5 * Created by westcott on 8/24/09.
6 * Copyright 2009 Schloss Lab. All rights reserved.
11 #include "ignoregaps.h"
12 #include "eachgapdist.h"
15 //***************************************************************************************************************
16 Ccode::Ccode(string filename, string temp, bool f, string mask, int win, int numW, string o) : Chimera() {
19 templateFileName = temp; templateSeqs = readSeqs(temp);
25 distCalc = new eachGapDist();
26 decalc = new DeCalculator();
28 mapInfo = outputDir + getRootName(getSimpleName(fastafile)) + "mapinfo";
32 //char* inFileName = new char[mapInfo.length()];
33 //memcpy(inFileName, mapInfo.c_str(), mapInfo.length());
35 char inFileName[1024];
36 strcpy(inFileName, mapInfo.c_str());
38 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
40 MPI_File_open(MPI_COMM_WORLD, inFileName, outMode, MPI_INFO_NULL, &outMap); //comm, filename, mode, info, filepointer
45 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
48 string outString = "Place in masked, filtered and trimmed sequence\tPlace in original alignment\n";
51 int length = outString.length();
52 char* buf2 = new char[length];
53 memcpy(buf2, outString.c_str(), length);
55 MPI_File_write_shared(outMap, buf2, length, MPI_CHAR, &status);
61 openOutputFile(mapInfo, out2);
63 out2 << "Place in masked, filtered and trimmed sequence\tPlace in original alignment" << endl;
67 //***************************************************************************************************************
73 MPI_File_close(&outMap);
76 //***************************************************************************************************************
77 int Ccode::print(ostream& out, ostream& outAcc) {
81 openOutputFileAppend(mapInfo, out2);
83 out2 << querySeq->getName() << endl;
84 for (it = spotMap.begin(); it!= spotMap.end(); it++) {
85 out2 << it->first << '\t' << it->second << endl;
88 out << querySeq->getName() << endl << endl << "Reference sequences used and distance to query:" << endl;
90 for (int j = 0; j < closest.size(); j++) {
91 out << closest[j].seq->getName() << '\t' << closest[j].dist << endl;
96 //window mapping info.
97 out << "Mapping information: ";
98 //you mask and did not filter
99 if ((seqMask != "") && (!filter)) { out << "mask and trim."; }
101 //you filtered and did not mask
102 if ((seqMask == "") && (filter)) { out << "filter and trim."; }
104 //you masked and filtered
105 if ((seqMask != "") && (filter)) { out << "mask, filter and trim."; }
107 out << endl << "Window\tStartPos\tEndPos" << endl;
109 for (int k = 0; k < windows.size()-1; k++) {
110 out << k+1 << '\t' << spotMap[windows[k]-it->first] << '\t' << spotMap[windows[k]-it->first+windowSizes] << endl;
113 out << windows.size() << '\t' << spotMap[windows[windows.size()-1]-it->first] << '\t' << spotMap[it->second-it->first-1] << endl;
115 out << "Window\tAvgQ\t(sdQ)\tAvgR\t(sdR)\tRatio\tAnova" << endl;
116 for (int k = 0; k < windows.size(); k++) {
117 float ds = averageQuery[k] / averageRef[k];
118 out << k+1 << '\t' << averageQuery[k] << '\t' << sdQuery[k] << '\t' << averageRef[k] << '\t'<< sdRef[k] << '\t' << ds << '\t' << anova[k] << endl;
124 /* F test for differences among variances.
