makecontigscommand.h

   1 #ifndef Mothur_makecontigscommand_h
   2 #define Mothur_makecontigscommand_h
   3
   4 //
   5 //  makecontigscommand.h
   6 //  Mothur
   7 //
   8 //  Created by Sarah Westcott on 5/15/12.
   9 //  Copyright (c) 2012 Schloss Lab. All rights reserved.
  10 //
  11
  12 #include "command.hpp"
  13 #include "sequence.hpp"
  14 #include "qualityscores.h"
  15 #include "alignment.hpp"
  16 #include "gotohoverlap.hpp"
  17 #include "needlemanoverlap.hpp"
  18 #include "blastalign.hpp"
  19 #include "noalign.hpp"
  20 #include "trimoligos.h"
  21
  22 struct fastqRead {
  23         vector<int> scores;
  24         string name;
  25         string sequence;
  26
  27         fastqRead() { name = ""; sequence = ""; scores.clear(); };
  28         fastqRead(string n, string s, vector<int> sc) : name(n), sequence(s), scores(sc) {};
  29         ~fastqRead() {};
  30 };
  31
  32 struct pairFastqRead {
  33         fastqRead forward;
  34     fastqRead reverse;
  35
  36         pairFastqRead() {};
  37         pairFastqRead(fastqRead f, fastqRead r) : forward(f), reverse(r){};
  38         ~pairFastqRead() {};
  39 };
  40 /**************************************************************************************************/
  41
  42 class MakeContigsCommand : public Command {
  43 public:
  44     MakeContigsCommand(string);
  45     MakeContigsCommand();
  46     ~MakeContigsCommand(){}
  47
  48     vector<string> setParameters();
  49     string getCommandName()                     { return "make.contigs";                        }
  50     string getCommandCategory()         { return "Sequence Processing";         }
  51     //commmand category choices: Sequence Processing, OTU-Based Approaches, Hypothesis Testing, Phylotype Analysis, General, Clustering and Hidden
  52
  53         string getHelpString();
  54     string getOutputPattern(string);
  55     string getCitation() { return "http://www.mothur.org/wiki/Make.contigs"; }
  56     string getDescription()             { return "description"; }
  57
  58     int execute();
  59     void help() { m->mothurOut(getHelpString()); }
  60
  61 private:
  62     bool abort, allFiles, createGroup;
  63     string outputDir, ffastqfile, rfastqfile, align, oligosfile, rfastafile, ffastafile, rqualfile, fqualfile, file, format;
  64         float match, misMatch, gapOpen, gapExtend;
  65         int processors, longestBase, threshold, tdiffs, bdiffs, pdiffs, ldiffs, sdiffs;
  66     vector<string> outputNames;
  67
  68     map<int, oligosPair> barcodes;
  69         map<int, oligosPair> primers;
  70     vector<string>  linker;
  71     vector<string>  spacer;
  72         vector<string> primerNameVector;
  73         vector<string> barcodeNameVector;
  74         vector<char> convertTable;
  75
  76         map<string, int> groupCounts;
  77     map<string, string> groupMap;
  78
  79     vector<int> convertQual(string);
  80     fastqRead readFastq(ifstream&, bool&);
  81     vector< vector< vector<string> > > preProcessData(unsigned long int&);
  82     vector< vector<string> > readFileNames(string);
  83     vector< vector<string> > readFastqFiles(unsigned long int&, string, string);
  84     vector< vector<string> > readFastaFiles(unsigned long int&, string, string);
  85     bool checkReads(fastqRead&, fastqRead&, string, string);
  86     int createProcesses(vector< vector<string> >, string, string, string, string, string, vector<vector<string> >, vector<vector<string> >);
  87     int driver(vector<string>, string, string, string, string, string, vector<vector<string> >, vector<vector<string> >);
  88     bool getOligos(vector<vector<string> >&, vector< vector<string> >&, string);
  89     string reverseOligo(string);
  90     vector<pairFastqRead> getReads(bool ignoref, bool ignorer, fastqRead forward, fastqRead reverse, map<string, fastqRead>& uniques);
  91 };
  92
  93 /**************************************************************************************************/
  94
  95 /**************************************************************************************************/
  96 //custom data structure for threads to use.
  97 // This is passed by void pointer so it can be any data type
  98 // that can be passed using a single void pointer (LPVOID).
