X-Git-Url: https://git.donarmstrong.com/?p=mothur.git;a=blobdiff_plain;f=slayer.cpp;h=a787f2727dd8290b89bffeb382bae6c15001110d;hp=d838a2e86857bbce9a3375b795abe17964cbeaf0;hb=615301e57c25e241356a9c2380648d117709458d;hpb=05a0195e07c42c996592ee1c8abb63adedb7f493 diff --git a/slayer.cpp b/slayer.cpp index d838a2e..a787f27 100644 --- a/slayer.cpp +++ b/slayer.cpp @@ -10,123 +10,577 @@ #include "slayer.h" /***********************************************************************/ -Slayer::Slayer(int win, int increment, int parentThreshold, float div) : - windowSize(win), windowStep(increment), parentFragmentThreshold(parentThreshold), divRThreshold(div) {} +Slayer::Slayer(int win, int increment, int parentThreshold, float div, int i, int snp, int mi) : + minBS(mi), windowSize(win), windowStep(increment), parentFragmentThreshold(parentThreshold), divRThreshold(div), iters(i), percentSNPSample(snp){ m = MothurOut::getInstance(); } /***********************************************************************/ -void Slayer::getResults(Sequence* query, vector refSeqs) { +string Slayer::getResults(Sequence query, vector refSeqs) { try { - + vector all; all.clear(); + myQuery = query; + for (int i = 0; i < refSeqs.size(); i++) { for (int j = i+1; j < refSeqs.size(); j++) { + if (m->control_pressed) { return "no"; } + //make copies of query and each parent because runBellerophon removes gaps and messes them up - Sequence* q = new Sequence(query->getName(), query->getAligned()); - Sequence* leftParent = new Sequence(refSeqs[i]->getName(), refSeqs[i]->getAligned()); - Sequence* rightParent = new Sequence(refSeqs[j]->getName(), refSeqs[j]->getAligned()); + Sequence q(query.getName(), query.getAligned()); + Sequence leftParent(refSeqs[i].getName(), refSeqs[i].getAligned()); + Sequence rightParent(refSeqs[j].getName(), refSeqs[j].getAligned()); - vector divs = runBellerophon(q, leftParent, rightParent); - - delete q; - delete leftParent; - delete rightParent; + //cout << q->getName() << endl << q->getAligned() << endl << endl; + //cout << leftParent.getName() << '\t' << leftParent.getAligned().length() << endl << endl; + //cout << rightParent.getName() << '\t' << rightParent.getAligned().length() << endl << endl; + //cout << q.getName() << '\t' << q.getAligned().length() << endl << endl; + //cout << rightParent->getName() << endl << rightParent->getAligned() << endl << endl; + //cout << " length = " << rightParent->getAligned().length() << endl; + + map spots; //map from spot in original sequence to spot in filtered sequence for query and both parents + vector divs = runBellerophon(q, leftParent, rightParent, spots); + + if (m->control_pressed) { return "no"; } +// cout << "examining:\t" << refSeqs[i]->getName() << '\t' << refSeqs[j]->getName() << endl; + vector selectedDivs; + for (int k = 0; k < divs.size(); k++) { + + vector snpsLeft = getSNPS(divs[k].parentA.getAligned(), divs[k].querySeq.getAligned(), divs[k].parentB.getAligned(), divs[k].winLStart, divs[k].winLEnd); + vector snpsRight = getSNPS(divs[k].parentA.getAligned(), divs[k].querySeq.getAligned(), divs[k].parentB.getAligned(), divs[k].winRStart, divs[k].winREnd); + + if (m->control_pressed) { return "no"; } + + int numSNPSLeft = snpsLeft.size(); + int numSNPSRight = snpsRight.size(); + +// cout << numSNPSLeft << '\t' << numSNPSRight << endl; + //require at least 4 SNPs on each side of the break + if ((numSNPSLeft >= 4) && (numSNPSRight >= 4)) { + + float BS_A, BS_B; + bootstrapSNPS(snpsLeft, snpsRight, BS_A, BS_B, iters); + + if (m->control_pressed) { return "no"; } + + divs[k].bsa = BS_A; + divs[k].bsb = BS_B; + divs[k].bsMax = max(BS_A, BS_B); + divs[k].chimeraMax = max(divs[k].qla_qrb, divs[k].qlb_qra); + + + //are we within 10 points of the bootstrap cutoff? +// if ((divs[k].bsMax >= (minBS-10)) && (iters < 1000)) { +// bootstrapSNPS(snpsLeft, snpsRight, BS_A, BS_B, 1000); +// +// if (m->control_pressed) { delete q; delete leftParent; delete rightParent; return "no"; } +// +// divs[k].bsa = BS_A; +// divs[k].bsb = BS_B; +// divs[k].bsMax = max(BS_A, BS_B); +// divs[k].chimeraMax = max(divs[k].qla_qrb, divs[k].qlb_qra); +// } + + //so results reflect orignal alignment + divs[k].winLStart = spots[divs[k].winLStart]; + divs[k].winLEnd = spots[divs[k].winLEnd]; + divs[k].winRStart = spots[divs[k].winRStart]; + divs[k].winREnd = spots[divs[k].winREnd]; + + selectedDivs.push_back(divs[k]); + } + } + + //save selected + for (int mi = 0; mi < selectedDivs.size(); mi++) { all.push_back(selectedDivs[mi]); } } - } - + + // compute bootstrap support + if (all.size() > 0) { + //sort them + sort(all.begin(), all.end(), compareDataStruct); + reverse(all.begin(), all.end()); + + outputResults = all; + return "yes"; + }else { + outputResults = all; + return "no"; + } } catch(exception& e) { - errorOut(e, "Slayer", "getResults"); + m->errorOut(e, "Slayer", "getResults"); exit(1); } } /***********************************************************************/ -vector Slayer::runBellerophon(Sequence* query, Sequence* parentA, Sequence* parentB) { +vector Slayer::runBellerophon(Sequence q, Sequence pA, Sequence pB, map& spots) { try{ vector data; //vertical filter - vector temp; - verticalFilter(temp); + //cout << q.getName() << endl << q.getAligned() << endl << endl; + //cout << pA.getName() << endl << pA.getUnaligned() << endl << endl; + //cout << pB.getName() << endl << pB.getUnaligned() << endl << endl; + + //maps spot in new alignment to spot in alignment before filter + spots = verticalFilter(q, pA, pB); //fills baseSpots + + //get these to avoid numerous function calls + string query = q.getAligned(); + string parentA = pA.getAligned(); + string parentB = pB.getAligned(); + int length = query.length(); +//cout << q.getName() << endl << q.getAligned() << endl << endl; +//cout << pA.getName() << endl << pA.getUnaligned() << endl << endl; +//cout << pB.getName() << endl << pB.getUnaligned() << endl << endl; +//cout << " length = " << length << endl; + + //check window size + if (length < (2*windowSize+windowStep)) { +// m->mothurOut("Your window size is too large for " + q->getName() + ". I will make the window size " + toString(length/4) + " which is 1/4 the filtered length."); m->mothurOutEndLine(); + windowSize = length / 4; + } + + for (int i = windowSize-1; i <= (length - windowSize); i += windowStep) { + + if (m->control_pressed) { return data; } - int alignLength = query->getAligned().length(); + int breakpoint = i; + int leftLength = breakpoint + 1; + int rightLength = length - leftLength; + + float QLA = computePercentID(query, parentA, 0, breakpoint); + float QRB = computePercentID(query, parentB, breakpoint+1, length-1); + + float QLB = computePercentID(query, parentB, 0, breakpoint); + float QRA = computePercentID(query, parentA, breakpoint+1, length-1); + + float LAB = computePercentID(parentA, parentB, 0, breakpoint); + float RAB = computePercentID(parentA, parentB, breakpoint+1, length-1); + + float AB = ((LAB*leftLength) + (RAB*rightLength)) / (float) length; + float QA = ((QLA*leftLength) + (QRA*rightLength)) / (float) length; + float QB = ((QLB*leftLength) + (QRB*rightLength)) / (float) length; + + float QLA_QRB = ((QLA*leftLength) + (QRB*rightLength)) / (float) length; + float QLB_QRA = ((QLB*leftLength) + (QRA*rightLength)) / (float) length; + //in original and not used + //float avgQA_QB = ((QA*leftLength) + (QB*rightLength)) / (float) length; + float divR_QLA_QRB = min((QLA_QRB/QA), (QLA_QRB/QB)); + float divR_QLB_QRA = min((QLB_QRA/QA), (QLB_QRA/QB)); + + + //cout << q->getName() << '\t'; + //cout << pA->getName() << '\t'; + //cout << pB->getName() << '\t'; + //cout << "bp: " << breakpoint << " CHIM_TYPE_A\t" << divR_QLA_QRB << "\tQLA: " << QLA << "\tQRB: " << QRB << "\tQLA_QRB: " << QLA_QRB; + //cout << "\tCHIM_TYPE_B\t" << divR_QLB_QRA << "\tQLB: " << QLB << "\tQRA: " << QRA << "\tQLB_QRA: " << QLB_QRA << endl; +//cout << leftLength << '\t' << rightLength << '\t' << QLA << '\t' << QRB << '\t' << QLB << '\t' << QRA << '\t' << LAB << '\t' << RAB << '\t' << AB << '\t' << QA << '\t' << QB << '\t' << QLA_QRB << '\t' << QLB_QRA << endl; + +//cout << divRThreshold << endl; +//cout << breakpoint << '\t' << divR_QLA_QRB << '\t' << divR_QLB_QRA << endl; + //is one of them above the + if (divR_QLA_QRB >= divRThreshold || divR_QLB_QRA >= divRThreshold) { + + if (((QLA_QRB > QA) && (QLA_QRB > QB) && (QLA >= parentFragmentThreshold) && (QRB >= parentFragmentThreshold)) || + ((QLB_QRA > QA) && (QLB_QRA > QB) && (QLB >=parentFragmentThreshold) && (QRA >= parentFragmentThreshold))) { + + data_struct member; + + member.divr_qla_qrb = divR_QLA_QRB; + member.divr_qlb_qra = divR_QLB_QRA; + member.qla_qrb = QLA_QRB; + member.qlb_qra = QLB_QRA; + member.qla = QLA; + member.qrb = QRB; + member.ab = AB; + member.qa = QA; + member.qb = QB; + member.lab = LAB; + member.rab = RAB; + member.qra = QRA; + member.qlb = QLB; + member.winLStart = 0; + member.winLEnd = breakpoint; + member.winRStart = breakpoint+1; + member.winREnd = length-1; + member.querySeq = q; + member.parentA = pA; + member.parentB = pB; + member.bsa = 0; + member.bsb = 0; + member.bsMax = 0; + member.chimeraMax = 0; + + data.push_back(member); + + }//if + }//if + }//for return data; } catch(exception& e) { - errorOut(e, "Slayer", "runBellerophon"); + m->errorOut(e, "Slayer", "runBellerophon"); exit(1); } } /***********************************************************************/ -float Slayer::computePercentID(string queryFrag, string parent, int left, int right) { +vector Slayer::getSNPS(string parentA, string query, string parentB, int left, int right) { try { - int total = 0; - int matches = 0; + vector data; + for (int i = left; i <= right; i++) { - total++; - if (queryFrag[i] == parent[i]) { - matches++; + + char A = parentA[i]; + char Q = query[i]; + char B = parentB[i]; + + if ((A != Q) || (B != Q)) { + + //ensure not neighboring a gap. change to 12/09 release of chimeraSlayer - not sure what this adds, but it eliminates alot of SNPS + + + if ( + //did query loose a base here during filter?? + ( i == 0 || abs (baseSpots[0][i] - baseSpots[0][i-1]) == 1) && + ( i == query.length()-1 || abs (baseSpots[0][i] - baseSpots[0][i+1]) == 1) + && + //did parentA loose a base here during filter?? + ( i == 0 || abs (baseSpots[1][i] - baseSpots[1][i-1]) == 1) && + ( i == parentA.length()-1 || abs (baseSpots[1][i] - baseSpots[1][i+1]) == 1) + && + //did parentB loose a base here during filter?? + ( i == 0 || abs (baseSpots[2][i] - baseSpots[2][i-1]) == 1) && + ( i == parentB.length()-1 || abs (baseSpots[2][i] - baseSpots[2][i+1]) == 1) + ) + { + snps member; + member.queryChar = Q; + member.parentAChar = A; + member.parentBChar = B; + data.push_back(member); + } } } - - float percentID =( matches/(float)total) * 100; - return percentID; + return data; + } catch(exception& e) { - errorOut(e, "Slayer", "computePercentID"); + m->errorOut(e, "Slayer", "getSNPS"); exit(1); } } /***********************************************************************/ -//this is a vertical filter -void Slayer::verticalFilter(vector seqs) { +int Slayer::bootstrapSNPS(vector left, vector right, float& BSA, float& BSB, int numIters) { try { - vector gaps; gaps.resize(seqs[0]->getAligned().length(), 0); + + srand((unsigned)time( NULL )); + + int count_A = 0; // sceneario QLA,QRB supported + int count_B = 0; // sceneario QLB,QRA supported + + int numLeft = max(1, int(left.size() * percentSNPSample/(float)100 + 0.5)); + int numRight = max(1, int(right.size() * percentSNPSample/(float)100 + 0.5)); + + for (int i = 0; i < numIters; i++) { + //random sampling with replacement. + + if (m->control_pressed) { return 0; } + + vector selectedLeft; + + for (int j = 0; j < numLeft; j++) { + int index = int(rand() % left.size()); + selectedLeft.push_back(left[index]); + } + + vector selectedRight; + for (int j = 0; j < numRight; j++) { + int index = int(rand() % right.size()); + selectedRight.push_back(right[index]); + } - string filterString = (string(seqs[0]->getAligned().length(), '1')); + /* A ------------------------------------------ + # QLA QRA + # Q ------------------------------------------ + # | + # | + # Q ------------------------------------------ + # QLB QRB + # B ------------------------------------------ */ - //for each sequence - for (int i = 0; i < seqs.size(); i++) { - string seqAligned = seqs[i]->getAligned(); + float QLA = snpQA(selectedLeft); + float QRA = snpQA(selectedRight); + + float QLB = snpQB(selectedLeft); + float QRB = snpQB(selectedRight); + + //in original - not used - not sure why? + //float ALB = snpAB(selectedLeft); + //float ARB = snpAB(selectedRight); + + if ((QLA > QLB) && (QRB > QRA)) { + count_A++; + } + + if ((QLB > QLA) && (QRA > QRB)) { + count_B++; + } - for (int j = 0; j < seqAligned.length(); j++) { - //if this spot is a gap - if ((seqAligned[j] == '-') || (seqAligned[j] == '.')) { gaps[j]++; } +//cout << "selected left snp: \n"; +//for (int j = 0; j < selectedLeft.size(); j++) { cout << selectedLeft[j].parentAChar; } +//cout << endl; +//for (int j = 0; j < selectedLeft.size(); j++) { cout << selectedLeft[j].queryChar; } +//cout << endl; +//for (int j = 0; j < selectedLeft.size(); j++) { cout << selectedLeft[j].parentBChar; } +//cout << endl; +//cout << "selected right snp: \n"; +//for (int j = 0; j < selectedRight.size(); j++) { cout << selectedRight[j].parentAChar; } +//cout << endl; +//for (int i = 0; i < selectedRight.size(); i++) { cout << selectedRight[i].queryChar; } +//cout << endl; +//for (int i = 0; i < selectedRight.size(); i++) { cout << selectedRight[i].parentBChar; } +//cout << endl; + } + + + //cout << count_A << '\t' << count_B << endl; + + BSA = (float) count_A / (float) numIters * 100; + BSB = (float) count_B / (float) numIters * 100; +//cout << "bsa = " << BSA << " bsb = " << BSB << endl; + + return 0; + + } + catch(exception& e) { + m->errorOut(e, "Slayer", "bootstrapSNPS"); + exit(1); + } +} +/***********************************************************************/ +float Slayer::snpQA(vector data) { + try { + + int numIdentical = 0; + + for (int i = 0; i < data.size(); i++) { + if (data[i].parentAChar == data[i].queryChar) { + numIdentical++; + } + } + + float percentID = (numIdentical / (float) data.size()) * 100; + + return percentID; + } + catch(exception& e) { + m->errorOut(e, "Slayer", "snpQA"); + exit(1); + } +} +/***********************************************************************/ +float Slayer::snpQB(vector data) { + try { + + int numIdentical = 0; + + for (int i = 0; i < data.size(); i++) { + if (data[i].parentBChar == data[i].queryChar) { + numIdentical++; + } + } + + float percentID = (numIdentical / (float) data.size()) * 100; + + return percentID; + + } + catch(exception& e) { + m->errorOut(e, "Slayer", "snpQB"); + exit(1); + } +} +/***********************************************************************/ +float Slayer::snpAB(vector data) { + try { + int numIdentical = 0; + + for (int i = 0; i < data.size(); i++) { + if (data[i].parentAChar == data[i].parentBChar) { + numIdentical++; + } + } + + float percentID = (numIdentical / (float) data.size()) * 100; + + return percentID; + + } + catch(exception& e) { + m->errorOut(e, "Slayer", "snpAB"); + exit(1); + } +} +/***********************************************************************/ +float Slayer::computePercentID(string queryAlign, string chimera, int left, int right) { + try { + + int numIdentical = 0; + int countA = 0; + int countB = 0; + for (int i = left; i <= right; i++) { + if (((queryAlign[i] != 'G') && (queryAlign[i] != 'T') && (queryAlign[i] != 'A') && (queryAlign[i] != 'C')&& (queryAlign[i] != '.') && (queryAlign[i] != '-')) || + ((chimera[i] != 'G') && (chimera[i] != 'T') && (chimera[i] != 'A') && (chimera[i] != 'C')&& (chimera[i] != '.') && (chimera[i] != '-'))) {} + else { + + bool charA = false; bool charB = false; + if ((queryAlign[i] == 