aligncommand.cpp

   1 /*
   2  *  aligncommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Sarah Westcott on 5/15/09.
   6  *  Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
   7  *
   8  *      This version of nast does everything I think that the greengenes nast server does and then some.  I have added the
   9  *      feature of allowing users to define their database, kmer size for searching, alignment penalty values and alignment
  10  *      method.  This latter feature is perhaps most significant.  nastPlus enables a user to use either a Needleman-Wunsch
  11  *      (non-affine gap penalty) or Gotoh (affine gap penalty) pairwise alignment algorithm.  This is significant because it
  12  *      allows for a global alignment and not the local alignment provided by bLAst.  Furthermore, it has the potential to
  13  *      provide a better alignment because of the banding method employed by blast (I'm not sure about this).
  14  *
  15  */
  16
  17 #include "aligncommand.h"
  18 #include "sequence.hpp"
  19
  20 #include "gotohoverlap.hpp"
  21 #include "needlemanoverlap.hpp"
  22 #include "blastalign.hpp"
  23 #include "noalign.hpp"
  24
  25 #include "nast.hpp"
  26 #include "nastreport.hpp"
  27
  28
  29 //**********************************************************************************************************************
  30
  31 AlignCommand::AlignCommand(string option){
  32         try {
  33                 //              globaldata = GlobalData::getInstance();
  34                 abort = false;
  35
  36                 //allow user to run help
  37                 if(option == "help") { help(); abort = true; }
  38
  39                 else {
  40
  41                         //valid paramters for this command
  42                         string AlignArray[] =  {"template","candidate","search","ksize","align","match","mismatch","gapopen","gapextend", "processors","flip","threshold"};
  43                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  44
  45                         OptionParser parser(option);
  46                         map<string, string> parameters = parser.getParameters();
  47
  48                         ValidParameters validParameter;
  49
  50                         //check to make sure all parameters are valid for command
  51                         for (map<string, string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  52                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  53                         }
  54
  55                         //check for required parameters
  56                         templateFileName = validParameter.validFile(parameters, "template", true);
  57                         if (templateFileName == "not found") {
  58                                 mothurOut("template is a required parameter for the align.seqs command.");
  59                                 mothurOutEndLine();
  60                                 abort = true;
  61                         }
  62                         else if (templateFileName == "not open") { abort = true; }
  63
  64                         candidateFileName = validParameter.validFile(parameters, "candidate", true);
  65                         if (candidateFileName == "not found") {
  66                                 mothurOut("candidate is a required parameter for the align.seqs command.");
  67                                 mothurOutEndLine();
  68                                 abort = true;
  69                         }
  70                         else if (candidateFileName == "not open") { abort = true; }
  71
  72                         //check for optional parameter and set defaults
  73                         // ...at some point should added some additional type checking...
  74                         string temp;
  75                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
  76                         convert(temp, kmerSize);
  77
  78                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
  79                         convert(temp, match);
  80
  81                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
  82                         convert(temp, misMatch);
  83
  84                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
  85                         convert(temp, gapOpen);
  86
  87                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
  88                         convert(temp, gapExtend);
  89
  90                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
  91                         convert(temp, processors);
  92
  93                         temp = validParameter.validFile(parameters, "flip", false);                     if (temp == "not found"){       temp = "f";                             }
  94                         flip = isTrue(temp);
  95
  96                         temp = validParameter.validFile(parameters, "threshold", false);        if (temp == "not found"){       temp = "0.50";                  }
  97                         convert(temp, threshold);
  98
  99                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
 100
 101                         align = validParameter.validFile(parameters, "align", false);           if (align == "not found"){      align = "needleman";    }
 102                 }
 103
 104         }
 105         catch(exception& e) {
 106                 errorOut(e, "AlignCommand", "AlignCommand");
 107                 exit(1);
 108         }
 109 }
 110
 111 //**********************************************************************************************************************
