maligner.cpp

   1 /*
   2  *  maligner.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 9/23/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "maligner.h"
  11 #include "kmerdb.hpp"
  12 #include "blastdb.hpp"
  13
  14 /***********************************************************************/
  15 Maligner::Maligner(vector<Sequence*> temp, int num, int match, int misMatch, float div, int ms, int minCov, string mode, Database* dataLeft, Database* dataRight) :
  16                 db(temp), numWanted(num), matchScore(match), misMatchPenalty(misMatch), minDivR(div), minSimilarity(ms), minCoverage(minCov), searchMethod(mode), databaseLeft(dataLeft), databaseRight(dataRight) { m = MothurOut::getInstance(); }
  17 /***********************************************************************/
  18 string Maligner::getResults(Sequence* q, DeCalculator* decalc) {
  19         try {
  20
  21                 outputResults.clear();
  22
  23                 //make copy so trimming doesn't destroy query from calling class - remember to deallocate
  24                 query = new Sequence(q->getName(), q->getAligned());
  25
  26                 string chimera;
  27
  28                 if (searchMethod == "distance") {
  29                         //find closest seqs to query in template - returns copies of seqs so trim does not destroy - remember to deallocate
  30                         refSeqs = decalc->findClosest(query, db, numWanted, indexes);
  31                 }else if (searchMethod == "blast")  {
  32                         refSeqs = getBlastSeqs(query, numWanted); //fills indexes
  33                 }else if (searchMethod == "kmer") {
  34                         refSeqs = getKmerSeqs(query, numWanted); //fills indexes
  35                 }else { m->mothurOut("not valid search."); exit(1);  } //should never get here
  36
  37                 if (m->control_pressed) { return chimera;  }
  38
  39                 refSeqs = minCoverageFilter(refSeqs);
  40
  41                 if (refSeqs.size() < 2)  {
  42                         for (int i = 0; i < refSeqs.size(); i++) {  delete refSeqs[i];  }
  43                         percentIdenticalQueryChimera = 0.0;
  44                         return "unknown";
  45                 }
  46
  47                 int chimeraPenalty = computeChimeraPenalty();
  48
  49                 //fills outputResults
  50                 chimera = chimeraMaligner(chimeraPenalty, decalc);
  51
  52                 if (m->control_pressed) { return chimera;  }
  53
  54                 //free memory
  55                 delete query;
  56
  57                 for (int i = 0; i < refSeqs.size(); i++) {  delete refSeqs[i];  }
  58
  59                 return chimera;
  60         }
  61         catch(exception& e) {
  62                 m->errorOut(e, "Maligner", "getResults");
  63                 exit(1);
  64         }
  65 }
  66 /***********************************************************************/
  67 string Maligner::chimeraMaligner(int chimeraPenalty, DeCalculator* decalc) {
  68         try {
  69
  70                 string chimera;
  71
  72                 //trims seqs to first non gap char in all seqs and last non gap char in all seqs
  73                 spotMap = decalc->trimSeqs(query, refSeqs);
  74
  75                 vector<Sequence*> temp = refSeqs;
  76                 temp.push_back(query);
  77
  78                 verticalFilter(temp);
  79 //for (int i = 0; i < temp.size(); i++) { cout << temp[i]->getName() << '\n' << temp[i]->getAligned() << endl; } return "no";
  80
  81                 vector< vector<score_struct> > matrix = buildScoreMatrix(query->getAligned().length(), refSeqs.size()); //builds and initializes
  82
  83                 if (m->control_pressed) { return chimera;  }
  84
  85                 fillScoreMatrix(matrix, refSeqs, chimeraPenalty);
  86
  87                 vector<score_struct> path = extractHighestPath(matrix);
  88
  89                 if (m->control_pressed) { return chimera;  }
  90
  91                 vector<trace_struct> trace = mapTraceRegionsToAlignment(path, refSeqs);
  92
  93                 if (trace.size() > 1) {         chimera = "yes";        }
  94                 else { chimera = "no";  }
  95
  96                 int traceStart = path[0].col;
