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;
 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         }
 149 };
 150
 151 /**************************************************************************************************/
 152 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
 153 #else
 154 static DWORD WINAPI MyContigsThreadFunction(LPVOID lpParam){
 155         contigsData* pDataArray;
 156         pDataArray = (contigsData*)lpParam;
 157
 158         try {
 159         int longestBase = 1000;
 160         Alignment* alignment;
 161         if(pDataArray->align == "gotoh")                        {       alignment = new GotohOverlap(pDataArray->gapOpen, pDataArray->gapExtend, pDataArray->match, pDataArray->misMatch, longestBase);                 }
 162                 else if(pDataArray->align == "needleman")       {       alignment = new NeedlemanOverlap(pDataArray->gapOpen, pDataArray->match, pDataArray->misMatch, longestBase);                            }
 163
 164         int num = 0;
 165         string thisffastafile = pDataArray->files[0];
 166         string thisfqualfile = pDataArray->files[1];
 167         string thisrfastafile = pDataArray->files[2];
 168         string thisrqualfile = pDataArray->files[3];
 169
 170         if (pDataArray->m->debug) {  pDataArray->m->mothurOut("[DEBUG]: ffasta = " + thisffastafile + ".\n[DEBUG]: fqual = " + thisfqualfile + ".\n[DEBUG]: rfasta = " + thisrfastafile + ".\n[DEBUG]: rqual = " + thisrqualfile + ".\n"); }
 171
 172                 if(pDataArray->allFiles){
 173                         for (int i = 0; i < pDataArray->fastaFileNames.size(); i++) { //clears old file
 174                                 for (int j = 0; j < pDataArray->fastaFileNames[i].size(); j++) { //clears old file
 175                                         if (pDataArray->fastaFileNames[i][j] != "") {
 176                                                 ofstream temp;
 177                                                 pDataArray->m->openOutputFile(pDataArray->fastaFileNames[i][j], temp);                  temp.close();
 178                         if (thisfqualfile != "") { pDataArray->m->openOutputFile(pDataArray->qualFileNames[i][j], temp);                        temp.close(); }
 179                                         }
 180                                 }
 181                         }
 182                 }
 183
 184         ifstream inFFasta, inRFasta, inFQual, inRQual;
 185         ofstream outFasta, outQual, outMisMatch, outScrapFasta, outScrapQual;
 186         pDataArray->m->openInputFile(thisffastafile, inFFasta);
 187         pDataArray->m->openInputFile(thisrfastafile, inRFasta);
 188         if (thisfqualfile != "") {
 189             pDataArray->m->openInputFile(thisfqualfile, inFQual);
 190             pDataArray->m->openInputFile(thisrqualfile, inRQual);
 191             pDataArray->m->openOutputFile(pDataArray->outputQual, outQual);
 192             pDataArray->m->openOutputFile(pDataArray->outputScrapQual, outScrapQual);
 193         }
 194         pDataArray->m->openOutputFile(pDataArray->outputFasta, outFasta);
 195         pDataArray->m->openOutputFile(pDataArray->outputMisMatches, outMisMatch);
 196         pDataArray->m->openOutputFile(pDataArray->outputScrapFasta, outScrapFasta);
 197
 198         outMisMatch << "Name\tLength\tMisMatches\n";
 199
 200         TrimOligos trimOligos(pDataArray->pdiffs, pDataArray->bdiffs, 0, 0, pDataArray->primers, pDataArray->barcodes);
 201
 202         while ((!inFFasta.eof()) && (!inRFasta.eof())) {
 203
 204             if (pDataArray->m->control_pressed) { break; }
 205
 206             int success = 1;
 207             string trashCode = "";
 208             int currentSeqsDiffs = 0;
 209
 210             //read seqs and quality info
 211             Sequence fSeq(inFFasta); pDataArray->m->gobble(inFFasta);
 212             Sequence rSeq(inRFasta); pDataArray->m->gobble(inRFasta);
 213             QualityScores* fQual = NULL; QualityScores* rQual = NULL;
 214             if (thisfqualfile != "") {
 215                 fQual = new QualityScores(inFQual); pDataArray->m->gobble(inFQual);
 216                 rQual = new QualityScores(inRQual); pDataArray->m->gobble(inRQual);
 217             }
 218
 219             int barcodeIndex = 0;
 220             int primerIndex = 0;
 221
 222             if(pDataArray->barcodes.size() != 0){
 223                 if (thisfqualfile != "") {
 224                     success = trimOligos.stripBarcode(fSeq, rSeq, *fQual, *rQual, barcodeIndex);
 225                 }else {
 226                     success = trimOligos.stripBarcode(fSeq, rSeq, barcodeIndex);
 227                 }
 228                 if(success > pDataArray->bdiffs)                {       trashCode += 'b';       }
 229                 else{ currentSeqsDiffs += success;  }
 230             }
 231
 232             if(pDataArray->primers.size() != 0){
 233                 if (thisfqualfile != "") {
 234                     success = trimOligos.stripForward(fSeq, rSeq, *fQual, *rQual, primerIndex);
 235                 }else {
 236                     success = trimOligos.stripForward(fSeq, rSeq, primerIndex);