  99 struct contigsData {
 100         string outputFasta;
 101         string outputQual;
 102     string outputScrapFasta;
 103         string outputScrapQual;
 104         string outputMisMatches;
 105         string align;
 106     vector<string> files;
 107     vector<vector<string> > fastaFileNames;
 108     vector<vector<string> > qualFileNames;
 109         MothurOut* m;
 110         float match, misMatch, gapOpen, gapExtend;
 111         int count, threshold, threadID, pdiffs, bdiffs, tdiffs;
 112     bool allFiles, createGroup, done;
 113     map<string, int> groupCounts;
 114     map<string, string> groupMap;
 115     vector<string> primerNameVector;
 116         vector<string> barcodeNameVector;
 117     map<int, oligosPair> barcodes;
 118         map<int, oligosPair> primers;
 119
 120         contigsData(){}
 121         contigsData(vector<string> f, string of, string oq, string osf, string osq, string om, string al, MothurOut* mout, float ma, float misMa, float gapO, float gapE, int thr, map<int, oligosPair> br, map<int, oligosPair> pr, vector<vector<string> > ffn, vector<vector<string> > qfn, vector<string>bnv, vector<string> pnv, int pdf, int bdf, int tdf, bool cg, bool all, int tid) {
 122         files = f;
 123                 outputFasta = of;
 124         outputQual = oq;
 125         outputMisMatches = om;
 126         m = mout;
 127                 match = ma;
 128                 misMatch = misMa;
 129                 gapOpen = gapO;
 130                 gapExtend = gapE;
 131         threshold = thr;
 132                 align = al;
 133                 count = 0;
 134         outputScrapFasta = osf;
 135         outputScrapQual = osq;
 136         fastaFileNames = ffn;
 137         qualFileNames = qfn;
 138         barcodes = br;
 139         primers = pr;
 140         barcodeNameVector = bnv;
 141         primerNameVector = pnv;
 142         pdiffs = pdf;
 143         bdiffs = bdf;
 144         tdiffs = tdf;
 145         allFiles = all;
 146         createGroup = cg;
 147                 threadID = tid;
 148         done=false;
 149         }
 150 };
 151
 152 /**************************************************************************************************/
 153 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
 154 #else
 155 static DWORD WINAPI MyContigsThreadFunction(LPVOID lpParam){
 156         contigsData* pDataArray;
 157         pDataArray = (contigsData*)lpParam;
 158
 159         try {
 160         int longestBase = 1000;
 161         Alignment* alignment;
 162         if(pDataArray->align == "gotoh")                        {       alignment = new GotohOverlap(pDataArray->gapOpen, pDataArray->gapExtend, pDataArray->match, pDataArray->misMatch, longestBase);                 }
 163                 else if(pDataArray->align == "needleman")       {       alignment = new NeedlemanOverlap(pDataArray->gapOpen, pDataArray->match, pDataArray->misMatch, longestBase);                            }
 164
 165         pDataArray->count = 0;
 166         string thisffastafile = pDataArray->files[0];
 167         string thisfqualfile = pDataArray->files[1];
 168         string thisrfastafile = pDataArray->files[2];
 169         string thisrqualfile = pDataArray->files[3];
 170
 171         if (pDataArray->m->debug) {  pDataArray->m->mothurOut("[DEBUG]: ffasta = " + thisffastafile + ".\n[DEBUG]: fqual = " + thisfqualfile + ".\n[DEBUG]: rfasta = " + thisrfastafile + ".\n[DEBUG]: rqual = " + thisrqualfile + ".\n"); }
 172
 173                 if(pDataArray->allFiles){
 174                         for (int i = 0; i < pDataArray->fastaFileNames.size(); i++) { //clears old file
 175                                 for (int j = 0; j < pDataArray->fastaFileNames[i].size(); j++) { //clears old file
 176                                         if (pDataArray->fastaFileNames[i][j] != "") {
 177                                                 ofstream temp;
 178                                                 pDataArray->m->openOutputFile(pDataArray->fastaFileNames[i][j], temp);                  temp.close();
 179                         if (thisfqualfile != "") { pDataArray->m->openOutputFile(pDataArray->qualFileNames[i][j], temp);                        temp.close(); }
 180                                         }
 181                                 }
 182                         }
 183                 }
 184
 185         ifstream inFFasta, inRFasta, inFQual, inRQual;
 186         ofstream outFasta, outQual, outMisMatch, outScrapFasta, outScrapQual;
