ccode.cpp

   1 /*
   2  *  ccode.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 8/24/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "ccode.h"
  11 #include "ignoregaps.h"
  12 #include "eachgapdist.h"
  13
  14
  15 //***************************************************************************************************************
  16 Ccode::Ccode(string filename, string o) {
  17         fastafile = filename;  outputDir = o;
  18         distCalc = new eachGapDist();
  19         decalc = new DeCalculator();
  20
  21         mapInfo = outputDir + getRootName(getSimpleName(fastafile)) + "mapinfo";
  22         ofstream out2;
  23         openOutputFile(mapInfo, out2);
  24
  25         out2 << "Place in masked, filtered and trimmed sequence\tPlace in original alignment" << endl;
  26         out2.close();
  27 }
  28 //***************************************************************************************************************
  29 Ccode::~Ccode() {
  30         delete distCalc;
  31         delete decalc;
  32 }
  33 //***************************************************************************************************************
  34 void Ccode::printHeader(ostream& out) {
  35         out << "For full window mapping info refer to " << mapInfo << endl << endl;
  36 }
  37 //***************************************************************************************************************
  38 void Ccode::print(ostream& out, ostream& outAcc) {
  39         try {
  40
  41                 ofstream out2;
  42                 openOutputFileAppend(mapInfo, out2);
  43
  44                 out2 << querySeq->getName() << endl;
  45                 for (it = spotMap.begin(); it!= spotMap.end(); it++) {
  46                         out2 << it->first << '\t' << it->second << endl;
  47                 }
  48                 out2.close();
  49                 out << querySeq->getName() << endl << endl << "Reference sequences used and distance to query:" << endl;
  50
  51                 for (int j = 0; j < closest.size(); j++) {
  52                         out << closest[j].seq->getName() << '\t' << closest[j].dist << endl;
  53                 }
  54                 out << endl << endl;
  55
  56                 //for each window
  57                 //window mapping info.
  58                 out << "Mapping information: ";
  59                 //you mask and did not filter
  60                 if ((seqMask != "") && (!filter)) { out << "mask and trim."; }
  61
  62                 //you filtered and did not mask
  63                 if ((seqMask == "") && (filter)) { out << "filter and trim."; }
  64
  65                 //you masked and filtered
  66                 if ((seqMask != "") && (filter)) { out << "mask, filter and trim."; }
  67
  68                 out << endl << "Window\tStartPos\tEndPos" << endl;
  69                 it = trim.begin();
  70                 for (int k = 0; k < windows.size()-1; k++) {
  71                         out << k+1 << '\t' << spotMap[windows[k]-it->first] << '\t' << spotMap[windows[k]-it->first+windowSizes] << endl;
  72                 }
  73
  74                 out << windows.size() << '\t' << spotMap[windows[windows.size()-1]-it->first] << '\t' << spotMap[it->second-it->first-1] << endl;
  75                 out << endl;
  76                 out << "Window\tAvgQ\t(sdQ)\tAvgR\t(sdR)\tRatio\tAnova" << endl;
  77                 for (int k = 0; k < windows.size(); k++) {
  78                         float ds = averageQuery[k] / averageRef[k];
  79                         out << k+1 << '\t' << averageQuery[k] << '\t' << sdQuery[k] << '\t' << averageRef[k] << '\t'<< sdRef[k] << '\t' << ds << '\t' << anova[k] << endl;
  80                 }
  81                 out << endl;
  82
  83                 //varRef
  84                 //varQuery
  85                 /* F test for differences among variances.