 237                 }
 238                 if(success > pDataArray->pdiffs)                {       trashCode += 'f';       }
 239                 else{ currentSeqsDiffs += success;  }
 240             }
 241
 242             if (currentSeqsDiffs > pDataArray->tdiffs)  {       trashCode += 't';   }
 243
 244             //flip the reverse reads
 245             rSeq.reverseComplement();
 246             if (thisfqualfile != "") { rQual->flipQScores(); }
 247
 248             //pairwise align
 249             alignment->align(fSeq.getUnaligned(), rSeq.getUnaligned());
 250             map<int, int> ABaseMap = alignment->getSeqAAlnBaseMap();
 251             map<int, int> BBaseMap = alignment->getSeqBAlnBaseMap();
 252             fSeq.setAligned(alignment->getSeqAAln());
 253             rSeq.setAligned(alignment->getSeqBAln());
 254             int length = fSeq.getAligned().length();
 255
 256             //traverse alignments merging into one contiguous seq
 257             string contig = "";
 258             vector<int> contigScores;
 259             int numMismatches = 0;
 260             string seq1 = fSeq.getAligned();
 261             string seq2 = rSeq.getAligned();
 262             vector<int> scores1, scores2;
 263             if (thisfqualfile != "") {
 264                 scores1 = fQual->getQualityScores();
 265                 scores2 = rQual->getQualityScores();
 266                 delete fQual; delete rQual;
 267             }
 268
 269             int overlapStart = fSeq.getStartPos();
 270             int seq2Start = rSeq.getStartPos();
 271             //bigger of the 2 starting positions is the location of the overlapping start
 272             if (overlapStart < seq2Start) { //seq2 starts later so take from 0 to seq2Start from seq1
 273                 overlapStart = seq2Start;
 274                 for (int i = 0; i < overlapStart; i++) {
 275                     contig += seq1[i];
 276                     if (thisfqualfile != "") { contigScores.push_back(scores1[ABaseMap[i]]); }
 277                 }
 278             }else { //seq1 starts later so take from 0 to overlapStart from seq2
 279                 for (int i = 0; i < overlapStart; i++) {
 280                     contig += seq2[i];
 281                     if (thisfqualfile != "") { contigScores.push_back(scores2[BBaseMap[i]]); }
 282                 }
 283             }
 284
 285             int seq1End = fSeq.getEndPos();
 286             int seq2End = rSeq.getEndPos();
 287             int overlapEnd = seq1End;
 288             if (seq2End < overlapEnd) { overlapEnd = seq2End; }  //smallest end position is where overlapping ends
 289
 290             for (int i = overlapStart; i < overlapEnd; i++) {
 291                 if (seq1[i] == seq2[i]) { //match, add base and choose highest score
 292                     contig += seq1[i];
 293                     if (thisfqualfile != "") {
 294                         contigScores.push_back(scores1[ABaseMap[i]]);
 295                         if (scores1[ABaseMap[i]] < scores2[BBaseMap[i]]) { contigScores[contigScores.size()-1] = scores2[BBaseMap[i]]; }
 296                     }
 297                 }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
 298                     if (thisfqualfile != "") {
 299                         if (scores2[BBaseMap[i]] < pDataArray->threshold) { } //
 300                         else {
 301                             contig += seq2[i];
 302                             contigScores.push_back(scores2[BBaseMap[i]]);
 303                         }
 304                     }else { contig += seq2[i]; }
 305                 }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
 306                     if (thisfqualfile != "") {
 307                         if (scores1[ABaseMap[i]] < pDataArray->threshold) { } //
 308                         else {
 309                             contig += seq1[i];
 310                             contigScores.push_back(scores1[ABaseMap[i]]);
 311                         }
 312                     }else { contig += seq1[i]; }
 313                 }else if (((seq1[i] != '-') && (seq1[i] != '.')) && ((seq2[i] != '-') && (seq2[i] != '.'))) { //both bases choose one with better quality
 314                     if (thisfqualfile != "") {
 315                         char c = seq1[i];
 316                         contigScores.push_back(scores1[ABaseMap[i]]);
 317                         if (scores1[ABaseMap[i]] < scores2[BBaseMap[i]]) { contigScores[contigScores.size()-1] = scores2[BBaseMap[i]]; c = seq2[i]; }
 318                         contig += c;
 319                         numMismatches++;
 320                     }else { numMismatches++; }
 321                 }else { //should never get here
 322                     pDataArray->m->mothurOut("[ERROR]: case I didn't think of seq1 = " + toString(seq1[i]) + " and seq2 = " + toString(seq2[i]) + "\n");
 323                 }
 324             }
 325
 326             if (seq1End < seq2End) { //seq1 ends before seq2 so take from overlap to length from seq2
 327                 for (int i = overlapEnd; i < length; i++) {
