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, insert, tdiffs, bdiffs, pdiffs, ldiffs, sdiffs, deltaq;
  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, vector<vector<string> >);
  87     int driver(vector<string>, string, string, string, vector<vector<string> >, int);
  88     bool getOligos(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 outputScrapFasta;
 102         string outputMisMatches;
 103         string align;
 104     vector<string> files;
 105     vector<vector<string> > fastaFileNames;
 106         MothurOut* m;
 107         float match, misMatch, gapOpen, gapExtend;
 108         int count, insert, threadID, pdiffs, bdiffs, tdiffs, deltaq;
 109     bool allFiles, createGroup, done;
 110     map<string, int> groupCounts;
 111     map<string, string> groupMap;
 112     vector<string> primerNameVector;
 113         vector<string> barcodeNameVector;
 114     map<int, oligosPair> barcodes;
 115         map<int, oligosPair> primers;
 116
 117         contigsData(){}
 118         contigsData(vector<string> f, string of, string osf, string om, string al, MothurOut* mout, float ma, float misMa, float gapO, float gapE, int thr, int delt, map<int, oligosPair> br, map<int, oligosPair> pr, vector<vector<string> > ffn, vector<string>bnv, vector<string> pnv, int pdf, int bdf, int tdf, bool cg, bool all, int tid) {
 119         files = f;
 120                 outputFasta = of;
 121         outputMisMatches = om;
 122         m = mout;
 123                 match = ma;
 124                 misMatch = misMa;
 125                 gapOpen = gapO;
 126                 gapExtend = gapE;
 127         insert = thr;
 128                 align = al;
 129                 count = 0;
 130         outputScrapFasta = osf;
 131         fastaFileNames = ffn;
 132         barcodes = br;
 133         primers = pr;
 134         barcodeNameVector = bnv;
 135         primerNameVector = pnv;
 136         pdiffs = pdf;
 137         bdiffs = bdf;
 138         tdiffs = tdf;
 139         allFiles = all;
 140         createGroup = cg;
 141                 threadID = tid;
 142         deltaq = delt;
 143         done=false;
 144         }
 145 };
 146
 147 /**************************************************************************************************/
 148 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
 149 #else
 150 static DWORD WINAPI MyContigsThreadFunction(LPVOID lpParam){
 151         contigsData* pDataArray;
 152         pDataArray = (contigsData*)lpParam;
 153
 154         try {
 155         int longestBase = 1000;
 156         Alignment* alignment;
 157         if(pDataArray->align == "gotoh")                        {       alignment = new GotohOverlap(pDataArray->gapOpen, pDataArray->gapExtend, pDataArray->match, pDataArray->misMatch, longestBase);                 }
 158                 else if(pDataArray->align == "needleman")       {       alignment = new NeedlemanOverlap(pDataArray->gapOpen, pDataArray->match, pDataArray->misMatch, longestBase);                            }
 159
 160         pDataArray->count = 0;
 161         string thisffastafile = pDataArray->files[0];
 162         string thisfqualfile = pDataArray->files[1];
 163         string thisrfastafile = pDataArray->files[2];
 164         string thisrqualfile = pDataArray->files[3];
 165
 166         if (pDataArray->m->debug) {  pDataArray->m->mothurOut("[DEBUG]: ffasta = " + thisffastafile + ".\n[DEBUG]: fqual = " + thisfqualfile + ".\n[DEBUG]: rfasta = " + thisrfastafile + ".\n[DEBUG]: rqual = " + thisrqualfile + ".\n"); }
 167
 168                 if(pDataArray->allFiles){
 169                         for (int i = 0; i < pDataArray->fastaFileNames.size(); i++) { //clears old file
 170                                 for (int j = 0; j < pDataArray->fastaFileNames[i].size(); j++) { //clears old file
 171                                         if (pDataArray->fastaFileNames[i][j] != "") {
 172                                                 ofstream temp;
 173                                                 pDataArray->m->openOutputFile(pDataArray->fastaFileNames[i][j], temp);                  temp.close();
 174                                         }
 175                                 }
 176                         }
 177                 }
 178
 179         ifstream inFFasta, inRFasta, inFQual, inRQual;
 180         ofstream outFasta, outMisMatch, outScrapFasta;
 181         pDataArray->m->openInputFile(thisffastafile, inFFasta);
 182         pDataArray->m->openInputFile(thisrfastafile, inRFasta);
 183         if (thisfqualfile != "") {
 184             pDataArray->m->openInputFile(thisfqualfile, inFQual);
 185             pDataArray->m->openInputFile(thisrqualfile, inRQual);
 186         }
 187         pDataArray->m->openOutputFile(pDataArray->outputFasta, outFasta);
 188         pDataArray->m->openOutputFile(pDataArray->outputMisMatches, outMisMatch);
 189         pDataArray->m->openOutputFile(pDataArray->outputScrapFasta, outScrapFasta);
 190
 191         if (pDataArray->threadID == 0) {  outMisMatch << "Name\tLength\tOverlap_Length\tOverlap_Start\tOverlap_End\tMisMatches\tNum_Ns\n";  }
 192