  86                 * varQuery is expected to be higher or similar than varRef */
  87                 //float fs = varQuery[query] / varRef[query];   /* F-Snedecor, test for differences of variances */
  88
  89                 bool results = false;
  90
  91                 //confidence limit, t - Student, anova
  92                 out << "Window\tConfidenceLimit\tt-Student\tAnova" << endl;
  93
  94                 for (int k = 0; k < windows.size(); k++) {
  95                         string temp = "";
  96                         if (isChimericConfidence[k]) {  temp += "*\t"; }
  97                         else { temp += "\t"; }
  98
  99                         if (isChimericTStudent[k]) {  temp += "*\t"; }
 100                         else { temp += "\t"; }
 101
 102                         if (isChimericANOVA[k]) {  temp += "*\t"; }
 103                         else { temp += "\t"; }
 104
 105                         out << k+1 << '\t' << temp << endl;
 106
 107                         if (temp == "*\t*\t*\t") {  results = true;  }
 108                 }
 109                 out << endl;
 110
 111                 if (results) {
 112                         m->mothurOut(querySeq->getName() + " was found have at least one chimeric window."); m->mothurOutEndLine();
 113                         outAcc << querySeq->getName() << endl;
 114                 }
 115
 116                 //free memory
 117                 for (int i = 0; i < closest.size(); i++) {  delete closest[i].seq;  }
 118
 119
 120         }
 121         catch(exception& e) {
 122                 m->errorOut(e, "Ccode", "print");
 123                 exit(1);
 124         }
 125 }
 126 //***************************************************************************************************************
 127 int Ccode::getChimeras(Sequence* query) {
 128         try {
 129
 130                 closest.clear();
 131                 refCombo = 0;
 132                 sumRef.clear();
 133                 varRef.clear();
 134                 varQuery.clear();
 135                 sdRef.clear();
 136                 sdQuery.clear();
 137                 sumQuery.clear();
 138                 sumSquaredRef.clear();
 139                 sumSquaredQuery.clear();
 140                 averageRef.clear();
 141                 averageQuery.clear();
 142                 anova.clear();
 143                 isChimericConfidence.clear();
 144                 isChimericTStudent.clear();
 145                 isChimericANOVA.clear();
 146                 trim.clear();
 147                 spotMap.clear();
 148                 windowSizes = window;
 149                 windows.clear();
 150
 151
 152                 querySeq = query;
 153
 154                 //find closest matches to query
 155                 closest = findClosest(query, numWanted);
 156
 157                 //initialize spotMap
 158                 for (int i = 0; i < query->getAligned().length(); i++) {        spotMap[i] = i;         }
 159
 160                 //mask sequences if the user wants to
 161                 if (seqMask != "") {
 162                         decalc->setMask(seqMask);
 163
 164                         decalc->runMask(query);
 165
 166                         //mask closest
 167                         for (int i = 0; i < closest.size(); i++) {      decalc->runMask(closest[i].seq);        }
 168
 169                         spotMap = decalc->getMaskMap();
 170                 }
 171
 172                 if (filter) {
 173                         vector<Sequence*> temp;
 174                         for (int i = 0; i < closest.size(); i++) { temp.push_back(closest[i].seq);  }
 175                         temp.push_back(query);
 176
 177                         createFilter(temp, 0.5);
 178
 179                         for (int i = 0; i < temp.size(); i++) { runFilter(temp[i]);  }
 180
 181                         //update spotMap
 182                         map<int, int> newMap;
 183                         int spot = 0;
 184
 185                         for (int i = 0; i < filterString.length(); i++) {
 186                                 if (filterString[i] == '1') {
 187                                         //add to newMap
 188                                         newMap[spot] = spotMap[i];
 189                                         spot++;
 190                                 }
 191                         }
 192                         spotMap = newMap;
 193                 }
 194
 195                 //trim sequences - this follows ccodes remove_extra_gaps
 196                 trimSequences(query);
 197
 198
 199                 //windows are equivalent to words - ccode paper recommends windows are between 5% and 20% on alignment length().
 200                 //Our default will be 10% and we will warn if user tries to use a window above or below these recommendations
 201                 windows = findWindows();
 202
 203
 204                 //remove sequences that are more than 20% different and less than 0.5% different - may want to allow user to specify this later
 205                 removeBadReferenceSeqs(closest);
 206
 207
 208                 //find the averages for each querys references
 209                 getAverageRef(closest);  //fills sumRef, averageRef, sumSquaredRef and refCombo.
 210                 getAverageQuery(closest, query);  //fills sumQuery, averageQuery, sumSquaredQuery.
 211
 212
 213                 //find the averages for each querys references
 214                 findVarianceRef();  //fills varRef and sdRef also sets minimum error rate to 0.001 to avoid divide by 0.
 215
 216
 217                 //find the averages for the query
 218                 findVarianceQuery();  //fills varQuery and sdQuery also sets minimum error rate to 0.001 to avoid divide by 0.
 219
 220                 determineChimeras();  //fills anova, isChimericConfidence, isChimericTStudent and isChimericANOVA.
 221
 222                 return 0;
 223         }
 224         catch(exception& e) {
 225                 m->errorOut(e, "Ccode", "getChimeras");
 226                 exit(1);
 227         }
 228 }
 229 /***************************************************************************************************************/
 230 //ccode algo says it does this to "Removes the initial and final gaps to avoid biases due to incomplete sequences."