 112
 113 AlignCommand::~AlignCommand(){
 114
 115         if (abort == false) {
 116                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 117                 delete templateDB;
 118                 delete alignment;
 119         }
 120 }
 121
 122 //**********************************************************************************************************************
 123
 124 void AlignCommand::help(){
 125         try {
 126                 mothurOut("The align.seqs command reads a file containing sequences and creates an alignment file and a report file.\n");
 127                 mothurOut("The align.seqs command parameters are template, candidate, search, ksize, align, match, mismatch, gapopen and gapextend.\n");
 128                 mothurOut("The template and candidate parameters are required.\n");
 129                 mothurOut("The search parameter allows you to specify the method to find most similar template.  Your options are: suffix, kmer and blast. The default is kmer.\n");
 130                 mothurOut("The align parameter allows you to specify the alignment method to use.  Your options are: gotoh, needleman, blast and noalign. The default is needleman.\n");
 131                 mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 132                 mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 133                 mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 134                 mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -2.0.\n");
 135                 mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -1.0.\n");
 136                 mothurOut("The flip parameter is used to specify whether or not you want mothur to try the reverse compement if a sequence falls below the threshold.  The default is false.\n");
 137                 mothurOut("The threshold is used to specify a cutoff at which an alignment is deemed 'bad' and the reverse complement may be tried. \n");
 138                 mothurOut("If the flip parameter is set to true the reverse complement of the sequence is aligned and the better alignment is reported.\n");
 139                 mothurOut("The default for the threshold parameter is 0.50, meaning at least 50% of the bases must remain or the sequence is reported as potentially reversed.\n");
 140                 mothurOut("The align.seqs command should be in the following format: \n");
 141                 mothurOut("align.seqs(template=yourTemplateFile, candidate=yourCandidateFile, align=yourAlignmentMethod, search=yourSearchmethod, ksize=yourKmerSize, match=yourMatchBonus, mismatch=yourMismatchpenalty, gapopen=yourGapopenPenalty, gapextend=yourGapExtendPenalty) \n");
 142                 mothurOut("Example align.seqs(candidate=candidate.fasta, template=core.filtered, align=kmer, search=gotoh, ksize=8, match=2.0, mismatch=3.0, gapopen=-2.0, gapextend=-1.0)\n");
 143                 mothurOut("Note: No spaces between parameter labels (i.e. candidate), '=' and parameters (i.e.yourFastaFile).\n\n");
 144         }
 145         catch(exception& e) {
 146                 errorOut(e, "AlignCommand", "help");
 147                 exit(1);
 148         }
 149 }
 150
 151
 152 //**********************************************************************************************************************
 153
 154 int AlignCommand::execute(){
 155         try {
 156                 if (abort == true) {    return 0;       }
 157
 158                 templateDB = new AlignmentDB(templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch);
 159                 int longestBase = templateDB->getLongestBase();
 160
 161                 if(align == "gotoh")                    {       alignment = new GotohOverlap(gapOpen, gapExtend, match, misMatch, longestBase);                 }
 162                 else if(align == "needleman")   {       alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase);                                }
 163                 else if(align == "blast")               {       alignment = new BlastAlignment(gapOpen, gapExtend, match, misMatch);            }
 164                 else if(align == "noalign")             {       alignment = new NoAlign();                                                                                                      }
 165                 else {
 166                         mothurOut(align + " is not a valid alignment option. I will run the command using needleman.");
 167                         mothurOutEndLine();
 168                         alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase);
 169                 }
 170                 mothurOut("Aligning sequences...");
 171                 mothurOutEndLine();
 172
 173                 string alignFileName = candidateFileName.substr(0,candidateFileName.find_last_of(".")+1) + "align";
 174                 string reportFileName = candidateFileName.substr(0,candidateFileName.find_last_of(".")+1) + "align.report";
 175                 string accnosFileName = candidateFileName.substr(0,candidateFileName.find_last_of(".")+1) + "flip.accnos";
 176
 177                 int numFastaSeqs = 0;
 178                 int start = time(NULL);
 179
 180 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 181                 if(processors == 1){
 182                         ifstream inFASTA;
 183                         openInputFile(candidateFileName, inFASTA);
 184                         numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 185                         inFASTA.close();
 186
 187                         lines.push_back(new linePair(0, numFastaSeqs));
 188