 193         TrimOligos trimOligos(pDataArray->pdiffs, pDataArray->bdiffs, 0, 0, pDataArray->primers, pDataArray->barcodes);
 194
 195         while ((!inFFasta.eof()) && (!inRFasta.eof())) {
 196
 197             if (pDataArray->m->control_pressed) { break; }
 198
 199             int success = 1;
 200             string trashCode = "";
 201             int currentSeqsDiffs = 0;
 202
 203             //read seqs and quality info
 204             Sequence fSeq(inFFasta); pDataArray->m->gobble(inFFasta);
 205             Sequence rSeq(inRFasta); pDataArray->m->gobble(inRFasta);
 206             QualityScores* fQual = NULL; QualityScores* rQual = NULL;
 207             if (thisfqualfile != "") {
 208                 fQual = new QualityScores(inFQual); pDataArray->m->gobble(inFQual);
 209                 rQual = new QualityScores(inRQual); pDataArray->m->gobble(inRQual);
 210             }
 211
 212             int barcodeIndex = 0;
 213             int primerIndex = 0;
 214
 215             if(pDataArray->barcodes.size() != 0){
 216                 if (thisfqualfile != "") {
 217                     success = trimOligos.stripBarcode(fSeq, rSeq, *fQual, *rQual, barcodeIndex);
 218                 }else {
 219                     success = trimOligos.stripBarcode(fSeq, rSeq, barcodeIndex);
 220                 }
 221                 if(success > pDataArray->bdiffs)                {       trashCode += 'b';       }
 222                 else{ currentSeqsDiffs += success;  }
 223             }
 224
 225             if(pDataArray->primers.size() != 0){
 226                 if (thisfqualfile != "") {
 227                     success = trimOligos.stripForward(fSeq, rSeq, *fQual, *rQual, primerIndex);
 228                 }else {
 229                     success = trimOligos.stripForward(fSeq, rSeq, primerIndex);
 230                 }
 231                 if(success > pDataArray->pdiffs)                {       trashCode += 'f';       }
 232                 else{ currentSeqsDiffs += success;  }
 233             }
 234
 235             if (currentSeqsDiffs > pDataArray->tdiffs)  {       trashCode += 't';   }
 236
 237             //flip the reverse reads
 238             rSeq.reverseComplement();
 239             if (thisfqualfile != "") { rQual->flipQScores(); }
 240
 241             //pairwise align
 242             alignment->align(fSeq.getUnaligned(), rSeq.getUnaligned());
 243             map<int, int> ABaseMap = alignment->getSeqAAlnBaseMap();
 244             map<int, int> BBaseMap = alignment->getSeqBAlnBaseMap();
 245             fSeq.setAligned(alignment->getSeqAAln());
 246             rSeq.setAligned(alignment->getSeqBAln());
 247             int length = fSeq.getAligned().length();
 248
 249             //traverse alignments merging into one contiguous seq
 250             string contig = "";
 251             int numMismatches = 0;
 252             string seq1 = fSeq.getAligned();
 253             string seq2 = rSeq.getAligned();
 254             vector<int> scores1, scores2;
 255             if (thisfqualfile != "") {
 256                 scores1 = fQual->getQualityScores();
 257                 scores2 = rQual->getQualityScores();
 258                 delete fQual; delete rQual;
 259             }
 260
 261             int overlapStart = fSeq.getStartPos();
 262             int seq2Start = rSeq.getStartPos();
 263             //bigger of the 2 starting positions is the location of the overlapping start
 264             if (overlapStart < seq2Start) { //seq2 starts later so take from 0 to seq2Start from seq1
 265                 overlapStart = seq2Start;
 266                 for (int i = 0; i < overlapStart; i++) { contig += seq1[i];  }
 267             }else { //seq1 starts later so take from 0 to overlapStart from seq2
 268                 for (int i = 0; i < overlapStart; i++) {  contig += seq2[i]; }
 269             }
 270
 271             int seq1End = fSeq.getEndPos();
 272             int seq2End = rSeq.getEndPos();
 273             int overlapEnd = seq1End;
 274             if (seq2End < overlapEnd) { overlapEnd = seq2End; }  //smallest end position is where overlapping ends
 275
 276             int oStart = contig.length();
 277             for (int i = overlapStart; i < overlapEnd; i++) {
 278                 if (seq1[i] == seq2[i]) { //match, add base and choose highest score
 279                     contig += seq1[i];
 280                 }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 insert. In that case eliminate base
 281                     if (thisfqualfile != "") {
 282                         if (scores2[BBaseMap[i]] < pDataArray->insert) { } //
 283                         else { contig += seq2[i];  }
 284                     }else { contig += seq2[i]; } //with no quality info, then we keep it?
 285                 }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 insert. In that case eliminate base
 286                     if (thisfqualfile != "") {
 287                         if (scores1[ABaseMap[i]] < pDataArray->insert) { } //
 288                         else { contig += seq1[i];  }
 289                     }else { contig += seq1[i]; } //with no quality info, then we keep it?