  97                 int traceEnd = path[path.size()-1].col;
  98                 string queryInRange = query->getAligned();
  99                 queryInRange = queryInRange.substr(traceStart, (traceEnd-traceStart+1));
 100
 101                 string chimeraSeq = constructChimericSeq(trace, refSeqs);
 102
 103                 percentIdenticalQueryChimera = computePercentID(queryInRange, chimeraSeq);
 104
 105                 if (m->control_pressed) { return chimera;  }
 106
 107                 //save output results
 108                 for (int i = 0; i < trace.size(); i++) {
 109                         int regionStart = trace[i].col;
 110                         int regionEnd = trace[i].oldCol;
 111                         int seqIndex = trace[i].row;
 112
 113                         results temp;
 114
 115                         temp.parent = refSeqs[seqIndex]->getName();
 116                         temp.parentAligned = db[indexes[seqIndex]]->getAligned();
 117                         temp.nastRegionStart = spotMap[regionStart];
 118                         temp.nastRegionEnd = spotMap[regionEnd];
 119                         temp.regionStart = regionStart;
 120                         temp.regionEnd = regionEnd;
 121
 122                         string parentInRange = refSeqs[seqIndex]->getAligned();
 123                         parentInRange = parentInRange.substr(traceStart, (traceEnd-traceStart+1));
 124
 125                         temp.queryToParent = computePercentID(queryInRange, parentInRange);
 126                         temp.divR = (percentIdenticalQueryChimera / temp.queryToParent);
 127
 128                         string queryInRegion = query->getAligned();
 129                         queryInRegion = queryInRegion.substr(regionStart, (regionEnd-regionStart+1));
 130
 131                         string parentInRegion = refSeqs[seqIndex]->getAligned();
 132                         parentInRegion = parentInRegion.substr(regionStart, (regionEnd-regionStart+1));
 133
 134                         temp.queryToParentLocal = computePercentID(queryInRegion, parentInRegion);
 135
 136                         outputResults.push_back(temp);
 137                 }
 138
 139                 return chimera;
 140         }
 141         catch(exception& e) {
 142                 m->errorOut(e, "Maligner", "chimeraMaligner");
 143                 exit(1);
 144         }
 145 }
 146 /***********************************************************************/
 147 //removes top matches that do not have minimum coverage with query.
 148 vector<Sequence*> Maligner::minCoverageFilter(vector<Sequence*> ref){
 149         try {
 150                 vector<Sequence*> newRefs;
 151
 152                 string queryAligned = query->getAligned();
 153
 154                 for (int i = 0; i < ref.size(); i++) {
 155
 156                         string refAligned = ref[i]->getAligned();
 157
 158                         int numBases = 0;
 159                         int numCovered = 0;
 160
 161                         //calculate coverage
 162                         for (int j = 0; j < queryAligned.length(); j++) {
 163
 164                                 if (isalpha(queryAligned[j])) {
 165                                         numBases++;
 166
 167                                         if (isalpha(refAligned[j])) {
 168                                                 numCovered++;
 169                                         }
 170                                 }
 171                         }
 172
 173                         int coverage = ((numCovered/(float)numBases)*100);
 174
 175                         //if coverage above minimum
 176                         if (coverage > minCoverage) {
 177                                 newRefs.push_back(ref[i]);
 178                         }
 179                 }
 180
 181                 return newRefs;
 182         }
 183         catch(exception& e) {
 184                 m->errorOut(e, "Maligner", "minCoverageFilter");
 185                 exit(1);
 186         }
 187 }
 188 /***********************************************************************/
 189 // a breakpoint should yield fewer mismatches than this number with respect to the best parent sequence.
