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) {
  39         try {
  40
  41                 mothurOutEndLine();
  42
  43                 ofstream out2;
  44                 openOutputFileAppend(mapInfo, out2);
  45
  46                 out2 << querySeq->getName() << endl;
  47                 for (it = spotMap.begin(); it!= spotMap.end(); it++) {
  48                         out2 << it->first << '\t' << it->second << endl;
  49                 }
  50                 out2.close();
  51
  52                 out.setf(ios::fixed, ios::floatfield); out.setf(ios::showpoint);
  53
  54                 out << querySeq->getName() << endl << endl << "Reference sequences used and distance to query:" << endl;
  55
  56                 for (int j = 0; j < closest.size(); j++) {
  57                         out << setprecision(3) << closest[j].seq->getName() << '\t' << closest[j].dist << endl;
  58                 }
  59                 out << endl << endl;
  60
  61                 //for each window
  62                 //window mapping info.
  63                 out << "Mapping information: ";
  64                 //you mask and did not filter
  65                 if ((seqMask != "") && (!filter)) { out << "mask and trim."; }
  66
  67                 //you filtered and did not mask
  68                 if ((seqMask == "") && (filter)) { out << "filter and trim."; }
  69
  70                 //you masked and filtered
  71                 if ((seqMask != "") && (filter)) { out << "mask, filter and trim."; }
  72
  73                 out << endl << "Window\tStartPos\tEndPos" << endl;
  74                 it = trim.begin();
  75
  76                 for (int k = 0; k < windows.size()-1; k++) {
  77                         out << k+1 << '\t' << spotMap[windows[k]-it->first] << '\t' << spotMap[windows[k]-it->first+windowSizes] << endl;
  78                 }
  79
  80                 out << windows.size() << '\t' << spotMap[windows[windows.size()-1]-it->first] << '\t' << spotMap[it->second-it->first-1] << endl;
  81                 out << endl;
  82
  83                 out << "Window\tAvgQ\t(sdQ)\tAvgR\t(sdR)\tRatio\tAnova" << endl;
  84                 for (int k = 0; k < windows.size(); k++) {
  85                         float ds = averageQuery[k] / averageRef[k];
  86                         out << k+1 << '\t' << averageQuery[k] << '\t' << sdQuery[k] << '\t' << averageRef[k] << '\t'<< sdRef[k] << '\t' << ds << '\t' << anova[k] << endl;
  87                 }
  88                 out << endl;
  89
  90                 //varRef
  91                 //varQuery
  92                 /* F test for differences among variances.
  93                 * varQuery is expected to be higher or similar than varRef */
  94                 //float fs = varQuery[query] / varRef[query];   /* F-Snedecor, test for differences of variances */
  95
  96                 bool results = false;
  97
  98                 //confidence limit, t - Student, anova
  99                 out << "Window\tConfidenceLimit\tt-Student\tAnova" << endl;
 100
 101                 for (int k = 0; k < windows.size(); k++) {
 102                         string temp = "";
 103                         if (isChimericConfidence[k]) {  temp += "*\t"; }
 104                         else { temp += "\t"; }
 105
 106                         if (isChimericTStudent[k]) {  temp += "*\t"; }
 107                         else { temp += "\t"; }
 108
 109                         if (isChimericANOVA[k]) {  temp += "*\t"; }
 110                         else { temp += "\t"; }
 111
 112                         out << k+1 << '\t' << temp << endl;
 113
 114                         if (temp == "*\t*\t*\t") {  results = true;  }
 115                 }
 116                 out << endl;
 117
 118                 if (results) {
 119                         mothurOut(querySeq->getName() + " was found have at least one chimeric window."); mothurOutEndLine();
 120                 }
 121
 122         }
 123         catch(exception& e) {
 124                 errorOut(e, "Ccode", "print");
 125                 exit(1);
 126         }
 127 }
 128 //***************************************************************************************************************
 129 int Ccode::getChimeras(Sequence* query) {
 130         try {
 131
 132                 closest.clear();
 133                 refCombo = 0;
 134                 sumRef.clear();
 135                 varRef.clear();
 136                 varQuery.clear();