125 * varQuery is expected to be higher or similar than varRef */
126 //float fs = varQuery[query] / varRef[query]; /* F-Snedecor, test for differences of variances */
128 bool results = false;
130 //confidence limit, t - Student, anova
131 out << "Window\tConfidenceLimit\tt-Student\tAnova" << endl;
133 for (int k = 0; k < windows.size(); k++) {
135 if (isChimericConfidence[k]) { temp += "*\t"; }
136 else { temp += "\t"; }
138 if (isChimericTStudent[k]) { temp += "*\t"; }
139 else { temp += "\t"; }
141 if (isChimericANOVA[k]) { temp += "*\t"; }
142 else { temp += "\t"; }
144 out << k+1 << '\t' << temp << endl;
146 if (temp == "*\t*\t*\t") { results = true; }
151 m->mothurOut(querySeq->getName() + " was found have at least one chimeric window."); m->mothurOutEndLine();
152 outAcc << querySeq->getName() << endl;
156 for (int i = 0; i < closest.size(); i++) { delete closest[i].seq; }
160 catch(exception& e) {
161 m->errorOut(e, "Ccode", "print");
166 //***************************************************************************************************************
167 int Ccode::print(MPI_File& out, MPI_File& outAcc) {
170 string outMapString = "";
172 outMapString += querySeq->getName() + "\n";
173 for (it = spotMap.begin(); it!= spotMap.end(); it++) {
174 outMapString += toString(it->first) + "\t" + toString(it->second) + "\n";
176 printMapping(outMapString);
179 string outString = "";
180 string outAccString = "";
182 outString += querySeq->getName() + "\n\nReference sequences used and distance to query:\n";
184 for (int j = 0; j < closest.size(); j++) {
185 outString += closest[j].seq->getName() + "\t" + toString(closest[j].dist) + "\n";
187 outString += "\n\nMapping information: ";
190 //window mapping info.
191 //you mask and did not filter
192 if ((seqMask != "") && (!filter)) { outString += "mask and trim."; }
194 //you filtered and did not mask
195 if ((seqMask == "") && (filter)) { outString += "filter and trim."; }
197 //you masked and filtered
198 if ((seqMask != "") && (filter)) { outString += "mask, filter and trim."; }
200 outString += "\nWindow\tStartPos\tEndPos\n";
202 for (int k = 0; k < windows.size()-1; k++) {
203 outString += toString(k+1) + "\t" + toString(spotMap[windows[k]-it->first]) + "\t" + toString(spotMap[windows[k]-it->first+windowSizes]) + "\n";
206 outString += toString(windows.size()) + "\t" + toString(spotMap[windows[windows.size()-1]-it->first]) + "\t" + toString(spotMap[it->second-it->first-1]) + "\n\n";
208 outString += "Window\tAvgQ\t(sdQ)\tAvgR\t(sdR)\tRatio\tAnova\n";
209 for (int k = 0; k < windows.size(); k++) {
210 float ds = averageQuery[k] / averageRef[k];
211 outString += toString(k+1) + "\t" + toString(averageQuery[k]) + "\t" + toString(sdQuery[k]) + "\t" + toString(averageRef[k]) + "\t" + toString(sdRef[k]) + "\t" + toString(ds) + "\t" + toString(anova[k]) + "\n";
216 /* F test for differences among variances.
217 * varQuery is expected to be higher or similar than varRef */
218 //float fs = varQuery[query] / varRef[query]; /* F-Snedecor, test for differences of variances */
220 bool results = false;
222 //confidence limit, t - Student, anova
223 outString += "\nWindow\tConfidenceLimit\tt-Student\tAnova\n";
225 for (int k = 0; k < windows.size(); k++) {
227 if (isChimericConfidence[k]) { temp += "*\t"; }
228 else { temp += "\t"; }
230 if (isChimericTStudent[k]) { temp += "*\t"; }
231 else { temp += "\t"; }
233 if (isChimericANOVA[k]) { temp += "*\t"; }
234 else { temp += "\t"; }
236 outString += toString(k+1) + "\t" + temp + "\n";
238 if (temp == "*\t*\t*\t") { results = true; }
243 int length = outString.length();
244 char* buf2 = new char[length];
245 memcpy(buf2, outString.c_str(), length);
247 MPI_File_write_shared(out, buf2, length, MPI_CHAR, &status);
251 m->mothurOut(querySeq->getName() + " was found have at least one chimeric window."); m->mothurOutEndLine();
252 outAccString += querySeq->getName() + "\n";
254 MPI_Status statusAcc;
255 length = outAccString.length();
256 char* buf = new char[length];
257 memcpy(buf, outAccString.c_str(), length);
259 MPI_File_write_shared(outAcc, buf, length, MPI_CHAR, &statusAcc);
264 for (int i = 0; i < closest.size(); i++) { delete closest[i].seq; }
268 catch(exception& e) {
269 m->errorOut(e, "Ccode", "print");