 190 int Maligner::computeChimeraPenalty() {
 191         try {
 192
 193                 int numAllowable = ((1.0 - (1.0/minDivR)) * query->getNumBases());
 194
 195                 int penalty = int(numAllowable + 1) * misMatchPenalty;
 196
 197                 return penalty;
 198
 199         }
 200         catch(exception& e) {
 201                 m->errorOut(e, "Maligner", "computeChimeraPenalty");
 202                 exit(1);
 203         }
 204 }
 205 /***********************************************************************/
 206 //this is a vertical filter
 207 void Maligner::verticalFilter(vector<Sequence*> seqs) {
 208         try {
 209                 vector<int> gaps;       gaps.resize(query->getAligned().length(), 0);
 210
 211                 string filterString = (string(query->getAligned().length(), '1'));
 212
 213                 //for each sequence
 214                 for (int i = 0; i < seqs.size(); i++) {
 215
 216                         string seqAligned = seqs[i]->getAligned();
 217
 218                         for (int j = 0; j < seqAligned.length(); j++) {
 219                                 //if this spot is a gap
 220                                 if ((seqAligned[j] == '-') || (seqAligned[j] == '.'))   {       gaps[j]++;      }
 221                         }
 222                 }
 223
 224                 //zero out spot where all sequences have blanks
 225                 int numColRemoved = 0;
 226                 for(int i = 0; i < seqs[0]->getAligned().length(); i++){
 227                         if(gaps[i] == seqs.size())      {       filterString[i] = '0';  numColRemoved++;  }
 228                 }
 229
 230                 map<int, int> newMap;
 231                 //for each sequence
 232                 for (int i = 0; i < seqs.size(); i++) {
 233
 234                         string seqAligned = seqs[i]->getAligned();
 235                         string newAligned = "";
 236                         int count = 0;
 237
 238                         for (int j = 0; j < seqAligned.length(); j++) {
 239                                 //if this spot is not a gap
 240                                 if (filterString[j] == '1') {
 241                                         newAligned += seqAligned[j];
 242                                         newMap[count] = spotMap[j];
 243                                         count++;
 244                                 }
 245                         }
 246
 247                         seqs[i]->setAligned(newAligned);
 248                 }
 249
 250                 spotMap = newMap;
 251         }
 252         catch(exception& e) {
 253                 m->errorOut(e, "Maligner", "verticalFilter");
 254                 exit(1);
 255         }
 256 }
 257 //***************************************************************************************************************
 258 vector< vector<score_struct> > Maligner::buildScoreMatrix(int cols, int rows) {
 259         try{
 260
 261                 vector< vector<score_struct> > m; m.resize(rows);
 262
 263                 for (int i = 0; i < m.size(); i++) {
 264                         for (int j = 0; j < cols; j++) {
 265
 266                                 //initialize each cell
 267                                 score_struct temp;
 268                                 temp.prev = -1;
 269                                 temp.score = -9999999;
 270                                 temp.col = j;
 271                                 temp.row = i;
 272
 273                                 m[i].push_back(temp);
 274                         }
 275                 }
 276
 277                 return m;
 278         }
 279         catch(exception& e) {
 280                 m->errorOut(e, "Maligner", "buildScoreMatrix");
 281                 exit(1);
 282         }
 283 }
 284 //***************************************************************************************************************
 285 void Maligner::fillScoreMatrix(vector<vector<score_struct> >& ms, vector<Sequence*> seqs, int penalty) {
 286         try{
 287
 288                 //get matrix dimensions
 289                 int numCols = query->getAligned().length();
 290                 int numRows = seqs.size();
 291
 292                 //initialize first col
 293                 string queryAligned = query->getAligned();
 294                 for (int i = 0; i < numRows; i++) {
 295                         string subjectAligned = seqs[i]->getAligned();
 296
 297                         //are you both gaps?
 298                         if ((!isalpha(queryAligned[0])) && (!isalpha(subjectAligned[0]))) {
 299                                 ms[i][0].score = 0;
 300                         }else if (queryAligned[0] == subjectAligned[0]) {
 301                                 ms[i][0].score = matchScore;
 302                         }else{
 303                                 ms[i][0].score = 0;
 304                         }
 305                 }
 306
 307                 //fill rest of matrix
 308                 for (int j = 1; j < numCols; j++) {  //iterate through matrix columns
 309
 310                         for (int i = 0; i < numRows; i++) {  //iterate through matrix rows
 311
 312                                 string subjectAligned = seqs[i]->getAligned();
 313
 314                                 int matchMisMatchScore = 0;
 315                                 //are you both gaps?