 137                 sdRef.clear();
 138                 sdQuery.clear();
 139                 sumQuery.clear();
 140                 sumSquaredRef.clear();
 141                 sumSquaredQuery.clear();
 142                 averageRef.clear();
 143                 averageQuery.clear();
 144                 anova.clear();
 145                 isChimericConfidence.clear();
 146                 isChimericTStudent.clear();
 147                 isChimericANOVA.clear();
 148                 trim.clear();
 149                 spotMap.clear();
 150                 windowSizes = window;
 151                 windows.clear();
 152
 153
 154                 querySeq = query;
 155
 156                 //find closest matches to query
 157                 closest = findClosest(query, numWanted);
 158
 159                 //initialize spotMap
 160                 for (int i = 0; i < query->getAligned().length(); i++) {        spotMap[i] = i;         }
 161
 162                 //mask sequences if the user wants to
 163                 if (seqMask != "") {
 164                         decalc->setMask(seqMask);
 165
 166                         decalc->runMask(query);
 167
 168                         //mask closest
 169                         for (int i = 0; i < closest.size(); i++) {      decalc->runMask(closest[i].seq);        }
 170
 171                         spotMap = decalc->getMaskMap();
 172                 }
 173
 174                 if (filter) {
 175                         vector<Sequence*> temp;
 176                         for (int i = 0; i < closest.size(); i++) { temp.push_back(closest[i].seq);  }
 177                         temp.push_back(query);
 178
 179                         createFilter(temp);
 180
 181                         for (int i = 0; i < temp.size(); i++) { runFilter(temp[i]);  }
 182
 183                         //update spotMap
 184                         map<int, int> newMap;
 185                         int spot = 0;
 186
 187                         for (int i = 0; i < filterString.length(); i++) {
 188                                 if (filterString[i] == '1') {
 189                                         //add to newMap
 190                                         newMap[spot] = spotMap[i];
 191                                         spot++;
 192                                 }
 193                         }
 194                         spotMap = newMap;
 195                 }
 196
 197                 //trim sequences - this follows ccodes remove_extra_gaps
 198                 trimSequences(query);
 199
 200
 201                 //windows are equivalent to words - ccode paper recommends windows are between 5% and 20% on alignment length().
 202                 //Our default will be 10% and we will warn if user tries to use a window above or below these recommendations
 203                 windows = findWindows();
 204
 205
 206                 //remove sequences that are more than 20% different and less than 0.5% different - may want to allow user to specify this later
 207                 removeBadReferenceSeqs(closest);
 208
 209
 210                 //find the averages for each querys references
 211                 getAverageRef(closest);  //fills sumRef, averageRef, sumSquaredRef and refCombo.
 212                 getAverageQuery(closest, query);  //fills sumQuery, averageQuery, sumSquaredQuery.
 213
 214
 215                 //find the averages for each querys references
 216                 findVarianceRef();  //fills varRef and sdRef also sets minimum error rate to 0.001 to avoid divide by 0.
 217
 218
 219                 //find the averages for the query
 220                 findVarianceQuery();  //fills varQuery and sdQuery also sets minimum error rate to 0.001 to avoid divide by 0.
 221
 222                 determineChimeras();  //fills anova, isChimericConfidence, isChimericTStudent and isChimericANOVA.
 223
 224                 //free memory
 225                 for (int i = 0; i < closest.size(); i++) {  delete closest[i].seq;  }
 226
 227                 return 0;
 228         }
 229         catch(exception& e) {
 230                 errorOut(e, "Ccode", "getChimeras");
 231                 exit(1);
 232         }
 233 }
 234 /***************************************************************************************************************/
 235 //ccode algo says it does this to "Removes the initial and final gaps to avoid biases due to incomplete sequences."