 187         pDataArray->m->openInputFile(thisffastafile, inFFasta);
 188         pDataArray->m->openInputFile(thisrfastafile, inRFasta);
 189         if (thisfqualfile != "") {
 190             pDataArray->m->openInputFile(thisfqualfile, inFQual);
 191             pDataArray->m->openInputFile(thisrqualfile, inRQual);
 192             pDataArray->m->openOutputFile(pDataArray->outputQual, outQual);
 193             pDataArray->m->openOutputFile(pDataArray->outputScrapQual, outScrapQual);
 194         }
 195         pDataArray->m->openOutputFile(pDataArray->outputFasta, outFasta);
 196         pDataArray->m->openOutputFile(pDataArray->outputMisMatches, outMisMatch);
 197         pDataArray->m->openOutputFile(pDataArray->outputScrapFasta, outScrapFasta);
 198
 199         outMisMatch << "Name\tLength\tMisMatches\n";
 200
 201         TrimOligos trimOligos(pDataArray->pdiffs, pDataArray->bdiffs, 0, 0, pDataArray->primers, pDataArray->barcodes);
 202
 203         while ((!inFFasta.eof()) && (!inRFasta.eof())) {
 204
 205             if (pDataArray->m->control_pressed) { break; }
 206
 207             int success = 1;
 208             string trashCode = "";
 209             int currentSeqsDiffs = 0;
 210
 211             //read seqs and quality info
 212             Sequence fSeq(inFFasta); pDataArray->m->gobble(inFFasta);
 213             Sequence rSeq(inRFasta); pDataArray->m->gobble(inRFasta);
 214             QualityScores* fQual = NULL; QualityScores* rQual = NULL;
 215             if (thisfqualfile != "") {
 216                 fQual = new QualityScores(inFQual); pDataArray->m->gobble(inFQual);
 217                 rQual = new QualityScores(inRQual); pDataArray->m->gobble(inRQual);
 218             }
 219
 220             int barcodeIndex = 0;
 221             int primerIndex = 0;
 222
 223             if(pDataArray->barcodes.size() != 0){
 224                 if (thisfqualfile != "") {
 225                     success = trimOligos.stripBarcode(fSeq, rSeq, *fQual, *rQual, barcodeIndex);
 226                 }else {
 227                     success = trimOligos.stripBarcode(fSeq, rSeq, barcodeIndex);
 228                 }
 229                 if(success > pDataArray->bdiffs)                {       trashCode += 'b';       }
 230                 else{ currentSeqsDiffs += success;  }
 231             }
 232
 233             if(pDataArray->primers.size() != 0){
 234                 if (thisfqualfile != "") {
 235                     success = trimOligos.stripForward(fSeq, rSeq, *fQual, *rQual, primerIndex);
 236                 }else {
 237                     success = trimOligos.stripForward(fSeq, rSeq, primerIndex);
 238                 }
 239                 if(success > pDataArray->pdiffs)                {       trashCode += 'f';       }
 240                 else{ currentSeqsDiffs += success;  }
 241             }
 242
 243             if (currentSeqsDiffs > pDataArray->tdiffs)  {       trashCode += 't';   }
 244
 245             //flip the reverse reads
 246             rSeq.reverseComplement();
 247             if (thisfqualfile != "") { rQual->flipQScores(); }
 248
 249             //pairwise align
 250             alignment->align(fSeq.getUnaligned(), rSeq.getUnaligned());
 251             map<int, int> ABaseMap = alignment->getSeqAAlnBaseMap();
 252             map<int, int> BBaseMap = alignment->getSeqBAlnBaseMap();
 253             fSeq.setAligned(alignment->getSeqAAln());
 254             rSeq.setAligned(alignment->getSeqBAln());
 255             int length = fSeq.getAligned().length();
 256
 257             //traverse alignments merging into one contiguous seq
 258             string contig = "";
 259             vector<int> contigScores;
 260             int numMismatches = 0;
 261             string seq1 = fSeq.getAligned();
 262             string seq2 = rSeq.getAligned();
 263             vector<int> scores1, scores2;
 264             if (thisfqualfile != "") {
 265                 scores1 = fQual->getQualityScores();
 266                 scores2 = rQual->getQualityScores();
 267                 delete fQual; delete rQual;
 268             }
 269
 270             int overlapStart = fSeq.getStartPos();
 271             int seq2Start = rSeq.getStartPos();
 272             //bigger of the 2 starting positions is the location of the overlapping start
 273             if (overlapStart < seq2Start) { //seq2 starts later so take from 0 to seq2Start from seq1
 274                 overlapStart = seq2Start;
 275                 for (int i = 0; i < overlapStart; i++) {
 276                     contig += seq1[i];