273 //***************************************************************************************************************
274 int Ccode::printMapping(string& output) {
277 int length = output.length();
278 char* buf = new char[length];
279 memcpy(buf, output.c_str(), length);
281 MPI_File_write_shared(outMap, buf, length, MPI_CHAR, &status);
285 catch(exception& e) {
286 m->errorOut(e, "Ccode", "printMapping");
291 //***************************************************************************************************************
292 int Ccode::getChimeras(Sequence* query) {
303 sumSquaredRef.clear();
304 sumSquaredQuery.clear();
306 averageQuery.clear();
308 isChimericConfidence.clear();
309 isChimericTStudent.clear();
310 isChimericANOVA.clear();
313 windowSizes = window;
319 //find closest matches to query
320 closest = findClosest(query, numWanted);
322 if (m->control_pressed) { return 0; }
325 for (int i = 0; i < query->getAligned().length(); i++) { spotMap[i] = i; }
327 //mask sequences if the user wants to
329 decalc->setMask(seqMask);
331 decalc->runMask(query);
334 for (int i = 0; i < closest.size(); i++) { decalc->runMask(closest[i].seq); }
336 spotMap = decalc->getMaskMap();
340 vector<Sequence*> temp;
341 for (int i = 0; i < closest.size(); i++) { temp.push_back(closest[i].seq); }
342 temp.push_back(query);
344 createFilter(temp, 0.5);
346 for (int i = 0; i < temp.size(); i++) {
347 if (m->control_pressed) { return 0; }
352 map<int, int> newMap;
355 for (int i = 0; i < filterString.length(); i++) {
356 if (filterString[i] == '1') {
358 newMap[spot] = spotMap[i];
365 //trim sequences - this follows ccodes remove_extra_gaps
366 trimSequences(query);
367 if (m->control_pressed) { return 0; }
369 //windows are equivalent to words - ccode paper recommends windows are between 5% and 20% on alignment length().
370 //Our default will be 10% and we will warn if user tries to use a window above or below these recommendations
371 windows = findWindows();
372 if (m->control_pressed) { return 0; }
374 //remove sequences that are more than 20% different and less than 0.5% different - may want to allow user to specify this later
375 removeBadReferenceSeqs(closest);
376 if (m->control_pressed) { return 0; }
378 //find the averages for each querys references
379 getAverageRef(closest); //fills sumRef, averageRef, sumSquaredRef and refCombo.
380 getAverageQuery(closest, query); //fills sumQuery, averageQuery, sumSquaredQuery.
381 if (m->control_pressed) { return 0; }
383 //find the averages for each querys references
384 findVarianceRef(); //fills varRef and sdRef also sets minimum error rate to 0.001 to avoid divide by 0.
385 if (m->control_pressed) { return 0; }
387 //find the averages for the query
388 findVarianceQuery(); //fills varQuery and sdQuery also sets minimum error rate to 0.001 to avoid divide by 0.
389 if (m->control_pressed) { return 0; }
391 determineChimeras(); //fills anova, isChimericConfidence, isChimericTStudent and isChimericANOVA.
392 if (m->control_pressed) { return 0; }
396 catch(exception& e) {
397 m->errorOut(e, "Ccode", "getChimeras");
401 /***************************************************************************************************************/
402 //ccode algo says it does this to "Removes the initial and final gaps to avoid biases due to incomplete sequences."
403 void Ccode::trimSequences(Sequence* query) {
406 int frontPos = 0; //should contain first position in all seqs that is not a gap character
407 int rearPos = query->getAligned().length();
409 //********find first position in closest seqs that is a non gap character***********//
410 //find first position all query seqs that is a non gap character
411 for (int i = 0; i < closest.size(); i++) {
413 string aligned = closest[i].seq->getAligned();
416 //find first spot in this seq
417 for (int j = 0; j < aligned.length(); j++) {
418 if (isalpha(aligned[j])) {
424 //save this spot if it is the farthest
425 if (pos > frontPos) { frontPos = pos; }
428 //find first position all querySeq[query] that is a non gap character
429 string aligned = query->getAligned();
432 //find first spot in this seq
433 for (int j = 0; j < aligned.length(); j++) {
434 if (isalpha(aligned[j])) {
440 //save this spot if it is the farthest
441 if (pos > frontPos) { frontPos = pos; }
444 //********find last position in closest seqs that is a non gap character***********//