 231 void Ccode::trimSequences(Sequence* query) {
 232         try {
 233
 234                 int frontPos = 0;  //should contain first position in all seqs that is not a gap character
 235                 int rearPos = query->getAligned().length();
 236
 237                 //********find first position in closest seqs that is a non gap character***********//
 238                 //find first position all query seqs that is a non gap character
 239                 for (int i = 0; i < closest.size(); i++) {
 240
 241                         string aligned = closest[i].seq->getAligned();
 242                         int pos = 0;
 243
 244                         //find first spot in this seq
 245                         for (int j = 0; j < aligned.length(); j++) {
 246                                 if (isalpha(aligned[j])) {
 247                                         pos = j;
 248                                         break;
 249                                 }
 250                         }
 251
 252                         //save this spot if it is the farthest
 253                         if (pos > frontPos) { frontPos = pos; }
 254                 }
 255
 256                 //find first position all querySeq[query] that is a non gap character
 257                 string aligned = query->getAligned();
 258                 int pos = 0;
 259
 260                 //find first spot in this seq
 261                 for (int j = 0; j < aligned.length(); j++) {
 262                         if (isalpha(aligned[j])) {
 263                                 pos = j;
 264                                 break;
 265                         }
 266                 }
 267
 268                 //save this spot if it is the farthest
 269                 if (pos > frontPos) { frontPos = pos; }
 270
 271
 272                 //********find last position in closest seqs that is a non gap character***********//
 273                 for (int i = 0; i < closest.size(); i++) {
 274
 275                         string aligned = closest[i].seq->getAligned();
 276                         int pos = aligned.length();
 277
 278                         //find first spot in this seq
 279                         for (int j = aligned.length()-1; j >= 0; j--) {
 280                                 if (isalpha(aligned[j])) {
 281                                         pos = j;
 282                                         break;
 283                                 }
 284                         }
 285
 286                         //save this spot if it is the farthest
 287                         if (pos < rearPos) { rearPos = pos; }
 288                 }
 289
 290                 //find last position all querySeqs[query] that is a non gap character
 291                 aligned = query->getAligned();
 292                 pos = aligned.length();
 293
 294                 //find first spot in this seq
 295                 for (int j = aligned.length()-1; j >= 0; j--) {
 296                         if (isalpha(aligned[j])) {
 297                                 pos = j;
 298                                 break;
 299                         }
 300                 }
 301
 302                 //save this spot if it is the farthest
 303                 if (pos < rearPos) { rearPos = pos; }
 304
 305
 306                 //check to make sure that is not whole seq
 307                 if ((rearPos - frontPos - 1) <= 0) {  m->mothurOut("Error, when I trim your sequences, the entire sequence is trimmed."); m->mothurOutEndLine(); exit(1);  }
 308
 309                 map<int, int> tempTrim;
 310                 tempTrim[frontPos] = rearPos;
 311
 312                 //save trimmed locations
 313                 trim = tempTrim;
 314
 315                 //update spotMask
 316                 map<int, int> newMap;
 317                 int spot = 0;
 318
 319                 for (int i = frontPos; i < rearPos; i++) {
 320                         //add to newMap
 321                         newMap[spot] = spotMap[i];
 322                         spot++;
 323                 }
 324                 spotMap = newMap;
 325         }
 326         catch(exception& e) {
 327                 m->errorOut(e, "Ccode", "trimSequences");
 328                 exit(1);
 329         }
 330 }
 331 /***************************************************************************************************************/
 332 vector<int> Ccode::findWindows() {
 333         try {
 334
 335                 vector<int> win;
 336                 it = trim.begin();
 337
 338                 int length = it->second - it->first;
 339
 340                 //default is wanted = 10% of total length
 341                 if (windowSizes > length) {
 342                         m->mothurOut("You have slected a window larger than your sequence length after all filters, masks and trims have been done. I will use the default 10% of sequence length.");
 343                         windowSizes = length / 10;
 344                 }else if (windowSizes == 0) { windowSizes = length / 10;  }
 345                 else if (windowSizes > (length * 0.20)) {
 346                         m->mothurOut("You have selected a window that is larger than 20% of your sequence length.  This is not recommended, but I will continue anyway."); m->mothurOutEndLine();
 347                 }else if (windowSizes < (length * 0.05)) {
 348                         m->mothurOut("You have selected a window that is smaller than 5% of your sequence length.  This is not recommended, but I will continue anyway."); m->mothurOutEndLine();
 349                 }
 350
 351                 //save starting points of each window
 352                 for (int m = it->first;  m < (it->second-windowSizes); m+=windowSizes) {  win.push_back(m);  }