 277                     if (thisfqualfile != "") { contigScores.push_back(scores1[ABaseMap[i]]); }
 278                 }
 279             }else { //seq1 starts later so take from 0 to overlapStart from seq2
 280                 for (int i = 0; i < overlapStart; i++) {
 281                     contig += seq2[i];
 282                     if (thisfqualfile != "") { contigScores.push_back(scores2[BBaseMap[i]]); }
 283                 }
 284             }
 285
 286             int seq1End = fSeq.getEndPos();
 287             int seq2End = rSeq.getEndPos();
 288             int overlapEnd = seq1End;
 289             if (seq2End < overlapEnd) { overlapEnd = seq2End; }  //smallest end position is where overlapping ends
 290
 291             for (int i = overlapStart; i < overlapEnd; i++) {
 292                 if (seq1[i] == seq2[i]) { //match, add base and choose highest score
 293                     contig += seq1[i];
 294                     if (thisfqualfile != "") {
 295                         contigScores.push_back(scores1[ABaseMap[i]]);
 296                         if (scores1[ABaseMap[i]] < scores2[BBaseMap[i]]) { contigScores[contigScores.size()-1] = scores2[BBaseMap[i]]; }
 297                     }
 298                 }else if (((seq1[i] == '.') || (seq1[i] == '-')) && ((seq2[i] != '-') && (seq2[i] != '.'))) { //seq1 is a gap and seq2 is a base, choose seq2, unless quality score for base is below threshold. In that case eliminate base
 299                     if (thisfqualfile != "") {
 300                         if (scores2[BBaseMap[i]] < pDataArray->threshold) { } //
 301                         else {
 302                             contig += seq2[i];
 303                             contigScores.push_back(scores2[BBaseMap[i]]);
 304                         }
 305                     }else { contig += seq2[i]; }
 306                 }else if (((seq2[i] == '.') || (seq2[i] == '-')) && ((seq1[i] != '-') && (seq1[i] != '.'))) { //seq2 is a gap and seq1 is a base, choose seq1, unless quality score for base is below threshold. In that case eliminate base
 307                     if (thisfqualfile != "") {
 308                         if (scores1[ABaseMap[i]] < pDataArray->threshold) { } //
 309                         else {
 310                             contig += seq1[i];
 311                             contigScores.push_back(scores1[ABaseMap[i]]);
 312                         }
 313                     }else { contig += seq1[i]; }
 314                 }else if (((seq1[i] != '-') && (seq1[i] != '.')) && ((seq2[i] != '-') && (seq2[i] != '.'))) { //both bases choose one with better quality
 315                     if (thisfqualfile != "") {
 316                         char c = seq1[i];
 317                         contigScores.push_back(scores1[ABaseMap[i]]);
 318                         if (scores1[ABaseMap[i]] < scores2[BBaseMap[i]]) { contigScores[contigScores.size()-1] = scores2[BBaseMap[i]]; c = seq2[i]; }
 319                         contig += c;
 320                         numMismatches++;
 321                     }else { numMismatches++; }
 322                 }else { //should never get here
 323                     pDataArray->m->mothurOut("[ERROR]: case I didn't think of seq1 = " + toString(seq1[i]) + " and seq2 = " + toString(seq2[i]) + "\n");
 324                 }
 325             }
 326
 327             if (seq1End < seq2End) { //seq1 ends before seq2 so take from overlap to length from seq2
 328                 for (int i = overlapEnd; i < length; i++) {
 329                     contig += seq2[i];
 330                     if (thisfqualfile != "") { contigScores.push_back(scores2[BBaseMap[i]]); }
 331                 }
 332             }else { //seq2 ends before seq1 so take from overlap to length from seq1
 333                 for (int i = overlapEnd; i < length; i++) {
 334                     contig += seq1[i];
 335                     if (thisfqualfile != "") { contigScores.push_back(scores1[ABaseMap[i]]); }
 336                 }
 337
 338             }
 339
 340             if(trashCode.length() == 0){
 341                 bool ignore = false;
 342                 if (pDataArray->createGroup) {
 343                     if(pDataArray->barcodes.size() != 0){
 344                         string thisGroup = pDataArray->barcodeNameVector[barcodeIndex];
 345                         if (pDataArray->primers.size() != 0) {
 346                             if (pDataArray->primerNameVector[primerIndex] != "") {
 347                                 if(thisGroup != "") {