 316                                 if ((!isalpha(queryAligned[j])) && (!isalpha(subjectAligned[j]))) {
 317                                         //leave the same
 318                                 }else if ((toupper(queryAligned[j]) == 'N') || (toupper(subjectAligned[j]) == 'N')) {
 319                                         //leave the same
 320                                 }else if (queryAligned[j] == subjectAligned[j]) {
 321                                         matchMisMatchScore = matchScore;
 322                                 }else if (queryAligned[j] != subjectAligned[j]) {
 323                                         matchMisMatchScore = misMatchPenalty;
 324                                 }
 325
 326                                 //compute score based on previous columns scores
 327                                 for (int prevIndex = 0; prevIndex < numRows; prevIndex++) { //iterate through rows
 328
 329                                         int sumScore = matchMisMatchScore + ms[prevIndex][j-1].score;
 330
 331                                         //you are not at yourself
 332                                         if (prevIndex != i) {   sumScore += penalty;    }
 333                                         if (sumScore < 0)       {       sumScore = 0;                   }
 334
 335                                         if (sumScore > ms[i][j].score) {
 336                                                 ms[i][j].score = sumScore;
 337                                                 ms[i][j].prev = prevIndex;
 338                                         }
 339                                 }
 340                         }
 341                 }
 342
 343         }
 344         catch(exception& e) {
 345                 m->errorOut(e, "Maligner", "fillScoreMatrix");
 346                 exit(1);
 347         }
 348 }
 349 //***************************************************************************************************************
 350 vector<score_struct> Maligner::extractHighestPath(vector<vector<score_struct> > ms) {
 351         try {
 352
 353                 //get matrix dimensions
 354                 int numCols = query->getAligned().length();
 355                 int numRows = ms.size();
 356
 357
 358                 //find highest score scoring matrix
 359                 score_struct highestStruct;
 360                 int highestScore = 0;
 361
 362                 for (int i = 0; i < numRows; i++) {
 363                         for (int j = 0; j < numCols; j++) {
 364                                 if (ms[i][j].score > highestScore) {
 365                                         highestScore = ms[i][j].score;
 366                                         highestStruct = ms[i][j];
 367                                 }
 368                         }
 369                 }
 370
 371                 vector<score_struct> path;
 372
 373                 int rowIndex = highestStruct.row;
 374                 int pos = highestStruct.col;
 375                 int score = highestStruct.score;
 376
 377                 while (pos >= 0 && score > 0) {
 378                         score_struct temp = ms[rowIndex][pos];
 379                         score = temp.score;
 380
 381                         if (score > 0) {        path.push_back(temp);   }
 382
 383                         rowIndex = temp.prev;
 384                         pos--;
 385                 }
 386
 387                 reverse(path.begin(), path.end());
 388
 389                 return path;
 390
 391         }
 392         catch(exception& e) {
 393                 m->errorOut(e, "Maligner", "extractHighestPath");
 394                 exit(1);
 395         }
 396 }
 397 //***************************************************************************************************************
 398 vector<trace_struct> Maligner::mapTraceRegionsToAlignment(vector<score_struct> path, vector<Sequence*> seqs) {
 399         try {
 400                 vector<trace_struct> trace;
 401
 402                 int region_index = path[0].row;
 403                 int region_start = path[0].col;
 404
 405                 for (int i = 1; i < path.size(); i++) {
 406
 407                         int next_region_index = path[i].row;
 408
 409                         if (next_region_index != region_index) {
 410
 411                                 // add trace region
 412                                 int col_index = path[i].col;
 413                                 trace_struct temp;
 414                                 temp.col = region_start;
 415                                 temp.oldCol = col_index-1;
 416                                 temp.row = region_index;
 417
 418                                 trace.push_back(temp);
 419
 420                                 region_index = path[i].row;
 421                                 region_start = col_index;
 422                         }
 423                 }
 424
 425                 // get last one