 236 void Ccode::trimSequences(Sequence* query) {
 237         try {
 238
 239                 int frontPos = 0;  //should contain first position in all seqs that is not a gap character
 240                 int rearPos = query->getAligned().length();
 241
 242                 //********find first position in closest seqs that is a non gap character***********//
 243                 //find first position all query seqs that is a non gap character
 244                 for (int i = 0; i < closest.size(); i++) {
 245
 246                         string aligned = closest[i].seq->getAligned();
 247                         int pos = 0;
 248
 249                         //find first spot in this seq
 250                         for (int j = 0; j < aligned.length(); j++) {
 251                                 if (isalpha(aligned[j])) {
 252                                         pos = j;
 253                                         break;
 254                                 }
 255                         }
 256
 257                         //save this spot if it is the farthest
 258                         if (pos > frontPos) { frontPos = pos; }
 259                 }
 260
 261                 //find first position all querySeq[query] that is a non gap character
 262                 string aligned = query->getAligned();
 263                 int pos = 0;
 264
 265                 //find first spot in this seq
 266                 for (int j = 0; j < aligned.length(); j++) {
 267                         if (isalpha(aligned[j])) {
 268                                 pos = j;
 269                                 break;
 270                         }
 271                 }
 272
 273                 //save this spot if it is the farthest
 274                 if (pos > frontPos) { frontPos = pos; }
 275
 276
 277                 //********find last position in closest seqs that is a non gap character***********//
 278                 for (int i = 0; i < closest.size(); i++) {
 279
 280                         string aligned = closest[i].seq->getAligned();
 281                         int pos = aligned.length();
 282
 283                         //find first spot in this seq
 284                         for (int j = aligned.length()-1; j >= 0; j--) {
 285                                 if (isalpha(aligned[j])) {
 286                                         pos = j;
 287                                         break;
 288                                 }
 289                         }
 290
 291                         //save this spot if it is the farthest
 292                         if (pos < rearPos) { rearPos = pos; }
 293                 }
 294
 295                 //find last position all querySeqs[query] that is a non gap character
 296                 aligned = query->getAligned();
 297                 pos = aligned.length();
 298
 299                 //find first spot in this seq
 300                 for (int j = aligned.length()-1; j >= 0; j--) {
 301                         if (isalpha(aligned[j])) {
 302                                 pos = j;
 303                                 break;
 304                         }
 305                 }
 306
 307                 //save this spot if it is the farthest
 308                 if (pos < rearPos) { rearPos = pos; }
 309
 310
 311                 //check to make sure that is not whole seq
 312                 if ((rearPos - frontPos - 1) <= 0) {  mothurOut("Error, when I trim your sequences, the entire sequence is trimmed."); mothurOutEndLine(); exit(1);  }
 313
 314                 map<int, int> tempTrim;
 315                 tempTrim[frontPos] = rearPos;
 316
 317                 //save trimmed locations
 318                 trim = tempTrim;
 319
 320                 //update spotMask
 321                 map<int, int> newMap;
 322                 int spot = 0;
 323
 324                 for (int i = frontPos; i < rearPos; i++) {
 325                         //add to newMap
 326                         newMap[spot] = spotMap[i];
 327                         spot++;
 328                 }
 329                 spotMap = newMap;
 330         }
 331         catch(exception& e) {
 332                 errorOut(e, "Ccode", "trimSequences");
 333                 exit(1);
 334         }
 335 }
 336 /***************************************************************************************************************/
 337 vector<int> Ccode::findWindows() {
 338         try {
 339
 340                 vector<int> win;
 341                 it = trim.begin();
 342
 343                 int length = it->second - it->first;
 344
 345                 //default is wanted = 10% of total length
 346                 if (windowSizes > length) {
 347                         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.");
 348                         windowSizes = length / 10;
 349                 }else if (windowSizes == 0) { windowSizes = length / 10;  }
 350                 else if (windowSizes > (length * 0.20)) {
 351                         mothurOut("You have selected a window that is larger than 20% of your sequence length.  This is not recommended, but I will continue anyway."); mothurOutEndLine();
 352                 }else if (windowSizes < (length * 0.05)) {
 353                         mothurOut("You have selected a window that is smaller than 5% of your sequence length.  This is not recommended, but I will continue anyway."); mothurOutEndLine();
 354                 }
 355
 356                 //save starting points of each window
 357                 for (int m = it->first;  m < (it->second-windowSizes); m+=windowSizes) {  win.push_back(m);  }
 358
 359                 //save last window
 360                 if (win[win.size()-1] < (it->first+length)) {
 361                         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
 362                 }                                                                                                                                                                                                       //with this you would get 1,25,50,75,100
 363
 364                 return win;
 365         }
 366         catch(exception& e) {
 367                 errorOut(e, "Ccode", "findWindows");
 368                 exit(1);
 369         }
 370 }
 371 //***************************************************************************************************************
 372 int Ccode::getDiff(string seqA, string seqB) {
 373         try {
 374
 375                 int numDiff = 0;
 376
 377                 for (int i = 0; i < seqA.length(); i++) {
 378                         //if you are both not gaps
 379                         //if (isalpha(seqA[i]) && isalpha(seqA[i])) {