 348                                     thisGroup += "." + pDataArray->primerNameVector[primerIndex];
 349                                 }else {
 350                                     thisGroup = pDataArray->primerNameVector[primerIndex];
 351                                 }
 352                             }
 353                         }
 354
 355                         if (pDataArray->m->debug) { pDataArray->m->mothurOut(", group= " + thisGroup + "\n"); }
 356
 357                         int pos = thisGroup.find("ignore");
 358                         if (pos == string::npos) {
 359                             pDataArray->groupMap[fSeq.getName()] = thisGroup;
 360
 361                             map<string, int>::iterator it = pDataArray->groupCounts.find(thisGroup);
 362                             if (it == pDataArray->groupCounts.end()) {  pDataArray->groupCounts[thisGroup] = 1; }
 363                             else { pDataArray->groupCounts[it->first] ++; }
 364                         }else { ignore = true; }
 365                     }
 366                 }
 367
 368                 if(pDataArray->allFiles && !ignore){
 369                     ofstream output;
 370                     pDataArray->m->openOutputFileAppend(pDataArray->fastaFileNames[barcodeIndex][primerIndex], output);
 371                     output << ">" << fSeq.getName() << endl << contig << endl;
 372                     output.close();
 373
 374                     if (thisfqualfile != "") {
 375                         pDataArray->m->openOutputFileAppend(pDataArray->qualFileNames[barcodeIndex][primerIndex], output);
 376                         output << ">" << fSeq.getName() << endl;
 377                         for (int i = 0; i < contigScores.size(); i++) { output << contigScores[i] << ' '; }
 378                         output << endl;
 379                         output.close();
 380                     }
 381                 }
 382
 383                 //output
 384                 outFasta << ">" << fSeq.getName() << endl << contig << endl;
 385                 if (thisfqualfile != "") {
 386                     outQual << ">" << fSeq.getName() << endl;
 387                     for (int i = 0; i < contigScores.size(); i++) { outQual << contigScores[i] << ' '; }
 388                     outQual << endl;
 389                 }
 390                 outMisMatch << fSeq.getName() << '\t' << contig.length() << '\t' << numMismatches << endl;
 391             }else {
 392                 //output
 393                 outScrapFasta << ">" << fSeq.getName() << " | " << trashCode << endl << contig << endl;
 394                 if (thisfqualfile != "") {
 395                     outScrapQual << ">" << fSeq.getName() << " | " << trashCode << endl;
 396                     for (int i = 0; i < contigScores.size(); i++) { outScrapQual << contigScores[i] << ' '; }
 397                     outScrapQual << endl;
 398                 }
 399             }
 400             pDataArray->count++;
 401
 402                         //report progress
 403                         if((pDataArray->count) % 1000 == 0){    pDataArray->m->mothurOut(toString(pDataArray->count)); pDataArray->m->mothurOutEndLine();               }
 404                 }
 405
 406                 //report progress
 407                 if((pDataArray->count) % 1000 != 0){    pDataArray->m->mothurOut(toString(pDataArray->count)); pDataArray->m->mothurOutEndLine();               }
 408
 409         inFFasta.close();
 410         inRFasta.close();
 411         outFasta.close();
 412         outMisMatch.close();
 413         outScrapFasta.close();
 414         if (thisfqualfile != "") {
 415             inFQual.close();
 416             inRQual.close();
 417             outQual.close();
 418             outScrapQual.close();
 419         }
 420         delete alignment;
 421
 422         pDataArray->done = true;
 423         if (pDataArray->m->control_pressed) {  pDataArray->m->mothurRemove(pDataArray->outputFasta);  pDataArray->m->mothurRemove(pDataArray->outputMisMatches);  pDataArray->m->mothurRemove(pDataArray->outputScrapFasta);  if (thisfqualfile != "") { pDataArray->m->mothurRemove(pDataArray->outputQual); pDataArray->m->mothurRemove(pDataArray->outputScrapQual); } }
 424
 425         return 0;
 426
 427         }
 428         catch(exception& e) {
 429                 pDataArray->m->errorOut(e, "AlignCommand", "MyContigsThreadFunction");
 430                 exit(1);
 431         }
 432 }
 433 #endif
 434
 435
 436 #endif