445 for (int i = 0; i < closest.size(); i++) {
447 string aligned = closest[i].seq->getAligned();
448 int pos = aligned.length();
450 //find first spot in this seq
451 for (int j = aligned.length()-1; j >= 0; j--) {
452 if (isalpha(aligned[j])) {
458 //save this spot if it is the farthest
459 if (pos < rearPos) { rearPos = pos; }
462 //find last position all querySeqs[query] that is a non gap character
463 aligned = query->getAligned();
464 pos = aligned.length();
466 //find first spot in this seq
467 for (int j = aligned.length()-1; j >= 0; j--) {
468 if (isalpha(aligned[j])) {
474 //save this spot if it is the farthest
475 if (pos < rearPos) { rearPos = pos; }
478 //check to make sure that is not whole seq
479 if ((rearPos - frontPos - 1) <= 0) { m->mothurOut("Error, when I trim your sequences, the entire sequence is trimmed."); m->mothurOutEndLine(); exit(1); }
481 map<int, int> tempTrim;
482 tempTrim[frontPos] = rearPos;
484 //save trimmed locations
488 map<int, int> newMap;
491 for (int i = frontPos; i < rearPos; i++) {
493 newMap[spot] = spotMap[i];
498 catch(exception& e) {
499 m->errorOut(e, "Ccode", "trimSequences");
503 /***************************************************************************************************************/
504 vector<int> Ccode::findWindows() {
510 int length = it->second - it->first;
512 //default is wanted = 10% of total length
513 if (windowSizes > length) {
514 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.");
515 windowSizes = length / 10;
516 }else if (windowSizes == 0) { windowSizes = length / 10; }
517 else if (windowSizes > (length * 0.20)) {
518 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();
519 }else if (windowSizes < (length * 0.05)) {
520 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();
523 //save starting points of each window
524 for (int m = it->first; m < (it->second-windowSizes); m+=windowSizes) { win.push_back(m); }
527 if (win[win.size()-1] < (it->first+length)) {
528 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
529 } //with this you would get 1,25,50,75,100
533 catch(exception& e) {
534 m->errorOut(e, "Ccode", "findWindows");
538 //***************************************************************************************************************
539 int Ccode::getDiff(string seqA, string seqB) {
544 for (int i = 0; i < seqA.length(); i++) {
545 //if you are both not gaps
546 //if (isalpha(seqA[i]) && isalpha(seqA[i])) {
548 if (seqA[i] != seqB[i]) {
549 int ok; /* ok=1 means equivalent base. Checks for degenerate bases */
551 /* the char in base_a and base_b have been checked and they are different */
552 if ((seqA[i] == 'N') && (seqB[i] != '-')) ok = 1;
553 else if ((seqB[i] == 'N') && (seqA[i] != '-')) ok = 1;
554 else if ((seqA[i] == 'Y') && ((seqB[i] == 'C') || (seqB[i] == 'T'))) ok = 1;
555 else if ((seqB[i] == 'Y') && ((seqA[i] == 'C') || (seqA[i] == 'T'))) ok = 1;
556 else if ((seqA[i] == 'R') && ((seqB[i] == 'G') || (seqB[i] == 'A'))) ok = 1;
557 else if ((seqB[i] == 'R') && ((seqA[i] == 'G') || (seqA[i] == 'A'))) ok = 1;
558 else if ((seqA[i] == 'S') && ((seqB[i] == 'C') || (seqB[i] == 'G'))) ok = 1;
559 else if ((seqB[i] == 'S') && ((seqA[i] == 'C') || (seqA[i] == 'G'))) ok = 1;
560 else if ((seqA[i] == 'W') && ((seqB[i] == 'T') || (seqB[i] == 'A'))) ok = 1;
561 else if ((seqB[i] == 'W') && ((seqA[i] == 'T') || (seqA[i] == 'A'))) ok = 1;
562 else if ((seqA[i] == 'M') && ((seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
563 else if ((seqB[i] == 'M') && ((seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
564 else if ((seqA[i] == 'K') && ((seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
565 else if ((seqB[i] == 'K') && ((seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
566 else if ((seqA[i] == 'V') && ((seqB[i] == 'C') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
567 else if ((seqB[i] == 'V') && ((seqA[i] == 'C') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
568 else if ((seqA[i] == 'H') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'C'))) ok = 1;
569 else if ((seqB[i] == 'H') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'C'))) ok = 1;
570 else if ((seqA[i] == 'D') && ((seqB[i] == 'T') || (seqB[i] == 'A') || (seqB[i] == 'G'))) ok = 1;
571 else if ((seqB[i] == 'D') && ((seqA[i] == 'T') || (seqA[i] == 'A') || (seqA[i] == 'G'))) ok = 1;
