#include "slayer.h"
/***********************************************************************/
-Slayer::Slayer(int win, int increment, int parentThreshold, float div, int i, int snp) :
- windowSize(win), windowStep(increment), parentFragmentThreshold(parentThreshold), divRThreshold(div), iters(i), percentSNPSample(snp){ m = MothurOut::getInstance(); }
+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(); }
/***********************************************************************/
-string Slayer::getResults(Sequence* query, vector<Sequence*> refSeqs) {
+string Slayer::getResults(Sequence query, vector<Sequence> refSeqs) {
try {
vector<data_struct> 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());
+
+ //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<int, int> spots; //map from spot in original sequence to spot in filtered sequence for query and both parents
vector<data_struct> divs = runBellerophon(q, leftParent, rightParent, spots);
-
- if (m->control_pressed) {
- delete q;
- delete leftParent;
- delete rightParent;
- return "no";
- }
-
-// cout << divs.size() << endl;
+
+ if (m->control_pressed) { return "no"; }
+// cout << "examining:\t" << refSeqs[i]->getName() << '\t' << refSeqs[j]->getName() << endl;
vector<data_struct> selectedDivs;
for (int k = 0; k < divs.size(); k++) {
vector<snps> snpsLeft = getSNPS(divs[k].parentA.getAligned(), divs[k].querySeq.getAligned(), divs[k].parentB.getAligned(), divs[k].winLStart, divs[k].winLEnd);
vector<snps> snpsRight = getSNPS(divs[k].parentA.getAligned(), divs[k].querySeq.getAligned(), divs[k].parentB.getAligned(), divs[k].winRStart, divs[k].winREnd);
-
- if (m->control_pressed) {
- delete q;
- delete leftParent;
- delete rightParent;
- return "no";
- }
+
+ if (m->control_pressed) { return "no"; }
int numSNPSLeft = snpsLeft.size();
int numSNPSRight = snpsRight.size();
- //require at least 3 SNPs on each side of the break
-// if ((numSNPSLeft >= 3) && (numSNPSRight >= 3)) {
+// cout << numSNPSLeft << '\t' << numSNPSRight << endl;
+ //require at least 4 SNPs on each side of the break
if ((numSNPSLeft >= 4) && (numSNPSRight >= 4)) {
-
- //removed in 12/09 version of chimeraSlayer
- //int winSizeLeft = divs[k].winLEnd - divs[k].winLStart + 1;
- //int winSizeRight = divs[k].winREnd - divs[k].winRStart + 1;
-
- //float snpRateLeft = numSNPSLeft / (float) winSizeLeft;
- //float snpRateRight = numSNPSRight / (float) winSizeRight;
- //float logR = log(snpRateLeft / snpRateRight) / log(2.0);
-
- // do not accept excess snp ratio on either side of the break
- //if (abs(logR) < 1 ) {
- float BS_A, BS_B;
- bootstrapSNPS(snpsLeft, snpsRight, BS_A, BS_B);
-
- if (m->control_pressed) {
- delete q;
- delete leftParent;
- delete rightParent;
- return "no";
- }
+ 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].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);
- 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];
+ //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);
+// }
- selectedDivs.push_back(divs[k]);
- //}
+ //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 m = 0; m < selectedDivs.size(); m++) { all.push_back(selectedDivs[m]); }
-
- delete q;
- delete leftParent;
- delete rightParent;
+ 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 {
+ }else {
outputResults = all;
return "no";
}
}
}
/***********************************************************************/
-vector<data_struct> Slayer::runBellerophon(Sequence* q, Sequence* pA, Sequence* pB, map<int, int>& spots) {
+vector<data_struct> Slayer::runBellerophon(Sequence q, Sequence pA, Sequence pB, map<int, int>& spots) {
try{
vector<data_struct> data;
//vertical filter
- vector<Sequence*> temp;
- temp.push_back(q); temp.push_back(pA); temp.push_back(pB);
+ //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(temp); //fills baseSpots
+ 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();
+ 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->getAligned() << endl << endl;
-cout << pB->getName() << endl << pB->getAligned() << endl << endl;
-cout << " length = " << length << endl;
-cout << q->getName() << endl;
-cout << pA->getName() << '\t';
-cout << pB->getName() << endl;*/
+//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)) {
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 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 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 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 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;
member.winLEnd = breakpoint;
member.winRStart = breakpoint+1;
member.winREnd = length-1;
- member.querySeq = *(q);
- member.parentA = *(pA);