 353
 354                 //save last window
 355                 if (win[win.size()-1] < (it->first+length)) {
 356                         win.push_back(win[win.size()-1]+windowSizes); // ex. string length is 115, window is 25, without this you would get 0, 25, 50, 75
 357                 }                                                                                                                                                                                                       //with this you would get 1,25,50,75,100
 358
 359                 return win;
 360         }
 361         catch(exception& e) {
 362                 m->errorOut(e, "Ccode", "findWindows");
 363                 exit(1);
 364         }
 365 }
 366 //***************************************************************************************************************
 367 int Ccode::getDiff(string seqA, string seqB) {
 368         try {
 369
 370                 int numDiff = 0;
 371
 372                 for (int i = 0; i < seqA.length(); i++) {
 373                         //if you are both not gaps
 374                         //if (isalpha(seqA[i]) && isalpha(seqA[i])) {
 375                                 //are you different
 376                                 if (seqA[i] != seqB[i]) {
 377                                          int ok; /* ok=1 means equivalent base. Checks for degenerate bases */
 378
 379                                         /* the char in base_a and base_b have been checked and they are different */
 380                                         if ((seqA[i] == 'N') && (seqB[i] != '-')) ok = 1;
 381                                         else if ((seqB[i] == 'N') && (seqA[i] != '-')) ok = 1;
 382                                         else if ((seqA[i] == 'Y') && ((seqB[i] == 'C') || (seqB[i] == 'T'))) ok = 1;
 383                                         else if ((seqB[i] == 'Y') && ((seqA[i] == 'C') || (seqA[i] == 'T'))) ok = 1;
 384                                         else if ((seqA[i] == 'R') && ((seqB[i] == 'G') || (seqB[i] == 'A'))) ok = 1;
 385                                         else if ((seqB[i] == 'R') && ((seqA[i] == 'G') || (seqA[i] == 'A'))) ok = 1;
 386                                         else if ((seqA[i] == 'S') && ((seqB[i] == 'C') || (seqB[i] == 'G'))) ok = 1;
 387                                         else if ((seqB[i] == 'S') && ((seqA[i] == 'C') || (seqA[i] == 'G'))) ok = 1;
 388                                         else if ((seqA[i] == 'W') && ((seqB[i] == 'T') || (seqB[i] == 'A'))) ok = 1;
 389                                         else if ((seqB[i] == 'W') && ((seqA[i] == 'T') || (seqA[i] == 'A'))) ok = 1;
 390                                         else if ((seqA[i] == 'M') && ((seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
 391                                         else if ((seqB[i] == 'M') && ((seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
 392                                         else if ((seqA[i] == 'K') && ((seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
 393                                         else if ((seqB[i] == 'K') && ((seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
 394                                         else if ((seqA[i] == 'V') && ((seqB[i] == 'C') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
 395                                         else if ((seqB[i] == 'V') && ((seqA[i] == 'C') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
 396                                         else if ((seqA[i] == 'H') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
 397                                         else if ((seqB[i] == 'H') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
 398                                         else if ((seqA[i] == 'D') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
 399                                         else if ((seqB[i] == 'D') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
 400                                         else if ((seqA[i] == 'B') && ((seqB[i] == 'C') || (seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
 401                                         else if ((seqB[i] == 'B') && ((seqA[i] == 'C') || (seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
 402                                         else ok = 0;  /* the bases are different and not equivalent */
 403
 404                                         //check if they are both blanks
 405                                         if ((seqA[i] == '.') && (seqB[i] == '-')) ok = 1;
 406                                         else if ((seqB[i] == '.') && (seqA[i] == '-')) ok = 1;
 407
 408                                         if (ok == 0) {  numDiff++;  }
 409                                 }
 410                         //}
 411                 }
 412
 413                 return numDiff;
 414
 415         }
 416         catch(exception& e) {
 417                 m->errorOut(e, "Ccode", "getDiff");
 418                 exit(1);
 419         }
 420 }
 421 //***************************************************************************************************************
 422 //tried to make this look most like ccode original implementation
 423 void Ccode::removeBadReferenceSeqs(vector<SeqDist>& seqs) {
 424         try {
 425
 426                 vector< vector<int> > numDiffBases;
 427                 numDiffBases.resize(seqs.size());
 428                 //initialize to 0
 429                 for (int i = 0; i < numDiffBases.size(); i++) { numDiffBases[i].resize(seqs.size(),0); }
 430
 431                 it = trim.begin();
 432                 int length = it->second - it->first;
 433
 434                 //calc differences from each sequence to everyother seq in the set