572 else if ((seqA[i] == 'B') && ((seqB[i] == 'C') || (seqB[i] == 'T') || (seqB[i] == 'G'))) ok = 1;
573 else if ((seqB[i] == 'B') && ((seqA[i] == 'C') || (seqA[i] == 'T') || (seqA[i] == 'G'))) ok = 1;
574 else ok = 0; /* the bases are different and not equivalent */
576 //check if they are both blanks
577 if ((seqA[i] == '.') && (seqB[i] == '-')) ok = 1;
578 else if ((seqB[i] == '.') && (seqA[i] == '-')) ok = 1;
580 if (ok == 0) { numDiff++; }
588 catch(exception& e) {
589 m->errorOut(e, "Ccode", "getDiff");
593 //***************************************************************************************************************
594 //tried to make this look most like ccode original implementation
595 void Ccode::removeBadReferenceSeqs(vector<SeqDist>& seqs) {
598 vector< vector<int> > numDiffBases;
599 numDiffBases.resize(seqs.size());
601 for (int i = 0; i < numDiffBases.size(); i++) { numDiffBases[i].resize(seqs.size(),0); }
604 int length = it->second - it->first;
606 //calc differences from each sequence to everyother seq in the set
607 for (int i = 0; i < seqs.size(); i++) {
609 string seqA = seqs[i].seq->getAligned().substr(it->first, length);
611 //so you don't calc i to j and j to i since they are the same
612 for (int j = 0; j < i; j++) {
614 string seqB = seqs[j].seq->getAligned().substr(it->first, length);
617 int numDiff = getDiff(seqA, seqB);
619 numDiffBases[i][j] = numDiff;
620 numDiffBases[j][i] = numDiff;
624 //initailize remove to 0
625 vector<int> remove; remove.resize(seqs.size(), 0);
626 float top = ((20*length) / (float) 100);
627 float bottom = ((0.5*length) / (float) 100);
629 //check each numDiffBases and if any are higher than threshold set remove to 1 so you can remove those seqs from the closest set
630 for (int i = 0; i < numDiffBases.size(); i++) {
631 for (int j = 0; j < i; j++) {
632 //are you more than 20% different
633 if (numDiffBases[i][j] > top) { remove[j] = 1; }
634 //are you less than 0.5% different
635 if (numDiffBases[i][j] < bottom) { remove[j] = 1; }
641 //count seqs that are not going to be removed
642 for (int i = 0; i < remove.size(); i++) {
643 if (remove[i] == 0) { numSeqsLeft++; }
646 //if you have enough then remove bad ones
647 if (numSeqsLeft >= 3) {
648 vector<SeqDist> goodSeqs;
650 for (int i = 0; i < remove.size(); i++) {
651 if (remove[i] == 0) {
652 goodSeqs.push_back(seqs[i]);
658 }else { //warn, but dont remove any
659 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();
663 catch(exception& e) {
664 m->errorOut(e, "Ccode", "removeBadReferenceSeqs");
668 //***************************************************************************************************************
669 //makes copy of templateseq for filter
670 vector<SeqDist> Ccode::findClosest(Sequence* q, int numWanted) {
673 vector<SeqDist> topMatches;
675 Sequence query = *(q);
677 //calc distance to each sequence in template seqs
678 for (int i = 0; i < templateSeqs.size(); i++) {
680 Sequence ref = *(templateSeqs[i]);
683 distCalc->calcDist(query, ref);
684 float dist = distCalc->getDist();
688 temp.seq = new Sequence(templateSeqs[i]->getName(), templateSeqs[i]->getAligned());
691 topMatches.push_back(temp);
694 sort(topMatches.begin(), topMatches.end(), compareSeqDist);
696 for (int i = numWanted; i < topMatches.size(); i++) { delete topMatches[i].seq; }
698 topMatches.resize(numWanted);
703 catch(exception& e) {
704 m->errorOut(e, "Ccode", "findClosestSides");
708 /**************************************************************************************************/
709 //find the distances from each reference sequence to every other reference sequence for each window for this query
710 void Ccode::getAverageRef(vector<SeqDist> ref) {
713 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.
715 //initialize diffs vector
716 diffs.resize(ref.size());
717 for (int i = 0; i < diffs.size(); i++) {
718 diffs[i].resize(ref.size());
719 for (int j = 0; j < diffs[i].size(); j++) {
720 diffs[i][j].resize(windows.size(), 0);
726 //find the distances from each reference sequence to every other reference sequence for each window for this query
727 for (int i = 0; i < ref.size(); i++) {
729 string refI = ref[i].seq->getAligned();
731 //j<i, so you don't find distances from i to j and then j to i.