 380                                 //are you different
 381                                 if (seqA[i] != seqB[i]) {
 382                                          int ok; /* ok=1 means equivalent base. Checks for degenerate bases */
 383
 384                                         /* the char in base_a and base_b have been checked and they are different */
 385                                         if ((seqA[i] == 'N') && (seqB[i] != '-')) ok = 1;
 386                                         else if ((seqB[i] == 'N') && (seqA[i] != '-')) ok = 1;
 387                                         else if ((seqA[i] == 'Y') && ((seqB[i] == 'C') || (seqB[i] == 'T'))) ok = 1;
 388                                         else if ((seqB[i] == 'Y') && ((seqA[i] == 'C') || (seqA[i] == 'T'))) ok = 1;
 389                                         else if ((seqA[i] == 'R') && ((seqB[i] == 'G') || (seqB[i] == 'A'))) ok = 1;
 390                                         else if ((seqB[i] == 'R') && ((seqA[i] == 'G') || (seqA[i] == 'A'))) ok = 1;
 391                                         else if ((seqA[i] == 'S') && ((seqB[i] == 'C') || (seqB[i] == 'G'))) ok = 1;
 392                                         else if ((seqB[i] == 'S') && ((seqA[i] == 'C') || (seqA[i] == 'G'))) ok = 1;
 393                                         else if ((seqA[i] == 'W') && ((seqB[i] == 'T') || (seqB[i] == 'A'))) ok = 1;
 394                                         else if ((seqB[i] == 'W') && ((seqA[i] == 'T') || (seqA[i] == 'A'))) ok = 1;
 395                                         else if ((seqA[i] == 'M') && ((seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
 396                                         else if ((seqB[i] == 'M') && ((seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
 397                                         else if ((seqA[i] == 'K') && ((seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
 398                                         else if ((seqB[i] == 'K') && ((seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
 399                                         else if ((seqA[i] == 'V') && ((seqB[i] == 'C') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
 400                                         else if ((seqB[i] == 'V') && ((seqA[i] == 'C') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
 401                                         else if ((seqA[i] == 'H') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
 402                                         else if ((seqB[i] == 'H') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
 403                                         else if ((seqA[i] == 'D') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
 404                                         else if ((seqB[i] == 'D') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
 405                                         else if ((seqA[i] == 'B') && ((seqB[i] == 'C') || (seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
 406                                         else if ((seqB[i] == 'B') && ((seqA[i] == 'C') || (seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
 407                                         else ok = 0;  /* the bases are different and not equivalent */
 408
 409                                         //check if they are both blanks
 410                                         if ((seqA[i] == '.') && (seqB[i] == '-')) ok = 1;
 411                                         else if ((seqB[i] == '.') && (seqA[i] == '-')) ok = 1;
 412
 413                                         if (ok == 0) {  numDiff++;  }
 414                                 }
 415                         //}
 416                 }
 417
 418                 return numDiff;
 419
 420         }
 421         catch(exception& e) {
 422                 errorOut(e, "Ccode", "getDiff");
 423                 exit(1);
 424         }
 425 }
 426 //***************************************************************************************************************
 427 //tried to make this look most like ccode original implementation
 428 void Ccode::removeBadReferenceSeqs(vector<SeqDist>& seqs) {
 429         try {
 430
 431                 vector< vector<int> > numDiffBases;
 432                 numDiffBases.resize(seqs.size());
 433                 //initialize to 0
 434                 for (int i = 0; i < numDiffBases.size(); i++) { numDiffBases[i].resize(seqs.size(),0); }
 435
 436                 it = trim.begin();
 437                 int length = it->second - it->first;
 438
 439                 //calc differences from each sequence to everyother seq in the set
 440                 for (int i = 0; i < seqs.size(); i++) {
 441
 442                         string seqA = seqs[i].seq->getAligned().substr(it->first, length);
 443
 444                         //so you don't calc i to j and j to i since they are the same
 445                         for (int j = 0; j < i; j++) {
 446
 447                                 string seqB = seqs[j].seq->getAligned().substr(it->first, length);
 448
 449                                 //compare strings
 450                                 int numDiff = getDiff(seqA, seqB);
 451
 452                                 numDiffBases[i][j] = numDiff;
 453                                 numDiffBases[j][i] = numDiff;
 454                         }
 455                 }
 456
 457                 //initailize remove to 0
 458                 vector<int> remove;  remove.resize(seqs.size(), 0);
 459                 float top = ((20*length) / (float) 100);
 460                 float bottom = ((0.5*length) / (float) 100);
 461
 462                 //check each numDiffBases and if any are higher than threshold set remove to 1 so you can remove those seqs from the closest set
 463                 for (int i = 0; i < numDiffBases.size(); i++) {
 464                         for (int j = 0; j < i; j++) {
 465                                 //are you more than 20% different
 466                                 if (numDiffBases[i][j] > top)           {  remove[j] = 1;  }
 467                                 //are you less than 0.5% different
 468                                 if (numDiffBases[i][j] < bottom)        {  remove[j] = 1;  }
 469                         }
 470                 }
 471
 472                 int numSeqsLeft = 0;
 473
 474                 //count seqs that are not going to be removed
 475                 for (int i = 0; i < remove.size(); i++) {
 476                         if (remove[i] == 0)  { numSeqsLeft++;  }
 477                 }
 478
 479                 //if you have enough then remove bad ones
 480                 if (numSeqsLeft >= 3) {
 481                         vector<SeqDist> goodSeqs;