 189                         driver(lines[0], alignFileName, reportFileName, accnosFileName);
 190
 191                         //delete accnos file if its blank else report to user
 192                         if (isBlank(accnosFileName)) {  remove(accnosFileName.c_str());  }
 193                         else {
 194                                 mothurOut("Some of you sequences generated alignments that eliminated too many bases, a list is provided in " + accnosFileName + ".");
 195                                 if (!flip) {
 196                                         mothurOut(" If you set the flip parameter to true mothur will try aligning the reverse compliment as well.");
 197                                 }else{  mothurOut(" If the reverse compliment proved to be better it was reported.");  }
 198                                 mothurOutEndLine();
 199                         }
 200                 }
 201                 else{
 202                         vector<int> positions;
 203                         processIDS.resize(0);
 204
 205                         ifstream inFASTA;
 206                         openInputFile(candidateFileName, inFASTA);
 207
 208                         string input;
 209                         while(!inFASTA.eof()){
 210                                 input = getline(inFASTA);
 211                                 if (input.length() != 0) {
 212                                         if(input[0] == '>'){    long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1);  }
 213                                 }
 214                         }
 215                         inFASTA.close();
 216
 217                         numFastaSeqs = positions.size();
 218
 219                         int numSeqsPerProcessor = numFastaSeqs / processors;
 220
 221                         for (int i = 0; i < processors; i++) {
 222                                 long int startPos = positions[ i * numSeqsPerProcessor ];
 223                                 if(i == processors - 1){
 224                                         numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
 225                                 }
 226                                 lines.push_back(new linePair(startPos, numSeqsPerProcessor));
 227                         }
 228
 229                         createProcesses(alignFileName, reportFileName, accnosFileName);
 230
 231                         rename((alignFileName + toString(processIDS[0]) + ".temp").c_str(), alignFileName.c_str());
 232                         rename((reportFileName + toString(processIDS[0]) + ".temp").c_str(), reportFileName.c_str());
 233
 234                         //append alignment and report files
 235                         for(int i=1;i<processors;i++){
 236                                 appendAlignFiles((alignFileName + toString(processIDS[i]) + ".temp"), alignFileName);
 237                                 remove((alignFileName + toString(processIDS[i]) + ".temp").c_str());
 238
 239                                 appendReportFiles((reportFileName + toString(processIDS[i]) + ".temp"), reportFileName);
 240                                 remove((reportFileName + toString(processIDS[i]) + ".temp").c_str());
 241                         }
 242
 243                         vector<string> nonBlankAccnosFiles;
 244                         //delete blank accnos files generated with multiple processes
 245                         for(int i=0;i<processors;i++){
 246                                 if (!(isBlank(accnosFileName + toString(processIDS[i]) + ".temp"))) {
 247                                         nonBlankAccnosFiles.push_back(accnosFileName + toString(processIDS[i]) + ".temp");
 248                                 }else { remove((accnosFileName + toString(processIDS[i]) + ".temp").c_str());  }
 249                         }
 250
 251                         //append accnos files
 252                         if (nonBlankAccnosFiles.size() != 0) {
 253                                 rename(nonBlankAccnosFiles[0].c_str(), accnosFileName.c_str());
 254
 255                                 for (int h=1; h < nonBlankAccnosFiles.size(); h++) {
 256                                         appendAlignFiles(nonBlankAccnosFiles[h], accnosFileName);
 257                                         remove(nonBlankAccnosFiles[h].c_str());
 258                                 }
 259                                 mothurOut("Some of you sequences generated alignments that eliminated too many bases, a list is provided in " + accnosFileName + ".");
 260                                 if (!flip) {
 261                                         mothurOut(" If you set the flip parameter to true mothur will try aligning the reverse compliment as well.");
 262                                 }else{  mothurOut(" If the reverse compliment proved to be better it was reported.");  }
 263                                 mothurOutEndLine();
 264                         }
 265                 }
 266 #else
 267                 ifstream inFASTA;
 268                 openInputFile(candidateFileName, inFASTA);
 269                 numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 270                 inFASTA.close();
 271
 272                 lines.push_back(new linePair(0, numFastaSeqs));
 273
 274                 driver(lines[0], alignFileName, reportFileName, accnosFileName);
 275
 276                 //delete accnos file if its blank else report to user
 277                 if (isBlank(accnosFileName)) {  remove(accnosFileName.c_str());  }
 278                 else {