732 for (int j = 0; j < i; j++) {
734 string refJ = ref[j].seq->getAligned();
736 for (int k = 0; k < windows.size(); k++) {
738 string refIWindowk, refJWindowk;
740 if (k < windows.size()-1) {
742 refIWindowk = refI.substr(windows[k], windowSizes);
743 refJWindowk = refJ.substr(windows[k], windowSizes);
744 }else { //last window may be smaller than rest - see findwindows
746 refIWindowk = refI.substr(windows[k], (it->second-windows[k]));
747 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
751 int diff = getDiff(refIWindowk, refJWindowk);
753 //save differences in [i][j][k] and [j][i][k] since they are the same
754 diffs[i][j][k] = diff;
755 diffs[j][i][k] = diff;
763 //initialize sumRef for this query
764 sumRef.resize(windows.size(), 0);
765 sumSquaredRef.resize(windows.size(), 0);
766 averageRef.resize(windows.size(), 0);
768 //find the sum of the differences for hte reference sequences
769 for (int i = 0; i < diffs.size(); i++) {
770 for (int j = 0; j < i; j++) {
772 //increment this querys reference sequences combos
775 for (int k = 0; k < diffs[i][j].size(); k++) {
776 sumRef[k] += diffs[i][j][k];
777 sumSquaredRef[k] += (diffs[i][j][k]*diffs[i][j][k]);
784 //find the average of the differences for the references for each window
785 for (int i = 0; i < windows.size(); i++) {
786 averageRef[i] = sumRef[i] / (float) refCombo;
790 catch(exception& e) {
791 m->errorOut(e, "Ccode", "getAverageRef");
795 /**************************************************************************************************/
796 void Ccode::getAverageQuery (vector<SeqDist> ref, Sequence* query) {
799 vector< vector<int> > diffs; //diffs[1][2] is the number of differences between querySeqs[query] and ref seq 1 at window 2.
801 //initialize diffs vector
802 diffs.resize(ref.size());
803 for (int j = 0; j < diffs.size(); j++) {
804 diffs[j].resize(windows.size(), 0);
809 string refQuery = query->getAligned();
811 //j<i, so you don't find distances from i to j and then j to i.
812 for (int j = 0; j < ref.size(); j++) {
814 string refJ = ref[j].seq->getAligned();
816 for (int k = 0; k < windows.size(); k++) {
818 string QueryWindowk, refJWindowk;
820 if (k < windows.size()-1) {
822 QueryWindowk = refQuery.substr(windows[k], windowSizes);
823 refJWindowk = refJ.substr(windows[k], windowSizes);
824 }else { //last window may be smaller than rest - see findwindows
826 QueryWindowk = refQuery.substr(windows[k], (it->second-windows[k]));
827 refJWindowk = refJ.substr(windows[k], (it->second-windows[k]));
831 int diff = getDiff(QueryWindowk, refJWindowk);
840 //initialize sumRef for this query
841 sumQuery.resize(windows.size(), 0);
842 sumSquaredQuery.resize(windows.size(), 0);
843 averageQuery.resize(windows.size(), 0);
845 //find the sum of the differences
846 for (int j = 0; j < diffs.size(); j++) {
847 for (int k = 0; k < diffs[j].size(); k++) {
848 sumQuery[k] += diffs[j][k];
849 sumSquaredQuery[k] += (diffs[j][k]*diffs[j][k]);
854 //find the average of the differences for the references for each window
855 for (int i = 0; i < windows.size(); i++) {
856 averageQuery[i] = sumQuery[i] / (float) ref.size();
859 catch(exception& e) {
860 m->errorOut(e, "Ccode", "getAverageQuery");
864 /**************************************************************************************************/
865 void Ccode::findVarianceRef() {
868 varRef.resize(windows.size(), 0);
869 sdRef.resize(windows.size(), 0);
872 for (int i = 0; i < windows.size(); i++) {
873 varRef[i] = (sumSquaredRef[i] - ((sumRef[i]*sumRef[i])/(float)refCombo)) / (float)(refCombo-1);
874 sdRef[i] = sqrt(varRef[i]);