 435                 for (int i = 0; i < seqs.size(); i++) {
 436
 437                         string seqA = seqs[i].seq->getAligned().substr(it->first, length);
 438
 439                         //so you don't calc i to j and j to i since they are the same
 440                         for (int j = 0; j < i; j++) {
 441
 442                                 string seqB = seqs[j].seq->getAligned().substr(it->first, length);
 443
 444                                 //compare strings
 445                                 int numDiff = getDiff(seqA, seqB);
 446
 447                                 numDiffBases[i][j] = numDiff;
 448                                 numDiffBases[j][i] = numDiff;
 449                         }
 450                 }
 451
 452                 //initailize remove to 0
 453                 vector<int> remove;  remove.resize(seqs.size(), 0);
 454                 float top = ((20*length) / (float) 100);
 455                 float bottom = ((0.5*length) / (float) 100);
 456
 457                 //check each numDiffBases and if any are higher than threshold set remove to 1 so you can remove those seqs from the closest set
 458                 for (int i = 0; i < numDiffBases.size(); i++) {
 459                         for (int j = 0; j < i; j++) {
 460                                 //are you more than 20% different
 461                                 if (numDiffBases[i][j] > top)           {  remove[j] = 1;  }
 462                                 //are you less than 0.5% different
 463                                 if (numDiffBases[i][j] < bottom)        {  remove[j] = 1;  }
 464                         }
 465                 }
 466
 467                 int numSeqsLeft = 0;
 468
 469                 //count seqs that are not going to be removed
 470                 for (int i = 0; i < remove.size(); i++) {
 471                         if (remove[i] == 0)  { numSeqsLeft++;  }
 472                 }
 473
 474                 //if you have enough then remove bad ones
 475                 if (numSeqsLeft >= 3) {
 476                         vector<SeqDist> goodSeqs;
 477                         //remove bad seqs
 478                         for (int i = 0; i < remove.size(); i++) {
 479                                 if (remove[i] == 0) {
 480                                         goodSeqs.push_back(seqs[i]);
 481                                 }
 482                         }
 483
 484                         seqs = goodSeqs;
 485
 486                 }else { //warn, but dont remove any
 487                         m->mothurOut(querySeq->getName() + " does not have an adaquate number of reference sequences that are within 20% and 0.5% similarity.  I will continue, but please check."); m->mothurOutEndLine();
 488                 }
 489
 490         }
 491         catch(exception& e) {
 492                 m->errorOut(e, "Ccode", "removeBadReferenceSeqs");
 493                 exit(1);
 494         }
 495 }
 496 //***************************************************************************************************************
 497 //makes copy of templateseq for filter
 498 vector<SeqDist>  Ccode::findClosest(Sequence* q, int numWanted) {
 499         try{
 500
 501                 vector<SeqDist>  topMatches;
 502
 503                 Sequence query = *(q);
 504
 505                 //calc distance to each sequence in template seqs
 506                 for (int i = 0; i < templateSeqs.size(); i++) {
 507
 508                         Sequence ref = *(templateSeqs[i]);
 509
 510                         //find overall dist
 511                         distCalc->calcDist(query, ref);
 512                         float dist = distCalc->getDist();
 513
 514                         //save distance
 515                         SeqDist temp;
 516                         temp.seq = new Sequence(templateSeqs[i]->getName(), templateSeqs[i]->getAligned());
 517                         temp.dist = dist;
 518
 519                         topMatches.push_back(temp);
 520                 }
 521
 522                 sort(topMatches.begin(), topMatches.end(), compareSeqDist);
 523
 524                 for (int i = numWanted; i < topMatches.size(); i++) {  delete topMatches[i].seq;  }
 525
 526                 topMatches.resize(numWanted);
 527
 528                 return topMatches;
 529
 530         }
 531         catch(exception& e) {
 532                 m->errorOut(e, "Ccode", "findClosestSides");
 533                 exit(1);
 534         }
 535 }
 536 /**************************************************************************************************/
 537 //find the distances from each reference sequence to every other reference sequence for each window for this query
 538 void Ccode::getAverageRef(vector<SeqDist> ref) {
 539         try {
 540
 541                 vector< vector< vector<int> > >  diffs;  //diffs[0][1][2] is the number of differences between ref seq 0 and ref seq 1 at window 2.
 542
 543                 //initialize diffs vector
 544                 diffs.resize(ref.size());
 545                 for (int i = 0; i < diffs.size(); i++) {
 546                         diffs[i].resize(ref.size());
 547                         for (int j = 0; j < diffs[i].size(); j++) {
 548                                 diffs[i][j].resize(windows.size(), 0);
 549                         }
 550                 }
 551
 552                 it = trim.begin();
 553
 554                 //find the distances from each reference sequence to every other reference sequence for each window for this query
 555                 for (int i = 0; i < ref.size(); i++) {
 556
 557                         string refI = ref[i].seq->getAligned();
 558
 559                         //j<i, so you don't find distances from i to j and then j to i.