 482                         //remove bad seqs
 483                         for (int i = 0; i < remove.size(); i++) {
 484                                 if (remove[i] == 0) {
 485                                         goodSeqs.push_back(seqs[i]);
 486                                 }
 487                         }
 488
 489                         seqs = goodSeqs;
 490
 491                 }else { //warn, but dont remove any
 492                         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."); mothurOutEndLine();
 493                 }
 494
 495         }
 496         catch(exception& e) {
 497                 errorOut(e, "Ccode", "removeBadReferenceSeqs");
 498                 exit(1);
 499         }
 500 }
 501 //***************************************************************************************************************
 502 //makes copy of templateseq for filter
 503 vector<SeqDist>  Ccode::findClosest(Sequence* q, int numWanted) {
 504         try{
 505
 506                 vector<SeqDist>  topMatches;
 507
 508                 Sequence query = *(q);
 509
 510                 //calc distance to each sequence in template seqs
 511                 for (int i = 0; i < templateSeqs.size(); i++) {
 512
 513                         Sequence ref = *(templateSeqs[i]);
 514
 515                         //find overall dist
 516                         distCalc->calcDist(query, ref);
 517                         float dist = distCalc->getDist();
 518
 519                         //save distance
 520                         SeqDist temp;
 521                         temp.seq = new Sequence(templateSeqs[i]->getName(), templateSeqs[i]->getAligned());
 522                         temp.dist = dist;
 523
 524                         topMatches.push_back(temp);
 525                 }
 526
 527                 sort(topMatches.begin(), topMatches.end(), compareSeqDist);
 528
 529                 return topMatches;
 530
 531         }
 532         catch(exception& e) {
 533                 errorOut(e, "Ccode", "findClosestSides");
 534                 exit(1);
 535         }
 536 }
 537 /**************************************************************************************************/
 538 //find the distances from each reference sequence to every other reference sequence for each window for this query
 539 void Ccode::getAverageRef(vector<SeqDist> ref) {
 540         try {
 541
 542                 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.
 543
 544                 //initialize diffs vector
 545                 diffs.resize(ref.size());
 546                 for (int i = 0; i < diffs.size(); i++) {
 547                         diffs[i].resize(ref.size());
 548                         for (int j = 0; j < diffs[i].size(); j++) {
 549                                 diffs[i][j].resize(windows.size(), 0);
 550                         }
 551                 }
 552
 553                 it = trim.begin();
 554
 555                 //find the distances from each reference sequence to every other reference sequence for each window for this query
 556                 for (int i = 0; i < ref.size(); i++) {
 557
 558                         string refI = ref[i].seq->getAligned();
 559
 560                         //j<i, so you don't find distances from i to j and then j to i.
 561                         for (int j = 0; j < i; j++) {
 562
 563                                 string refJ = ref[j].seq->getAligned();
 564
 565                                 for (int k = 0; k < windows.size(); k++) {
 566
 567                                         string refIWindowk, refJWindowk;
 568
 569                                         if (k < windows.size()-1) {
 570                                                 //get window strings
 571                                                 refIWindowk = refI.substr(windows[k], windowSizes);
 572                                                 refJWindowk = refJ.substr(windows[k], windowSizes);
 573                                         }else { //last window may be smaller than rest - see findwindows
 574                                                 //get window strings
 575                                                 refIWindowk = refI.substr(windows[k], (it->second-windows[k]));
 576                                                 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
 577                                         }
 578
 579                                         //find differences
 580                                         int diff = getDiff(refIWindowk, refJWindowk);
 581
 582                                         //save differences in [i][j][k] and [j][i][k] since they are the same
 583                                         diffs[i][j][k] = diff;
 584                                         diffs[j][i][k] = diff;
 585
 586                                 }//k
 587
 588                         }//j
 589
 590                 }//i
 591
 592                 //initialize sumRef for this query
 593                 sumRef.resize(windows.size(), 0);
 594                 sumSquaredRef.resize(windows.size(), 0);
 595                 averageRef.resize(windows.size(), 0);
 596
 597                 //find the sum of the differences for hte reference sequences
 598                 for (int i = 0; i < diffs.size(); i++) {
 599                         for (int j = 0; j < i; j++) {
 600
 601                                 //increment this querys reference sequences combos
 602                                 refCombo++;
 603
 604                                 for (int k = 0; k < diffs[i][j].size(); k++) {
 605                                         sumRef[k] += diffs[i][j][k];
 606                                         sumSquaredRef[k] += (diffs[i][j][k]*diffs[i][j][k]);
 607                                 }//k
 608
 609                         }//j
 610                 }//i
 611
 612
 613                 //find the average of the differences for the references for each window
 614                 for (int i = 0; i < windows.size(); i++) {
 615                         averageRef[i] = sumRef[i] / (float) refCombo;
 616                 }
 617
 618         }
 619         catch(exception& e) {
 620                 errorOut(e, "Ccode", "getAverageRef");
 621                 exit(1);
 622         }
 623 }
 624 /**************************************************************************************************/
 625 void Ccode::getAverageQuery (vector<SeqDist> ref, Sequence* query) {
 626         try {
 627
 628                 vector< vector<int> >  diffs;  //diffs[1][2] is the number of differences between querySeqs[query] and ref seq 1 at window 2.