876 //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
877 if (averageRef[i] < 0.001) { averageRef[i] = 0.001; }
878 if (sumRef[i] < 0.001) { sumRef[i] = 0.001; }
879 if (varRef[i] < 0.001) { varRef[i] = 0.001; }
880 if (sumSquaredRef[i] < 0.001) { sumSquaredRef[i] = 0.001; }
881 if (sdRef[i] < 0.001) { sdRef[i] = 0.001; }
885 catch(exception& e) {
886 m->errorOut(e, "Ccode", "findVarianceRef");
890 /**************************************************************************************************/
891 void Ccode::findVarianceQuery() {
893 varQuery.resize(windows.size(), 0);
894 sdQuery.resize(windows.size(), 0);
897 for (int i = 0; i < windows.size(); i++) {
898 varQuery[i] = (sumSquaredQuery[i] - ((sumQuery[i]*sumQuery[i])/(float) closest.size())) / (float) (closest.size()-1);
899 sdQuery[i] = sqrt(varQuery[i]);
901 //set minimum error rate to 0.001 - to avoid potential divide by zero - not sure if this is necessary but it follows ccode implementation
902 if (averageQuery[i] < 0.001) { averageQuery[i] = 0.001; }
903 if (sumQuery[i] < 0.001) { sumQuery[i] = 0.001; }
904 if (varQuery[i] < 0.001) { varQuery[i] = 0.001; }
905 if (sumSquaredQuery[i] < 0.001) { sumSquaredQuery[i] = 0.001; }
906 if (sdQuery[i] < 0.001) { sdQuery[i] = 0.001; }
910 catch(exception& e) {
911 m->errorOut(e, "Ccode", "findVarianceQuery");
915 /**************************************************************************************************/
916 void Ccode::determineChimeras() {
919 isChimericConfidence.resize(windows.size(), false);
920 isChimericTStudent.resize(windows.size(), false);
921 isChimericANOVA.resize(windows.size(), false);
922 anova.resize(windows.size());
926 for (int i = 0; i < windows.size(); i++) {
928 //get confidence limits
929 float t = getT(closest.size()-1); //how many seqs you are comparing to this querySeq
930 float dsUpper = (averageQuery[i] + (t * sdQuery[i])) / averageRef[i];
931 float dsLower = (averageQuery[i] - (t * sdQuery[i])) / averageRef[i];
933 if ((dsUpper > 1.0) && (dsLower > 1.0) && (averageQuery[i] > averageRef[i])) { /* range does not include 1 */
934 isChimericConfidence[i] = true; /* significantly higher at P<0.05 */
939 int degreeOfFreedom = refCombo + closest.size() - 2;
940 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 */
942 float ts = fabs((averageQuery[i] - averageRef[i]) / (sqrt(denomForT))); /* value of ts for t-student test */
943 t = getT(degreeOfFreedom);
945 if ((ts >= t) && (averageQuery[i] > averageRef[i])) {
946 isChimericTStudent[i] = true; /* significantly higher at P<0.05 */
950 float value1 = sumQuery[i] + sumRef[i];
951 float value2 = sumSquaredQuery[i] + sumSquaredRef[i];
952 float value3 = ((sumQuery[i]*sumQuery[i]) / (float) (closest.size())) + ((sumRef[i] * sumRef[i]) / (float) refCombo);
953 float value4 = (value1 * value1) / ( (float) (closest.size() + refCombo) );
954 float value5 = value2 - value4;
955 float value6 = value3 - value4;
956 float value7 = value5 - value6;
957 float value8 = value7 / ((float) degreeOfFreedom);
958 float anovaValue = value6 / value8;
960 float f = getF(degreeOfFreedom);
962 if ((anovaValue >= f) && (averageQuery[i] > averageRef[i])) {
963 isChimericANOVA[i] = true; /* significant P<0.05 */
966 if (isnan(anovaValue) || isinf(anovaValue)) { anovaValue = 0.0; }
968 anova[i] = anovaValue;
972 catch(exception& e) {
973 m->errorOut(e, "Ccode", "determineChimeras");
977 /**************************************************************************************************/
978 float Ccode::getT(int numseq) {
983 /* t-student critical values for different degrees of freedom and alpha 0.1 in one-tail tests (equivalent to 0.05) */