 560                         for (int j = 0; j < i; j++) {
 561
 562                                 string refJ = ref[j].seq->getAligned();
 563
 564                                 for (int k = 0; k < windows.size(); k++) {
 565
 566                                         string refIWindowk, refJWindowk;
 567
 568                                         if (k < windows.size()-1) {
 569                                                 //get window strings
 570                                                 refIWindowk = refI.substr(windows[k], windowSizes);
 571                                                 refJWindowk = refJ.substr(windows[k], windowSizes);
 572                                         }else { //last window may be smaller than rest - see findwindows
 573                                                 //get window strings
 574                                                 refIWindowk = refI.substr(windows[k], (it->second-windows[k]));
 575                                                 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
 576                                         }
 577
 578                                         //find differences
 579                                         int diff = getDiff(refIWindowk, refJWindowk);
 580
 581                                         //save differences in [i][j][k] and [j][i][k] since they are the same
 582                                         diffs[i][j][k] = diff;
 583                                         diffs[j][i][k] = diff;
 584
 585                                 }//k
 586
 587                         }//j
 588
 589                 }//i
 590
 591                 //initialize sumRef for this query
 592                 sumRef.resize(windows.size(), 0);
 593                 sumSquaredRef.resize(windows.size(), 0);
 594                 averageRef.resize(windows.size(), 0);
 595
 596                 //find the sum of the differences for hte reference sequences
 597                 for (int i = 0; i < diffs.size(); i++) {
 598                         for (int j = 0; j < i; j++) {
 599
 600                                 //increment this querys reference sequences combos
 601                                 refCombo++;
 602
 603                                 for (int k = 0; k < diffs[i][j].size(); k++) {
 604                                         sumRef[k] += diffs[i][j][k];
 605                                         sumSquaredRef[k] += (diffs[i][j][k]*diffs[i][j][k]);
 606                                 }//k
 607
 608                         }//j
 609                 }//i
 610
 611
 612                 //find the average of the differences for the references for each window
 613                 for (int i = 0; i < windows.size(); i++) {
 614                         averageRef[i] = sumRef[i] / (float) refCombo;
 615                 }
 616
 617         }
 618         catch(exception& e) {
 619                 m->errorOut(e, "Ccode", "getAverageRef");
 620                 exit(1);
 621         }
 622 }
 623 /**************************************************************************************************/
 624 void Ccode::getAverageQuery (vector<SeqDist> ref, Sequence* query) {
 625         try {
 626
 627                 vector< vector<int> >  diffs;  //diffs[1][2] is the number of differences between querySeqs[query] and ref seq 1 at window 2.
 628
 629                 //initialize diffs vector
 630                 diffs.resize(ref.size());
 631                 for (int j = 0; j < diffs.size(); j++) {
 632                         diffs[j].resize(windows.size(), 0);
 633                 }
 634
 635                 it = trim.begin();
 636
 637                 string refQuery = query->getAligned();
 638
 639                 //j<i, so you don't find distances from i to j and then j to i.
 640                 for (int j = 0; j < ref.size(); j++) {
 641
 642                          string refJ = ref[j].seq->getAligned();
 643
 644                          for (int k = 0; k < windows.size(); k++) {
 645
 646                                         string QueryWindowk, refJWindowk;
 647
 648                                         if (k < windows.size()-1) {
 649                                                 //get window strings
 650                                                 QueryWindowk = refQuery.substr(windows[k], windowSizes);
 651                                                 refJWindowk = refJ.substr(windows[k], windowSizes);
 652                                         }else { //last window may be smaller than rest - see findwindows
 653                                                 //get window strings
 654                                                 QueryWindowk = refQuery.substr(windows[k], (it->second-windows[k]));
 655                                                 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
 656                                         }
 657
 658                                         //find differences
 659                                         int diff = getDiff(QueryWindowk, refJWindowk);
 660
 661                                         //save differences
 662                                         diffs[j][k] = diff;
 663
 664                          }//k
 665                 }//j
 666
 667
 668                 //initialize sumRef for this query
 669                 sumQuery.resize(windows.size(), 0);
 670                 sumSquaredQuery.resize(windows.size(), 0);
 671                 averageQuery.resize(windows.size(), 0);
 672
 673                 //find the sum of the differences
 674                 for (int j = 0; j < diffs.size(); j++) {
 675                         for (int k = 0; k < diffs[j].size(); k++) {
 676                                 sumQuery[k] += diffs[j][k];
 677                                 sumSquaredQuery[k] += (diffs[j][k]*diffs[j][k]);