 290                 }else if (((seq1[i] != '-') && (seq1[i] != '.')) && ((seq2[i] != '-') && (seq2[i] != '.'))) { //both bases choose one with better quality
 291                     if (thisfqualfile != "") {
 292                         if (abs(scores1[ABaseMap[i]] - scores2[BBaseMap[i]]) >= pDataArray->deltaq) { //is the difference in qual scores >= deltaq, if yes choose base with higher score
 293                             char c = seq1[i];
 294                             if (scores1[ABaseMap[i]] < scores2[BBaseMap[i]]) { c = seq2[i]; }
 295                             contig += c;
 296                         }else { //if no, base becomes n
 297                             contig += 'N';
 298                         }
 299                         numMismatches++;
 300                     }else { numMismatches++; } //cant decide, so eliminate and mark as mismatch
 301                 }else { //should never get here
 302                     pDataArray->m->mothurOut("[ERROR]: case I didn't think of seq1 = " + toString(seq1[i]) + " and seq2 = " + toString(seq2[i]) + "\n");
 303                 }
 304             }
 305             int oend = contig.length();
 306
 307             if (seq1End < seq2End) { //seq1 ends before seq2 so take from overlap to length from seq2
 308                 for (int i = overlapEnd; i < length; i++) { contig += seq2[i];  }
 309             }else { //seq2 ends before seq1 so take from overlap to length from seq1
 310                 for (int i = overlapEnd; i < length; i++) {  contig += seq1[i]; }
 311             }
 312
 313             if(trashCode.length() == 0){
 314                 bool ignore = false;
 315                 if (pDataArray->createGroup) {
 316                     if(pDataArray->barcodes.size() != 0){
 317                         string thisGroup = pDataArray->barcodeNameVector[barcodeIndex];
 318                         if (pDataArray->primers.size() != 0) {
 319                             if (pDataArray->primerNameVector[primerIndex] != "") {
 320                                 if(thisGroup != "") {
 321                                     thisGroup += "." + pDataArray->primerNameVector[primerIndex];
 322                                 }else {
 323                                     thisGroup = pDataArray->primerNameVector[primerIndex];
 324                                 }
 325                             }
 326                         }
 327
 328                         if (pDataArray->m->debug) { pDataArray->m->mothurOut(", group= " + thisGroup + "\n"); }
 329
 330                         int pos = thisGroup.find("ignore");
 331                         if (pos == string::npos) {
 332                             pDataArray->groupMap[fSeq.getName()] = thisGroup;
 333
 334                             map<string, int>::iterator it = pDataArray->groupCounts.find(thisGroup);
 335                             if (it == pDataArray->groupCounts.end()) {  pDataArray->groupCounts[thisGroup] = 1; }
 336                             else { pDataArray->groupCounts[it->first] ++; }
 337                         }else { ignore = true; }
 338                     }
 339                 }
 340
 341                 if(pDataArray->allFiles && !ignore){
 342                     ofstream output;
 343                     pDataArray->m->openOutputFileAppend(pDataArray->fastaFileNames[barcodeIndex][primerIndex], output);
 344                     output << ">" << fSeq.getName() << endl << contig << endl;
 345                     output.close();
 346                 }
 347
 348                 //output
 349                 outFasta << ">" << fSeq.getName() << endl << contig << endl;
 350                 int numNs = 0;
 351                 for (int i = 0; i < contig.length(); i++) { if (contig[i] == 'N') { numNs++; }  }
 352                 outMisMatch << fSeq.getName() << '\t' << contig.length() << '\t' << (oend-oStart) << '\t' << oStart << '\t' << oend << '\t' << numMismatches << '\t' << numNs << endl;
 353             }else {
 354                 //output
 355                 outScrapFasta << ">" << fSeq.getName() << " | " << trashCode << endl << contig << endl;
 356             }
 357             pDataArray->count++;
 358
 359                         //report progress
 360                         if((pDataArray->count) % 1000 == 0){    pDataArray->m->mothurOut(toString(pDataArray->count)); pDataArray->m->mothurOutEndLine();               }
 361                 }
 362
 363                 //report progress
 364                 if((pDataArray->count) % 1000 != 0){    pDataArray->m->mothurOut(toString(pDataArray->count)); pDataArray->m->mothurOutEndLine();               }
 365
 366         inFFasta.close();
 367         inRFasta.close();
 368         outFasta.close();
 369         outMisMatch.close();
 370         outScrapFasta.close();
 371         if (thisfqualfile != "") {
 372             inFQual.close();
 373             inRQual.close();
 374         }
 375         delete alignment;
 376
 377         pDataArray->done = true;
 378         if (pDataArray->m->control_pressed) {  pDataArray->m->mothurRemove(pDataArray->outputFasta);  pDataArray->m->mothurRemove(pDataArray->outputMisMatches);  pDataArray->m->mothurRemove(pDataArray->outputScrapFasta); }
 379
 380         return 0;
 381
 382         }
 383         catch(exception& e) {
 384                 pDataArray->m->errorOut(e, "AlignCommand", "MyContigsThreadFunction");
 385                 exit(1);
 386         }
 387 }
 388 #endif
 389
 390
 391 #endif