984 if (numseq > 120) tvalue = 1.645;
985 else if (numseq > 60) tvalue = 1.658;
986 else if (numseq > 40) tvalue = 1.671;
987 else if (numseq > 30) tvalue = 1.684;
988 else if (numseq > 29) tvalue = 1.697;
989 else if (numseq > 28) tvalue = 1.699;
990 else if (numseq > 27) tvalue = 1.701;
991 else if (numseq > 26) tvalue = 1.703;
992 else if (numseq > 25) tvalue = 1.706;
993 else if (numseq > 24) tvalue = 1.708;
994 else if (numseq > 23) tvalue = 1.711;
995 else if (numseq > 22) tvalue = 1.714;
996 else if (numseq > 21) tvalue = 1.717;
997 else if (numseq > 20) tvalue = 1.721;
998 else if (numseq > 19) tvalue = 1.725;
999 else if (numseq > 18) tvalue = 1.729;
1000 else if (numseq > 17) tvalue = 1.734;
1001 else if (numseq > 16) tvalue = 1.740;
1002 else if (numseq > 15) tvalue = 1.746;
1003 else if (numseq > 14) tvalue = 1.753;
1004 else if (numseq > 13) tvalue = 1.761;
1005 else if (numseq > 12) tvalue = 1.771;
1006 else if (numseq > 11) tvalue = 1.782;
1007 else if (numseq > 10) tvalue = 1.796;
1008 else if (numseq > 9) tvalue = 1.812;
1009 else if (numseq > 8) tvalue = 1.833;
1010 else if (numseq > 7) tvalue = 1.860;
1011 else if (numseq > 6) tvalue = 1.895;
1012 else if (numseq > 5) tvalue = 1.943;
1013 else if (numseq > 4) tvalue = 2.015;
1014 else if (numseq > 3) tvalue = 2.132;
1015 else if (numseq > 2) tvalue = 2.353;
1016 else if (numseq > 1) tvalue = 2.920;
1017 else if (numseq <= 1) {
1018 m->mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); m->mothurOutEndLine();
1023 catch(exception& e) {
1024 m->errorOut(e, "Ccode", "getT");
1028 /**************************************************************************************************/
1029 float Ccode::getF(int numseq) {
1034 /* F-Snedecor critical values for v1=1 and different degrees of freedom v2 and alpha 0.05 */
1035 if (numseq > 120) fvalue = 3.84;
1036 else if (numseq > 60) fvalue = 3.92;
1037 else if (numseq > 40) fvalue = 4.00;
1038 else if (numseq > 30) fvalue = 4.08;
1039 else if (numseq > 29) fvalue = 4.17;
1040 else if (numseq > 28) fvalue = 4.18;
1041 else if (numseq > 27) fvalue = 4.20;
1042 else if (numseq > 26) fvalue = 4.21;
1043 else if (numseq > 25) fvalue = 4.23;
1044 else if (numseq > 24) fvalue = 4.24;
1045 else if (numseq > 23) fvalue = 4.26;
1046 else if (numseq > 22) fvalue = 4.28;
1047 else if (numseq > 21) fvalue = 4.30;
1048 else if (numseq > 20) fvalue = 4.32;
1049 else if (numseq > 19) fvalue = 4.35;
1050 else if (numseq > 18) fvalue = 4.38;
1051 else if (numseq > 17) fvalue = 4.41;
1052 else if (numseq > 16) fvalue = 4.45;
1053 else if (numseq > 15) fvalue = 4.49;
1054 else if (numseq > 14) fvalue = 4.54;
1055 else if (numseq > 13) fvalue = 4.60;
1056 else if (numseq > 12) fvalue = 4.67;
1057 else if (numseq > 11) fvalue = 4.75;
1058 else if (numseq > 10) fvalue = 4.84;
1059 else if (numseq > 9) fvalue = 4.96;
1060 else if (numseq > 8) fvalue = 5.12;
1061 else if (numseq > 7) fvalue = 5.32;
1062 else if (numseq > 6) fvalue = 5.59;
1063 else if (numseq > 5) fvalue = 5.99;
1064 else if (numseq > 4) fvalue = 6.61;
1065 else if (numseq > 3) fvalue = 7.71;
1066 else if (numseq > 2) fvalue = 10.1;
1067 else if (numseq > 1) fvalue = 18.5;
1068 else if (numseq > 0) fvalue = 161;
1069 else if (numseq <= 0) {
1070 m->mothurOut("Two or more reference sequences are required, your data will be flawed.\n"); m->mothurOutEndLine();
1075 catch(exception& e) {
1076 m->errorOut(e, "Ccode", "getF");
1080 //***************************************************************************************************************