 678                         }//k
 679                 }//j
 680
 681
 682                 //find the average of the differences for the references for each window
 683                 for (int i = 0; i < windows.size(); i++) {
 684                         averageQuery[i] = sumQuery[i] / (float) ref.size();
 685                 }
 686         }
 687         catch(exception& e) {
 688                 m->errorOut(e, "Ccode", "getAverageQuery");
 689                 exit(1);
 690         }
 691 }
 692 /**************************************************************************************************/
 693 void Ccode::findVarianceRef() {
 694         try {
 695
 696                 varRef.resize(windows.size(), 0);
 697                 sdRef.resize(windows.size(), 0);
 698
 699                 //for each window
 700                 for (int i = 0; i < windows.size(); i++) {
 701                         varRef[i] = (sumSquaredRef[i] - ((sumRef[i]*sumRef[i])/(float)refCombo)) / (float)(refCombo-1);
 702                         sdRef[i] = sqrt(varRef[i]);
 703
 704                         //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
 705                         if (averageRef[i] < 0.001)                      {       averageRef[i] = 0.001;          }
 706                         if (sumRef[i] < 0.001)                          {       sumRef[i] = 0.001;                      }
 707                         if (varRef[i] < 0.001)                          {       varRef[i] = 0.001;                      }
 708                         if (sumSquaredRef[i] < 0.001)           {       sumSquaredRef[i] = 0.001;       }
 709                         if (sdRef[i] < 0.001)                           {       sdRef[i] = 0.001;                       }
 710
 711                 }
 712         }
 713         catch(exception& e) {
 714                 m->errorOut(e, "Ccode", "findVarianceRef");
 715                 exit(1);
 716         }
 717 }
 718 /**************************************************************************************************/
 719 void Ccode::findVarianceQuery() {
 720         try {
 721                 varQuery.resize(windows.size(), 0);
 722                 sdQuery.resize(windows.size(), 0);
 723
 724                 //for each window
 725                 for (int i = 0; i < windows.size(); i++) {
 726                         varQuery[i] = (sumSquaredQuery[i] - ((sumQuery[i]*sumQuery[i])/(float) closest.size())) / (float) (closest.size()-1);
 727                         sdQuery[i] = sqrt(varQuery[i]);
 728
 729                         //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
 730                         if (averageQuery[i] < 0.001)                    {       averageQuery[i] = 0.001;                }
 731                         if (sumQuery[i] < 0.001)                                {       sumQuery[i] = 0.001;                    }
 732                         if (varQuery[i] < 0.001)                                {       varQuery[i] = 0.001;                    }
 733                         if (sumSquaredQuery[i] < 0.001)         {       sumSquaredQuery[i] = 0.001;     }
 734                         if (sdQuery[i] < 0.001)                         {       sdQuery[i] = 0.001;                     }
 735                 }
 736
 737         }
 738         catch(exception& e) {
 739                 m->errorOut(e, "Ccode", "findVarianceQuery");
 740                 exit(1);
 741         }
 742 }
 743 /**************************************************************************************************/
 744 void Ccode::determineChimeras() {
 745         try {
 746
 747                 isChimericConfidence.resize(windows.size(), false);
 748                 isChimericTStudent.resize(windows.size(), false);
 749                 isChimericANOVA.resize(windows.size(), false);
 750                 anova.resize(windows.size());
 751
 752
 753                 //for each window
 754                 for (int i = 0; i < windows.size(); i++) {
 755
 756                         //get confidence limits
 757                         float t = getT(closest.size()-1);  //how many seqs you are comparing to this querySeq
 758                         float dsUpper = (averageQuery[i] + (t * sdQuery[i])) / averageRef[i];
 759                         float dsLower = (averageQuery[i] - (t * sdQuery[i])) / averageRef[i];
 760
 761                         if ((dsUpper > 1.0) && (dsLower > 1.0) && (averageQuery[i] > averageRef[i])) {  /* range does not include 1 */
 762                                         isChimericConfidence[i] = true;   /* significantly higher at P<0.05 */
 763
 764                         }
 765
 766                         //student t test
 767                         int degreeOfFreedom = refCombo + closest.size() - 2;
 768                         float denomForT = (((refCombo-1) * varQuery[i] + (closest.size() - 1) * varRef[i]) / (float) degreeOfFreedom) * ((refCombo + closest.size()) / (float) (refCombo * closest.size()));    /* denominator, without sqrt(), for ts calculations */
 769
 770                         float ts = fabs((averageQuery[i] - averageRef[i]) / (sqrt(denomForT)));  /* value of ts for t-student test */
 771                         t = getT(degreeOfFreedom);
 772
 773                         if ((ts >= t) && (averageQuery[i] > averageRef[i])) {
 774                                 isChimericTStudent[i] = true;   /* significantly higher at P<0.05 */
 775                         }
 776
 777                         //anova test
 778                         float value1 = sumQuery[i] + sumRef[i];
 779                         float value2 = sumSquaredQuery[i] + sumSquaredRef[i];
 780                         float value3 = ((sumQuery[i]*sumQuery[i]) / (float) (closest.size())) + ((sumRef[i] * sumRef[i]) / (float) refCombo);
 781                         float value4 = (value1 * value1) / ( (float) (closest.size() + refCombo) );