 279                         mothurOut("Some of you sequences generated alignments that eliminated too many bases, a list is provided in " + accnosFileName + ".");
 280                         if (!flip) {
 281                                  mothurOut(" If you set the flip parameter to true mothur will try aligning the reverse compliment as well.");
 282                         }else{  mothurOut(" If the reverse compliment proved to be better it was reported.");  }
 283                         mothurOutEndLine();
 284                 }
 285
 286 #endif
 287
 288
 289
 290                 mothurOut("It took " + toString(time(NULL) - start) + " secs to align " + toString(numFastaSeqs) + " sequences.");
 291                 mothurOutEndLine();
 292                 mothurOutEndLine();
 293
 294                 return 0;
 295         }
 296         catch(exception& e) {
 297                 errorOut(e, "AlignCommand", "execute");
 298                 exit(1);
 299         }
 300 }
 301
 302 //**********************************************************************************************************************
 303
 304 int AlignCommand::driver(linePair* line, string alignFName, string reportFName, string accnosFName){
 305         try {
 306                 ofstream alignmentFile;
 307                 openOutputFile(alignFName, alignmentFile);
 308
 309                 ofstream accnosFile;
 310                 openOutputFile(accnosFName, accnosFile);
 311
 312                 NastReport report(reportFName);
 313
 314                 ifstream inFASTA;
 315                 openInputFile(candidateFileName, inFASTA);
 316
 317                 inFASTA.seekg(line->start);
 318
 319                 for(int i=0;i<line->numSeqs;i++){
 320
 321                         Sequence* candidateSeq = new Sequence(inFASTA);  gobble(inFASTA);
 322                         int origNumBases = candidateSeq->getNumBases();
 323                         string originalUnaligned = candidateSeq->getUnaligned();
 324                         int numBasesNeeded = origNumBases * threshold;
 325
 326                         if (candidateSeq->getName() != "") { //incase there is a commented sequence at the end of a file
 327                                 if (candidateSeq->getUnaligned().length() > alignment->getnRows()) {
 328                                         alignment->resize(candidateSeq->getUnaligned().length()+1);
 329                                 }
 330
 331                                 Sequence temp = templateDB->findClosestSequence(candidateSeq);
 332                                 Sequence* templateSeq = &temp;
 333
 334                                 float searchScore = templateDB->getSearchScore();
 335
 336                                 Nast* nast = new Nast(alignment, candidateSeq, templateSeq);
 337                                 Sequence* copy;
 338
 339                                 Nast* nast2;
 340                                 bool needToDeleteCopy = false;  //this is needed in case you have you enter the ifs below
 341                                                                                                 //since nast does not make a copy of hte sequence passed, and it is used by the reporter below
 342                                                                                                 //you can't delete the copy sequence til after you report, but you may choose not to create it in the first place
 343                                                                                                 //so this bool tells you if you need to delete it
 344
 345                                 //if there is a possibility that this sequence should be reversed
 346                                 if (candidateSeq->getNumBases() < numBasesNeeded) {
 347
 348                                         string wasBetter = "";
 349                                         //if the user wants you to try the reverse
 350                                         if (flip) {
 351                                                 //get reverse compliment
 352                                                 copy = new Sequence(candidateSeq->getName(), originalUnaligned);
 353                                                 copy->reverseComplement();
 354
 355                                                 //rerun alignment
 356                                                 Sequence temp2 = templateDB->findClosestSequence(copy);
 357                                                 Sequence* templateSeq2 = &temp2;
 358
 359                                                 searchScore = templateDB->getSearchScore();
 360
 361                                                 nast2 = new Nast(alignment, copy, templateSeq2);
 362
 363                                                 //check if any better
 364                                                 if (copy->getNumBases() > candidateSeq->getNumBases()) {
 365                                                         candidateSeq->setAligned(copy->getAligned());  //use reverse compliments alignment since its better
 366                                                         templateSeq = templateSeq2;
 367                                                         delete nast;
 368                                                         nast = nast2;
 369                                                         needToDeleteCopy = true;
 370                                                 }else{
 371                                                         wasBetter = "\treverse complement did NOT produce a better alignment, please check sequence.";