 426                 trace_struct temp;
 427                 temp.col = region_start;
 428                 temp.oldCol = path[path.size()-1].col;
 429                 temp.row = region_index;
 430                 trace.push_back(temp);
 431
 432                 return trace;
 433
 434         }
 435         catch(exception& e) {
 436                 m->errorOut(e, "Maligner", "mapTraceRegionsToAlignment");
 437                 exit(1);
 438         }
 439 }
 440 //***************************************************************************************************************
 441 string Maligner::constructChimericSeq(vector<trace_struct> trace, vector<Sequence*> seqs) {
 442         try {
 443                 string chimera = "";
 444
 445                 for (int i = 0; i < trace.size(); i++) {
 446                         string seqAlign = seqs[trace[i].row]->getAligned();
 447                         seqAlign = seqAlign.substr(trace[i].col, (trace[i].oldCol-trace[i].col+1));
 448                         chimera += seqAlign;
 449                 }
 450
 451                 return chimera;
 452         }
 453         catch(exception& e) {
 454                 m->errorOut(e, "Maligner", "constructChimericSeq");
 455                 exit(1);
 456         }
 457 }
 458 //***************************************************************************************************************
 459 float Maligner::computePercentID(string queryAlign, string chimera) {
 460         try {
 461
 462                 if (queryAlign.length() != chimera.length()) {
 463                         m->mothurOut("Error, alignment strings are of different lengths: "); m->mothurOutEndLine();
 464                         m->mothurOut(toString(queryAlign.length())); m->mothurOutEndLine(); m->mothurOutEndLine(); m->mothurOutEndLine(); m->mothurOutEndLine();
 465                         m->mothurOut(toString(chimera.length())); m->mothurOutEndLine();
 466                         return -1.0;
 467                 }
 468
 469
 470                 int numBases = 0;
 471                 int numIdentical = 0;
 472
 473                 for (int i = 0; i < queryAlign.length(); i++) {
 474                         if ((isalpha(queryAlign[i])) || (isalpha(chimera[i])))  {
 475                                 numBases++;
 476                                 if (queryAlign[i] == chimera[i]) {
 477                                         numIdentical++;
 478                                 }
 479                         }
 480                 }
 481
 482                 if (numBases == 0) { return 0; }
 483
 484                 float percentIdentical = (numIdentical/(float)numBases) * 100;
 485
 486                 return percentIdentical;
 487
 488         }
 489         catch(exception& e) {
 490                 m->errorOut(e, "Maligner", "computePercentID");
 491                 exit(1);
 492         }
 493 }
 494 //***************************************************************************************************************
 495 vector<Sequence*> Maligner::getBlastSeqs(Sequence* q, int num) {
 496         try {
 497                 indexes.clear();
 498                 vector<Sequence*> refResults;
 499
 500                 //get parts of query
 501                 string queryUnAligned = q->getUnaligned();
 502                 string leftQuery = queryUnAligned.substr(0, int(queryUnAligned.length() * 0.33)); //first 1/3 of the sequence
 503                 string rightQuery = queryUnAligned.substr(int(queryUnAligned.length() * 0.66)); //last 1/3 of the sequence
 504
 505                 Sequence* queryLeft = new Sequence(q->getName(), leftQuery);
 506                 Sequence* queryRight = new Sequence(q->getName(), rightQuery);
 507
 508                 vector<int> tempIndexesRight = databaseLeft->findClosestMegaBlast(queryRight, num+1);
 509                 vector<int> tempIndexesLeft = databaseLeft->findClosestMegaBlast(queryLeft, num+1);
 510
 511                 //if ((tempIndexesRight.size() != (num+1)) || (tempIndexesLeft.size() != (num+1)))  {  m->mothurOut("megablast returned " + toString(tempIndexesRight.size()) + " results for the right end, and " + toString(tempIndexesLeft.size()) + " for the left end. Needed " + toString(num+1) + ". Unable to porcess sequence " + q->getName()); m->mothurOutEndLine(); return refResults; }
 512
 513                 vector<int> smaller;
 514                 vector<int> larger;
 515
 516                 if (tempIndexesRight.size() < tempIndexesLeft.size()) { smaller = tempIndexesRight;  larger = tempIndexesLeft; }
 517                 else { smaller = tempIndexesLeft; larger = tempIndexesRight; }
 518
 519                 //merge results
 520                 map<int, int> seen;
 521                 map<int, int>::iterator it;
 522
 523                 vector<int> mergedResults;
 524                 for (int i = 0; i < smaller.size(); i++) {
 525                         //add left if you havent already