 629
 630                 //initialize diffs vector
 631                 diffs.resize(ref.size());
 632                 for (int j = 0; j < diffs.size(); j++) {
 633                         diffs[j].resize(windows.size(), 0);
 634                 }
 635
 636                 it = trim.begin();
 637
 638                 string refQuery = query->getAligned();
 639
 640                 //j<i, so you don't find distances from i to j and then j to i.
 641                 for (int j = 0; j < ref.size(); j++) {
 642
 643                          string refJ = ref[j].seq->getAligned();
 644
 645                          for (int k = 0; k < windows.size(); k++) {
 646
 647                                         string QueryWindowk, refJWindowk;
 648
 649                                         if (k < windows.size()-1) {
 650                                                 //get window strings
 651                                                 QueryWindowk = refQuery.substr(windows[k], windowSizes);
 652                                                 refJWindowk = refJ.substr(windows[k], windowSizes);
 653                                         }else { //last window may be smaller than rest - see findwindows
 654                                                 //get window strings
 655                                                 QueryWindowk = refQuery.substr(windows[k], (it->second-windows[k]));
 656                                                 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
 657                                         }
 658
 659                                         //find differences
 660                                         int diff = getDiff(QueryWindowk, refJWindowk);
 661
 662                                         //save differences
 663                                         diffs[j][k] = diff;
 664
 665                          }//k
 666                 }//j
 667
 668
 669                 //initialize sumRef for this query
 670                 sumQuery.resize(windows.size(), 0);
 671                 sumSquaredQuery.resize(windows.size(), 0);
 672                 averageQuery.resize(windows.size(), 0);
 673
 674                 //find the sum of the differences
 675                 for (int j = 0; j < diffs.size(); j++) {
 676                         for (int k = 0; k < diffs[j].size(); k++) {
 677                                 sumQuery[k] += diffs[j][k];
 678                                 sumSquaredQuery[k] += (diffs[j][k]*diffs[j][k]);
 679                         }//k
 680                 }//j
 681
 682
 683                 //find the average of the differences for the references for each window
 684                 for (int i = 0; i < windows.size(); i++) {
 685                         averageQuery[i] = sumQuery[i] / (float) ref.size();
 686                 }
 687         }
 688         catch(exception& e) {
 689                 errorOut(e, "Ccode", "getAverageQuery");
 690                 exit(1);
 691         }
 692 }
 693 /**************************************************************************************************/
 694 void Ccode::findVarianceRef() {
 695         try {
 696
 697                 varRef.resize(windows.size(), 0);
 698                 sdRef.resize(windows.size(), 0);
 699
 700                 //for each window
 701                 for (int i = 0; i < windows.size(); i++) {
 702                         varRef[i] = (sumSquaredRef[i] - ((sumRef[i]*sumRef[i])/(float)refCombo)) / (float)(refCombo-1);
 703                         sdRef[i] = sqrt(varRef[i]);
 704
 705                         //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
 706                         if (averageRef[i] < 0.001)                      {       averageRef[i] = 0.001;          }
 707                         if (sumRef[i] < 0.001)                          {       sumRef[i] = 0.001;                      }
 708                         if (varRef[i] < 0.001)                          {       varRef[i] = 0.001;                      }
 709                         if (sumSquaredRef[i] < 0.001)           {       sumSquaredRef[i] = 0.001;       }
 710                         if (sdRef[i] < 0.001)                           {       sdRef[i] = 0.001;                       }
 711
 712                 }
 713         }
 714         catch(exception& e) {
 715                 errorOut(e, "Ccode", "findVarianceRef");
 716                 exit(1);
 717         }
 718 }
 719 /**************************************************************************************************/
 720 void Ccode::findVarianceQuery() {
 721         try {
 722                 varQuery.resize(windows.size(), 0);