- member.parentB = *(pB);
+ member.querySeq = q;
+ member.parentA = pA;
+ member.parentB = pB;
member.bsa = 0;
member.bsb = 0;
member.bsMax = 0;
}//if
}//for
+
return data;
}
try {
vector<snps> data;
-//cout << left << '\t' << right << endl;
+
for (int i = left; i <= right; i++) {
char A = parentA[i];
char B = parentB[i];
if ((A != Q) || (B != Q)) {
-//cout << "not equal " << Q << '\t' << A << '\t' << B << endl;
-
+
//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() || 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() || 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() || 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;
-//cout << "not neighboring a gap " << Q << '\t' << A << '\t' << B << '\t' << baseSpots[0][i] << '\t' << baseSpots[0][i+1] << '\t' << baseSpots[0][i-1] << '\t' << baseSpots[1][i] << '\t' << baseSpots[1][i+1] << '\t' << baseSpots[1][i-1] << '\t' << baseSpots[2][i] << '\t' << baseSpots[2][i+1] << '\t' << baseSpots[2][i-1] << endl;
data.push_back(member);
}
}
-// cout << i << '\t' << data.size() << endl;
}
return data;
}
}
/***********************************************************************/
-int Slayer::bootstrapSNPS(vector<snps> left, vector<snps> right, float& BSA, float& BSB) {
+int Slayer::bootstrapSNPS(vector<snps> left, vector<snps> right, float& BSA, float& BSB, int numIters) {
try {
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));
+ 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 < iters; i++) {
+ for (int i = 0; i < numIters; i++) {
//random sampling with replacement.
if (m->control_pressed) { return 0; }
}
+ //cout << count_A << '\t' << count_B << endl;
-
- BSA = ((float) count_A / (float) iters) * 100;
- BSB = ((float) count_B / (float) iters) * 100;
+ BSA = (float) count_A / (float) numIters * 100;
+ BSB = (float) count_B / (float) numIters * 100;
//cout << "bsa = " << BSA << " bsb = " << BSB << endl;
return 0;
}
}
/***********************************************************************/
-float Slayer::computePercentID(string queryFrag, string parent, int left, int right) {
+float Slayer::computePercentID(string queryAlign, string chimera, int left, int right) {
try {
- int total = 0;
- int matches = 0;
-
+
+ int numIdentical = 0;
+ int countA = 0;
+ int countB = 0;
for (int i = left; i <= right; i++) {
- total++;
- if (queryFrag[i] == parent[i]) {
- matches++;
+ 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;
- float percentID =( matches/(float)total) * 100;
+ return percentIdentical;
- return percentID;
}
catch(exception& e) {
m->errorOut(e, "Slayer", "computePercentID");
}
/***********************************************************************/
//remove columns that contain any gaps
-map<int, int> Slayer::verticalFilter(vector<Sequence*> seqs) {
+map<int, int> Slayer::verticalFilter(Sequence& q, Sequence& pA, Sequence& pB) {
try {
//find baseSpots
baseSpots.clear();
baseSpots.resize(3); //query, parentA, parentB
- vector<int> gaps; gaps.resize(seqs[0]->getAligned().length(), 0);
+ vector<int> gaps; gaps.resize(q.getAligned().length(), 0);
- string filterString = (string(seqs[0]->getAligned().length(), '1'));
+ string filterString = (string(q.getAligned().length(), '1'));
- //for each sequence
- for (int i = 0; i < seqs.size(); i++) {
+ 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]++; }
+ }
- string seqAligned = seqs[i]->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]++; }
+ }
+
+ 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;
int count = 0;
map<int, int> maskMap; maskMap.clear();
- for(int i = 0; i < seqs[0]->getAligned().length(); i++){
+ for(int i = 0; i < q.getAligned().length(); i++){
if(gaps[i] != 0) { filterString[i] = '0'; numColRemoved++; }
else {
maskMap[count] = i;
}
}
- //for each sequence
- for (int i = 0; i < seqs.size(); i++) {
-
- string seqAligned = seqs[i]->getAligned();
- string newAligned = "";
+ seqAligned = q.getAligned();
+ string newAligned = "";
- int baseCount = 0;
- int count = 0;
- for (int j = 0; j < seqAligned.length(); j++) {
- //are you a base
- if ((seqAligned[j] != '-') && (seqAligned[j] != '.') && (toupper(seqAligned[j]) != 'N')) { baseCount++; }
+ 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[i][count] = baseCount;
- count++;
- }
+ //if this spot is not a gap
+ if (filterString[j] == '1') {
+ newAligned += seqAligned[j];
+ baseSpots[0][count] = baseCount;
+ count++;
}
+ }
- seqs[i]->setAligned(newAligned);
+ q.setAligned(newAligned);
+
+ seqAligned = pA.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++; }
+
+ //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++; }
+
+ //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) {