 782                         float value5 = value2 - value4;
 783                         float value6 = value3 - value4;
 784                         float value7 = value5 - value6;
 785                         float value8 = value7 / ((float) degreeOfFreedom);
 786                         float anovaValue = value6 / value8;
 787
 788                         float f = getF(degreeOfFreedom);
 789
 790                          if ((anovaValue >= f) && (averageQuery[i] > averageRef[i]))  {
 791                        isChimericANOVA[i] = true;   /* significant P<0.05 */
 792                 }
 793
 794                         if (isnan(anovaValue) || isinf(anovaValue)) { anovaValue = 0.0; }
 795
 796                         anova[i] = anovaValue;
 797                 }
 798
 799         }
 800         catch(exception& e) {
 801                 m->errorOut(e, "Ccode", "determineChimeras");
 802                 exit(1);
 803         }
 804 }
 805 /**************************************************************************************************/
 806 float Ccode::getT(int numseq) {
 807         try {
 808
 809                 float tvalue = 0;
 810
 811                 /* t-student critical values for different degrees of freedom and alpha 0.1 in one-tail tests (equivalent to 0.05) */
 812                 if (numseq > 120) tvalue = 1.645;
 813                 else if (numseq > 60) tvalue = 1.658;
 814         else if (numseq > 40) tvalue = 1.671;
 815         else if (numseq > 30) tvalue = 1.684;
 816         else if (numseq > 29) tvalue = 1.697;
 817         else if (numseq > 28) tvalue = 1.699;
 818         else if (numseq > 27) tvalue = 1.701;
 819         else if (numseq > 26) tvalue = 1.703;
 820         else if (numseq > 25) tvalue = 1.706;
 821         else if (numseq > 24) tvalue = 1.708;
 822         else if (numseq > 23) tvalue = 1.711;
 823         else if (numseq > 22) tvalue = 1.714;
 824         else if (numseq > 21) tvalue = 1.717;
 825         else if (numseq > 20) tvalue = 1.721;
 826         else if (numseq > 19) tvalue = 1.725;
 827         else if (numseq > 18) tvalue = 1.729;
 828         else if (numseq > 17) tvalue = 1.734;
 829         else if (numseq > 16) tvalue = 1.740;
 830         else if (numseq > 15) tvalue = 1.746;
 831         else if (numseq > 14) tvalue = 1.753;
 832         else if (numseq > 13) tvalue = 1.761;
 833         else if (numseq > 12) tvalue = 1.771;
 834         else if (numseq > 11) tvalue = 1.782;
 835         else if (numseq > 10) tvalue = 1.796;
 836         else if (numseq > 9) tvalue = 1.812;
 837         else if (numseq > 8) tvalue = 1.833;
 838         else if (numseq > 7) tvalue = 1.860;
 839         else if (numseq > 6) tvalue = 1.895;
 840         else if (numseq > 5) tvalue = 1.943;
 841         else if (numseq > 4) tvalue = 2.015;
 842         else if (numseq > 3) tvalue = 2.132;
 843         else if (numseq > 2) tvalue = 2.353;
 844         else if (numseq > 1) tvalue = 2.920;
 845                 else if (numseq <= 1) {
 846                         m->mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); m->mothurOutEndLine();
 847                 }
 848
 849                 return tvalue;
 850         }
 851         catch(exception& e) {
 852                 m->errorOut(e, "Ccode", "getT");
 853                 exit(1);
 854         }
 855 }
 856 /**************************************************************************************************/
 857 float Ccode::getF(int numseq) {
 858         try {
 859
 860                 float fvalue = 0;
 861
 862                  /* F-Snedecor critical values for v1=1 and different degrees of freedom v2 and alpha 0.05 */
 863         if (numseq > 120) fvalue = 3.84;
 864         else if (numseq > 60) fvalue = 3.92;
 865         else if (numseq > 40) fvalue = 4.00;
 866         else if (numseq > 30) fvalue = 4.08;
 867         else if (numseq > 29) fvalue = 4.17;
 868         else if (numseq > 28) fvalue = 4.18;
 869         else if (numseq > 27) fvalue = 4.20;
 870         else if (numseq > 26) fvalue = 4.21;
 871         else if (numseq > 25) fvalue = 4.23;
 872         else if (numseq > 24) fvalue = 4.24;
 873         else if (numseq > 23) fvalue = 4.26;
 874         else if (numseq > 22) fvalue = 4.28;
 875         else if (numseq > 21) fvalue = 4.30;
 876         else if (numseq > 20) fvalue = 4.32;
 877         else if (numseq > 19) fvalue = 4.35;
 878         else if (numseq > 18) fvalue = 4.38;
 879         else if (numseq > 17) fvalue = 4.41;
 880         else if (numseq > 16) fvalue = 4.45;
 881         else if (numseq > 15) fvalue = 4.49;
 882         else if (numseq > 14) fvalue = 4.54;
 883         else if (numseq > 13) fvalue = 4.60;
 884         else if (numseq > 12) fvalue = 4.67;
 885         else if (numseq > 11) fvalue = 4.75;
 886         else if (numseq > 10) fvalue = 4.84;
 887         else if (numseq > 9) fvalue = 4.96;
 888         else if (numseq > 8) fvalue = 5.12;
 889         else if (numseq > 7) fvalue = 5.32;
 890         else if (numseq > 6) fvalue = 5.59;
 891         else if (numseq > 5) fvalue = 5.99;
 892         else if (numseq > 4) fvalue = 6.61;
 893         else if (numseq > 3) fvalue = 7.71;
 894         else if (numseq > 2) fvalue = 10.1;
 895         else if (numseq > 1) fvalue = 18.5;
 896         else if (numseq > 0) fvalue = 161;
 897                 else if (numseq <= 0) {
 898                         m->mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); m->mothurOutEndLine();
 899         }
 900
 901                 return fvalue;
 902         }
 903         catch(exception& e) {
 904                 m->errorOut(e, "Ccode", "getF");
 905                 exit(1);
 906         }
 907 }
 908 //***************************************************************************************************************
 909
 910
 911