 723                 sdQuery.resize(windows.size(), 0);
 724
 725                 //for each window
 726                 for (int i = 0; i < windows.size(); i++) {
 727                         varQuery[i] = (sumSquaredQuery[i] - ((sumQuery[i]*sumQuery[i])/(float) closest.size())) / (float) (closest.size()-1);
 728                         sdQuery[i] = sqrt(varQuery[i]);
 729
 730                         //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
 731                         if (averageQuery[i] < 0.001)                    {       averageQuery[i] = 0.001;                }
 732                         if (sumQuery[i] < 0.001)                                {       sumQuery[i] = 0.001;                    }
 733                         if (varQuery[i] < 0.001)                                {       varQuery[i] = 0.001;                    }
 734                         if (sumSquaredQuery[i] < 0.001)         {       sumSquaredQuery[i] = 0.001;     }
 735                         if (sdQuery[i] < 0.001)                         {       sdQuery[i] = 0.001;                     }
 736                 }
 737
 738         }
 739         catch(exception& e) {
 740                 errorOut(e, "Ccode", "findVarianceQuery");
 741                 exit(1);
 742         }
 743 }
 744 /**************************************************************************************************/
 745 void Ccode::determineChimeras() {
 746         try {
 747
 748                 isChimericConfidence.resize(windows.size(), false);
 749                 isChimericTStudent.resize(windows.size(), false);
 750                 isChimericANOVA.resize(windows.size(), false);
 751                 anova.resize(windows.size());
 752
 753
 754                 //for each window
 755                 for (int i = 0; i < windows.size(); i++) {
 756
 757                         //get confidence limits
 758                         float t = getT(closest.size()-1);  //how many seqs you are comparing to this querySeq
 759                         float dsUpper = (averageQuery[i] + (t * sdQuery[i])) / averageRef[i];
 760                         float dsLower = (averageQuery[i] - (t * sdQuery[i])) / averageRef[i];
 761
 762                         if ((dsUpper > 1.0) && (dsLower > 1.0) && (averageQuery[i] > averageRef[i])) {  /* range does not include 1 */
 763                                         isChimericConfidence[i] = true;   /* significantly higher at P<0.05 */
 764
 765                         }
 766
 767                         //student t test
 768                         int degreeOfFreedom = refCombo + closest.size() - 2;
 769                         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 */
 770
 771                         float ts = fabs((averageQuery[i] - averageRef[i]) / (sqrt(denomForT)));  /* value of ts for t-student test */
 772                         t = getT(degreeOfFreedom);
 773
 774                         if ((ts >= t) && (averageQuery[i] > averageRef[i])) {
 775                                 isChimericTStudent[i] = true;   /* significantly higher at P<0.05 */
 776                         }
 777
 778                         //anova test
 779                         float value1 = sumQuery[i] + sumRef[i];
 780                         float value2 = sumSquaredQuery[i] + sumSquaredRef[i];
 781                         float value3 = ((sumQuery[i]*sumQuery[i]) / (float) (closest.size())) + ((sumRef[i] * sumRef[i]) / (float) refCombo);
 782                         float value4 = (value1 * value1) / ( (float) (closest.size() + refCombo) );
 783                         float value5 = value2 - value4;
 784                         float value6 = value3 - value4;
 785                         float value7 = value5 - value6;
 786                         float value8 = value7 / ((float) degreeOfFreedom);
 787                         float anovaValue = value6 / value8;
 788
 789                         float f = getF(degreeOfFreedom);
 790
 791                          if ((anovaValue >= f) && (averageQuery[i] > averageRef[i]))  {
 792                        isChimericANOVA[i] = true;   /* significant P<0.05 */
 793                 }
 794
 795                         if (isnan(anovaValue) || isinf(anovaValue)) { anovaValue = 0.0; }
 796
 797                         anova[i] = anovaValue;
 798                 }
 799
 800         }
 801         catch(exception& e) {
 802                 errorOut(e, "Ccode", "determineChimeras");
 803                 exit(1);
 804         }
 805 }
 806 /**************************************************************************************************/
 807 float Ccode::getT(int numseq) {
 808         try {
 809
 810                 float tvalue = 0;
