         bool success = 1;
 442         int length = oligo.length();
 443
 444         for(int i=0;i<length;i++){
 445
 446                 if(oligo[i] != seq[i]){
 447                         if(oligo[i] == 'A' || oligo[i] == 'T' || oligo[i] == 'G' || oligo[i] == 'C')    {       success = 0;    }
 448                         else if((oligo[i] == 'N' || oligo[i] == 'I') && (seq[i] == 'N'))                                {       success = 0;    }
 449                         else if(oligo[i] == 'R' && (seq[i] != 'A' && seq[i] != 'G'))                                    {       success = 0;    }
 450                         else if(oligo[i] == 'Y' && (seq[i] != 'C' && seq[i] != 'T'))                                    {       success = 0;    }
 451                         else if(oligo[i] == 'M' && (seq[i] != 'C' && seq[i] != 'A'))                                    {       success = 0;    }
 452                         else if(oligo[i] == 'K' && (seq[i] != 'T' && seq[i] != 'G'))                                    {       success = 0;    }
 453                         else if(oligo[i] == 'W' && (seq[i] != 'T' && seq[i] != 'A'))                                    {       success = 0;    }
 454                         else if(oligo[i] == 'S' && (seq[i] != 'C' && seq[i] != 'G'))                                    {       success = 0;    }
 455                         else if(oligo[i] == 'B' && (seq[i] != 'C' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 456                         else if(oligo[i] == 'D' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'G'))   {       success = 0;    }
 457                         else if(oligo[i] == 'H' && (seq[i] != 'A' && seq[i] != 'T' && seq[i] != 'C'))   {       success = 0;    }
 458                         else if(oligo[i] == 'V' && (seq[i] != 'A' && seq[i] != 'C' && seq[i] != 'G'))   {       success = 0;    }
 459
 460                         if(success == 0)        {       break;  }
 461                 }
 462                 else{
 463                         success = 1;
 464                 }
 465         }
 466
 467         return success;
 468 }
 469
 470 //***************************************************************************************************************
 471
 472 bool TrimSeqsCommand::stripQualThreshold(Sequence& seq, ifstream& qFile){
 473
 474         string rawSequence = seq.getUnaligned();
 475         int seqLength = rawSequence.length();
 476         string name;
 477
 478         qFile >> name;
 479         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 480         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 481
 482         int score;
 483         int end = seqLength;
 484
 485         for(int i=0;i<seqLength;i++){
 486                 qFile >> score;
 487
 488                 if(score <= qThreshold){
 489                         end = i;
 490                         break;
 491                 }
 492         }
 493         for(int i=end+1;i<seqLength;i++){
 494                 qFile >> score;
 495         }
 496
 497         seq.setUnaligned(rawSequence.substr(0,end));
 498
 499         return 1;
 500 }
 501
 502 //***************************************************************************************************************
 503
 504 bool TrimSeqsCommand::cullQualAverage(Sequence& seq, ifstream& qFile){
 505
 506         string rawSequence = seq.getUnaligned();
 507         int seqLength = seq.getNumBases();
 508         bool success = 0;       //guilty until proven innocent
 509         string name;
 510
 511         qFile >> name;
 512         if(name.substr(1) != seq.getName())     {       cout << "sequence name mismatch btwn fasta and qual file" << endl;      }
 513         while (!qFile.eof())    {       char c = qFile.get(); if (c == 10 || c == 13){  break;  }       }
 514
 515         float score;
 516         float average = 0;
 517
 518         for(int i=0;i<seqLength;i++){
 519                 qFile >> score;
 520                 average += score;
 521         }
 522         average /= seqLength;
 523
 524         if(average >= qAverage) {       success = 1;    }
 525         else                                    {       success = 0;    }
 526
 527         return success;
 528 }
 529
 530 //***************************************************************************************************************
 531
 532