'G') || (queryAlign[i] == 'T') || (queryAlign[i] == 'A') || (queryAlign[i] == 'C')) { charA = true; } + if ((chimera[i] == 'G') || (chimera[i] == 'T') || (chimera[i] == 'A') || (chimera[i] == 'C')) { charB = true; } + + if (charA || charB) { + + if (charA) { countA++; } + if (charB) { countB++; } + + if (queryAlign[i] == chimera[i]) { + numIdentical++; + } + } } + + } + + float numBases = (countA + countB) /(float) 2; + + if (numBases == 0) { return 0; } + + float percentIdentical = (numIdentical/(float)numBases) * 100; + + return percentIdentical; + + } + catch(exception& e) { + m->errorOut(e, "Slayer", "computePercentID"); + exit(1); + } +} +/***********************************************************************/ +//remove columns that contain any gaps +map Slayer::verticalFilter(Sequence& q, Sequence& pA, Sequence& pB) { + try { + //find baseSpots + baseSpots.clear(); + baseSpots.resize(3); //query, parentA, parentB + + vector gaps; gaps.resize(q.getAligned().length(), 0); + + string filterString = (string(q.getAligned().length(), '1')); + + string seqAligned = q.getAligned(); + for (int j = 0; j < seqAligned.length(); j++) { + //if this spot is a gap + if ((seqAligned[j] == '-') || (seqAligned[j] == '.') || (toupper(seqAligned[j]) == 'N')) { gaps[j]++; } + } + + seqAligned = pA.getAligned(); + for (int j = 0; j < seqAligned.length(); j++) { + //if this spot is a gap + if ((seqAligned[j] == '-') || (seqAligned[j] == '.') || (toupper(seqAligned[j]) == 'N')) { gaps[j]++; } } - //zero out spot where all sequences have blanks + seqAligned = pB.getAligned(); + for (int j = 0; j < seqAligned.length(); j++) { + //if this spot is a gap + if ((seqAligned[j] == '-') || (seqAligned[j] == '.') || (toupper(seqAligned[j]) == 'N')) { gaps[j]++; } + } + + + //zero out spot where any sequences have blanks int numColRemoved = 0; - for(int i = 0; i < seqs[0]->getAligned().length(); i++){ - if(gaps[i] == seqs.size()) { filterString[i] = '0'; numColRemoved++; } + int count = 0; + map maskMap; maskMap.clear(); + + for(int i = 0; i < q.getAligned().length(); i++){ + if(gaps[i] != 0) { filterString[i] = '0'; numColRemoved++; } + else { + maskMap[count] = i; + count++; + } + } + + seqAligned = q.getAligned(); + string newAligned = ""; + + int baseCount = 0; + count = 0; + for (int j = 0; j < seqAligned.length(); j++) { + //are you a base + if ((seqAligned[j] != '-') && (seqAligned[j] != '.') && (toupper(seqAligned[j]) != 'N')) { baseCount++; } + + //if this spot is not a gap + if (filterString[j] == '1') { + newAligned += seqAligned[j]; + baseSpots[0][count] = baseCount; + count++; + } } + + q.setAligned(newAligned); - //for each sequence - for (int i = 0; i < seqs.size(); i++) { + seqAligned = pA.getAligned(); + newAligned = ""; - string seqAligned = seqs[i]->getAligned(); - string newAligned = ""; + baseCount = 0; + count = 0; + for (int j = 0; j < seqAligned.length(); j++) { + //are you a base + if ((seqAligned[j] != '-') && (seqAligned[j] != '.') && (toupper(seqAligned[j]) != 'N')) { baseCount++; } - for (int j = 0; j < seqAligned.length(); j++) { - //if this spot is not a gap - if (filterString[j] == '1') { newAligned += seqAligned[j]; } + //if this spot is not a gap + if (filterString[j] == '1') { + newAligned += seqAligned[j]; + baseSpots[1][count] = baseCount; + count++; } + } + + pA.setAligned(newAligned); + + seqAligned = pB.getAligned(); + newAligned = ""; + + baseCount = 0; + count = 0; + for (int j = 0; j < seqAligned.length(); j++) { + //are you a base + if ((seqAligned[j] != '-') && (seqAligned[j] != '.') && (toupper(seqAligned[j]) != 'N')) { baseCount++; } - seqs[i]->setAligned(newAligned); + //if this spot is not a gap + if (filterString[j] == '1') { + newAligned += seqAligned[j]; + baseSpots[2][count] = baseCount; + count++; + } } + + pB.setAligned(newAligned); + + + return maskMap; } catch(exception& e) { - errorOut(e, "Slayer", "verticalFilter"); + m->errorOut(e, "Slayer", "verticalFilter"); exit(1); } }