 328                     contig += seq2[i];
 329                     if (thisfqualfile != "") { contigScores.push_back(scores2[BBaseMap[i]]); }
 330                 }
 331             }else { //seq2 ends before seq1 so take from overlap to length from seq1
 332                 for (int i = overlapEnd; i < length; i++) {
 333                     contig += seq1[i];
 334                     if (thisfqualfile != "") { contigScores.push_back(scores1[ABaseMap[i]]); }
 335                 }
 336
 337             }
 338
 339             if(trashCode.length() == 0){
 340                 bool ignore = false;
 341                 if (pDataArray->createGroup) {
 342                     if(pDataArray->barcodes.size() != 0){
 343                         string thisGroup = pDataArray->barcodeNameVector[barcodeIndex];
 344                         if (pDataArray->primers.size() != 0) {
 345                             if (pDataArray->primerNameVector[primerIndex] != "") {
 346                                 if(thisGroup != "") {
 347                                     thisGroup += "." + pDataArray->primerNameVector[primerIndex];
 348                                 }else {
 349                                     thisGroup = pDataArray->primerNameVector[primerIndex];
 350                                 }
 351                             }
 352                         }
 353
 354                         if (pDataArray->m->debug) { pDataArray->m->mothurOut(", group= " + thisGroup + "\n"); }
 355
 356                         int pos = thisGroup.find("ignore");
 357                         if (pos == string::npos) {
 358                             pDataArray->groupMap[fSeq.getName()] = thisGroup;
 359
 360                             map<string, int>::iterator it = pDataArray->groupCounts.find(thisGroup);
 361                             if (it == pDataArray->groupCounts.end()) {  pDataArray->groupCounts[thisGroup] = 1; }
 362                             else { pDataArray->groupCounts[it->first] ++; }
 363                         }else { ignore = true; }
 364                     }
 365                 }
 366
 367                 if(pDataArray->allFiles && !ignore){
 368                     ofstream output;
 369                     pDataArray->m->openOutputFileAppend(pDataArray->fastaFileNames[barcodeIndex][primerIndex], output);
 370                     output << ">" << fSeq.getName() << endl << contig << endl;
 371                     output.close();
 372
 373                     if (thisfqualfile != "") {
 374                         pDataArray->m->openOutputFileAppend(pDataArray->qualFileNames[barcodeIndex][primerIndex], output);
 375                         output << ">" << fSeq.getName() << endl;
 376                         for (int i = 0; i < contigScores.size(); i++) { output << contigScores[i] << ' '; }
 377                         output << endl;
 378                         output.close();
 379                     }
 380                 }
 381
 382                 //output
 383                 outFasta << ">" << fSeq.getName() << endl << contig << endl;
 384                 if (thisfqualfile != "") {
 385                     outQual << ">" << fSeq.getName() << endl;
 386                     for (int i = 0; i < contigScores.size(); i++) { outQual << contigScores[i] << ' '; }
 387                     outQual << endl;
 388                 }
 389                 outMisMatch << fSeq.getName() << '\t' << contig.length() << '\t' << numMismatches << endl;
 390             }else {
 391                 //output
 392                 outScrapFasta << ">" << fSeq.getName() << " | " << trashCode << endl << contig << endl;
 393                 if (thisfqualfile != "") {
 394                     outScrapQual << ">" << fSeq.getName() << " | " << trashCode << endl;
 395                     for (int i = 0; i < contigScores.size(); i++) { outScrapQual << contigScores[i] << ' '; }
 396                     outScrapQual << endl;
 397                 }
 398             }
 399             num++;
 400
 401                         //report progress
 402                         if((num) % 1000 == 0){  pDataArray->m->mothurOut(toString(num)); pDataArray->m->mothurOutEndLine();             }
 403                 }
 404
 405                 //report progress
 406                 if((num) % 1000 != 0){  pDataArray->m->mothurOut(toString(num)); pDataArray->m->mothurOutEndLine();             }
 407
 408         inFFasta.close();
 409         inRFasta.close();
 410         outFasta.close();
 411         outMisMatch.close();
 412         outScrapFasta.close();
 413         if (thisfqualfile != "") {
 414             inFQual.close();
 415             inRQual.close();
 416             outQual.close();
 417             outScrapQual.close();
 418         }
 419         delete alignment;
 420
 421         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); } }
 422
 423         return 0;
 424
 425         }
 426         catch(exception& e) {
 427                 pDataArray->m->errorOut(e, "AlignCommand", "MyContigsThreadFunction");
 428                 exit(1);
 429         }
 430 }
 431 #endif
 432
 433
 434 #endif