 811
 812                 /* t-student critical values for different degrees of freedom and alpha 0.1 in one-tail tests (equivalent to 0.05) */
 813                 if (numseq > 120) tvalue = 1.645;
 814                 else if (numseq > 60) tvalue = 1.658;
 815         else if (numseq > 40) tvalue = 1.671;
 816         else if (numseq > 30) tvalue = 1.684;
 817         else if (numseq > 29) tvalue = 1.697;
 818         else if (numseq > 28) tvalue = 1.699;
 819         else if (numseq > 27) tvalue = 1.701;
 820         else if (numseq > 26) tvalue = 1.703;
 821         else if (numseq > 25) tvalue = 1.706;
 822         else if (numseq > 24) tvalue = 1.708;
 823         else if (numseq > 23) tvalue = 1.711;
 824         else if (numseq > 22) tvalue = 1.714;
 825         else if (numseq > 21) tvalue = 1.717;
 826         else if (numseq > 20) tvalue = 1.721;
 827         else if (numseq > 19) tvalue = 1.725;
 828         else if (numseq > 18) tvalue = 1.729;
 829         else if (numseq > 17) tvalue = 1.734;
 830         else if (numseq > 16) tvalue = 1.740;
 831         else if (numseq > 15) tvalue = 1.746;
 832         else if (numseq > 14) tvalue = 1.753;
 833         else if (numseq > 13) tvalue = 1.761;
 834         else if (numseq > 12) tvalue = 1.771;
 835         else if (numseq > 11) tvalue = 1.782;
 836         else if (numseq > 10) tvalue = 1.796;
 837         else if (numseq > 9) tvalue = 1.812;
 838         else if (numseq > 8) tvalue = 1.833;
 839         else if (numseq > 7) tvalue = 1.860;
 840         else if (numseq > 6) tvalue = 1.895;
 841         else if (numseq > 5) tvalue = 1.943;
 842         else if (numseq > 4) tvalue = 2.015;
 843         else if (numseq > 3) tvalue = 2.132;
 844         else if (numseq > 2) tvalue = 2.353;
 845         else if (numseq > 1) tvalue = 2.920;
 846                 else if (numseq <= 1) {
 847                         mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); mothurOutEndLine();
 848                 }
 849
 850                 return tvalue;
 851         }
 852         catch(exception& e) {
 853                 errorOut(e, "Ccode", "getT");
 854                 exit(1);
 855         }
 856 }
 857 /**************************************************************************************************/
 858 float Ccode::getF(int numseq) {
 859         try {
 860
 861                 float fvalue = 0;
 862
 863                  /* F-Snedecor critical values for v1=1 and different degrees of freedom v2 and alpha 0.05 */
 864         if (numseq > 120) fvalue = 3.84;
 865         else if (numseq > 60) fvalue = 3.92;
 866         else if (numseq > 40) fvalue = 4.00;
 867         else if (numseq > 30) fvalue = 4.08;
 868         else if (numseq > 29) fvalue = 4.17;
 869         else if (numseq > 28) fvalue = 4.18;
 870         else if (numseq > 27) fvalue = 4.20;
 871         else if (numseq > 26) fvalue = 4.21;
 872         else if (numseq > 25) fvalue = 4.23;
 873         else if (numseq > 24) fvalue = 4.24;
 874         else if (numseq > 23) fvalue = 4.26;
 875         else if (numseq > 22) fvalue = 4.28;
 876         else if (numseq > 21) fvalue = 4.30;
 877         else if (numseq > 20) fvalue = 4.32;
 878         else if (numseq > 19) fvalue = 4.35;
 879         else if (numseq > 18) fvalue = 4.38;
 880         else if (numseq > 17) fvalue = 4.41;
 881         else if (numseq > 16) fvalue = 4.45;
 882         else if (numseq > 15) fvalue = 4.49;
 883         else if (numseq > 14) fvalue = 4.54;
 884         else if (numseq > 13) fvalue = 4.60;
 885         else if (numseq > 12) fvalue = 4.67;
 886         else if (numseq > 11) fvalue = 4.75;
 887         else if (numseq > 10) fvalue = 4.84;
 888         else if (numseq > 9) fvalue = 4.96;
 889         else if (numseq > 8) fvalue = 5.12;
 890         else if (numseq > 7) fvalue = 5.32;
 891         else if (numseq > 6) fvalue = 5.59;
 892         else if (numseq > 5) fvalue = 5.99;
 893         else if (numseq > 4) fvalue = 6.61;
 894         else if (numseq > 3) fvalue = 7.71;
 895         else if (numseq > 2) fvalue = 10.1;
 896         else if (numseq > 1) fvalue = 18.5;
 897         else if (numseq > 0) fvalue = 161;
 898                 else if (numseq <= 0) {
 899                         mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); mothurOutEndLine();
 900         }
 901
 902                 return fvalue;
 903         }
 904         catch(exception& e) {
 905                 errorOut(e, "Ccode", "getF");
 906                 exit(1);
 907         }
 908 }
 909 //***************************************************************************************************************
 910
 911
 912