 372                                                         delete nast2;
 373                                                         delete copy;
 374                                                 }
 375                                         }
 376
 377                                         //create accnos file with names
 378                                         accnosFile << candidateSeq->getName() << wasBetter << endl;
 379                                 }
 380
 381                                 report.setCandidate(candidateSeq);
 382                                 report.setTemplate(templateSeq);
 383                                 report.setSearchParameters(search, searchScore);
 384                                 report.setAlignmentParameters(align, alignment);
 385                                 report.setNastParameters(*nast);
 386
 387                                 alignmentFile << '>' << candidateSeq->getName() << '\n' << candidateSeq->getAligned() << endl;
 388
 389                                 report.print();
 390                                 delete nast;
 391                                 if (needToDeleteCopy) {   delete copy;   }
 392                         }
 393                         delete candidateSeq;
 394
 395                         //report progress
 396                         if((i+1) % 100 == 0){   mothurOut(toString(i+1)); mothurOutEndLine();           }
 397
 398                 }
 399                 //report progress
 400                 if((line->numSeqs) % 100 != 0){ mothurOut(toString(line->numSeqs)); mothurOutEndLine();         }
 401
 402                 alignmentFile.close();
 403                 inFASTA.close();
 404                 accnosFile.close();
 405
 406                 return 1;
 407         }
 408         catch(exception& e) {
 409                 errorOut(e, "AlignCommand", "driver");
 410                 exit(1);
 411         }
 412 }
 413
 414 /**************************************************************************************************/
 415
 416 void AlignCommand::createProcesses(string alignFileName, string reportFileName, string accnosFName) {
 417         try {
 418 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 419                 int process = 0;
 420                 //              processIDS.resize(0);
 421
 422                 //loop through and create all the processes you want
 423                 while (process != processors) {
 424                         int pid = fork();
 425
 426                         if (pid > 0) {
 427                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 428                                 process++;
 429                         }else if (pid == 0){
 430                                 driver(lines[process], alignFileName + toString(getpid()) + ".temp", reportFileName + toString(getpid()) + ".temp", accnosFName + toString(getpid()) + ".temp");
 431                                 exit(0);
 432                         }else { mothurOut("unable to spawn the necessary processes."); mothurOutEndLine(); exit(0); }
 433                 }
 434
 435                 //force parent to wait until all the processes are done
 436                 for (int i=0;i<processors;i++) {
 437                         int temp = processIDS[i];
 438                         wait(&temp);
 439                 }
 440 #endif
 441         }
 442         catch(exception& e) {
 443                 errorOut(e, "AlignCommand", "createProcesses");
 444                 exit(1);
 445         }
 446 }
 447
 448 /**************************************************************************************************/
 449
 450 void AlignCommand::appendAlignFiles(string temp, string filename) {
 451         try{
 452
 453                 ofstream output;
 454                 ifstream input;
 455                 openOutputFileAppend(filename, output);
 456                 openInputFile(temp, input);
 457
 458                 while(char c = input.get()){
 459                         if(input.eof())         {       break;                  }
 460                         else                            {       output << c;    }
 461                 }
 462
 463                 input.close();
 464                 output.close();
 465         }
 466         catch(exception& e) {
 467                 errorOut(e, "AlignCommand", "appendAlignFiles");
 468                 exit(1);
 469         }
 470 }
 471 //**********************************************************************************************************************
 472
 473 void AlignCommand::appendReportFiles(string temp, string filename) {
 474         try{
 475
 476                 ofstream output;
 477                 ifstream input;
 478                 openOutputFileAppend(filename, output);
 479                 openInputFile(temp, input);
 480
 481                 while (!input.eof())    {       char c = input.get(); if (c == 10 || c == 13){  break;  }       } // get header line
 482
 483                 while(char c = input.get()){
 484                         if(input.eof())         {       break;                  }
 485                         else                            {       output << c;    }
 486                 }
 487
 488                 input.close();
 489                 output.close();
 490         }
 491         catch(exception& e) {
 492                 errorOut(e, "AlignCommand", "appendReportFiles");
 493                 exit(1);
 494         }
 495 }
 496
 497 //**********************************************************************************************************************