 526                         it = seen.find(smaller[i]);
 527                         if (it == seen.end()) {
 528                                 mergedResults.push_back(smaller[i]);
 529                                 seen[smaller[i]] = smaller[i];
 530                         }
 531
 532                         //add right if you havent already
 533                         it = seen.find(larger[i]);
 534                         if (it == seen.end()) {
 535                                 mergedResults.push_back(larger[i]);
 536                                 seen[larger[i]] = larger[i];
 537                         }
 538                 }
 539
 540                 for (int i = smaller.size(); i < larger.size(); i++) {
 541                         it = seen.find(larger[i]);
 542                         if (it == seen.end()) {
 543                                 mergedResults.push_back(larger[i]);
 544                                 seen[larger[i]] = larger[i];
 545                         }
 546                 }
 547
 548                 if (mergedResults.size() < numWanted) { numWanted = mergedResults.size(); }
 549 //cout << q->getName() <<  endl;
 550                 for (int i = 0; i < numWanted; i++) {
 551 //cout << db[mergedResults[i]]->getName() << endl;
 552                         if (db[mergedResults[i]]->getName() != q->getName()) {
 553                                 Sequence* temp = new Sequence(db[mergedResults[i]]->getName(), db[mergedResults[i]]->getAligned());
 554                                 refResults.push_back(temp);
 555                                 indexes.push_back(mergedResults[i]);
 556                         }
 557 //cout << mergedResults[i] << endl;
 558                 }
 559 //cout << endl;
 560                 delete queryRight;
 561                 delete queryLeft;
 562
 563                 return refResults;
 564         }
 565         catch(exception& e) {
 566                 m->errorOut(e, "Maligner", "getBlastSeqs");
 567                 exit(1);
 568         }
 569 }
 570 //***************************************************************************************************************
 571 vector<Sequence*> Maligner::getKmerSeqs(Sequence* q, int num) {
 572         try {
 573                 indexes.clear();
 574
 575                 //get parts of query
 576                 string queryUnAligned = q->getUnaligned();
 577                 string leftQuery = queryUnAligned.substr(0, int(queryUnAligned.length() * 0.33)); //first 1/3 of the sequence
 578                 string rightQuery = queryUnAligned.substr(int(queryUnAligned.length() * 0.66)); //last 1/3 of the sequence
 579
 580                 Sequence* queryLeft = new Sequence(q->getName(), leftQuery);
 581                 Sequence* queryRight = new Sequence(q->getName(), rightQuery);
 582
 583                 vector<int> tempIndexesLeft = databaseLeft->findClosestSequences(queryLeft, numWanted);
 584                 vector<int> tempIndexesRight = databaseRight->findClosestSequences(queryRight, numWanted);
 585
 586                 //merge results
 587                 map<int, int> seen;
 588                 map<int, int>::iterator it;
 589
 590                 vector<int> mergedResults;
 591                 for (int i = 0; i < tempIndexesLeft.size(); i++) {
 592                         //add left if you havent already
 593                         it = seen.find(tempIndexesLeft[i]);
 594                         if (it == seen.end()) {
 595                                 mergedResults.push_back(tempIndexesLeft[i]);
 596                                 seen[tempIndexesLeft[i]] = tempIndexesLeft[i];
 597                         }
 598
 599                         //add right if you havent already
 600                         it = seen.find(tempIndexesRight[i]);
 601                         if (it == seen.end()) {
 602                                 mergedResults.push_back(tempIndexesRight[i]);
 603                                 seen[tempIndexesRight[i]] = tempIndexesRight[i];
 604                         }
 605                 }
 606
 607 //cout << q->getName() << endl;
 608                 vector<Sequence*> refResults;
 609                 for (int i = 0; i < numWanted; i++) {
 610 //cout << db[mergedResults[i]]->getName() << endl;
 611                         Sequence* temp = new Sequence(db[mergedResults[i]]->getName(), db[mergedResults[i]]->getAligned());
 612                         refResults.push_back(temp);
 613                         indexes.push_back(mergedResults[i]);
 614                 }
 615 //cout << endl;
 616                 delete queryRight;
 617                 delete queryLeft;
 618
 619                 return refResults;
 620         }
 621         catch(exception& e) {
 622                 m->errorOut(e, "Maligner", "getBlastSeqs");
 623                 exit(1);
 624         }
 625 }
 626 //***************************************************************************************************************
 627