#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) :
+ windowSize(win), windowStep(increment), parentFragmentThreshold(parentThreshold), divRThreshold(div), iters(i), percentSNPSample(snp){ m = MothurOut::getInstance(); }
/***********************************************************************/
string Slayer::getResults(Sequence* query, vector<Sequence*> refSeqs) {
try {
-cout << "refSeqs = " << refSeqs.size() << endl;
vector<data_struct> all; all.clear();
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());
- vector<data_struct> divs = runBellerophon(q, leftParent, rightParent);
+ 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";
+ }
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";
+ }
+
int numSNPSLeft = snpsLeft.size();
int numSNPSRight = snpsRight.size();
//require at least 3 SNPs on each side of the break
if ((numSNPSLeft >= 3) && (numSNPSRight >= 3)) {
+
+ //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;
- 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);
+ //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 ) {
+ //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";
+ }
+
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]);
- }
+ //}
}
}
}
}
-
+
// compute bootstrap support
if (all.size() > 0) {
//sort them
}
}
catch(exception& e) {
- errorOut(e, "Slayer", "getResults");
+ m->errorOut(e, "Slayer", "getResults");
exit(1);
}
}
/***********************************************************************/
-vector<data_struct> Slayer::runBellerophon(Sequence* q, Sequence* pA, Sequence* pB) {
+vector<data_struct> Slayer::runBellerophon(Sequence* q, Sequence* pA, Sequence* pB, map<int, int>& spots) {
try{
vector<data_struct> data;
-cout << q->getName() << '\t' << q->getAligned().length() << endl;
+
//vertical filter
vector<Sequence*> temp;
temp.push_back(q); temp.push_back(pA); temp.push_back(pB);
- verticalFilter(temp);
+
+ //maps spot in new alignment to spot in alignment before filter
+ spots = verticalFilter(temp); //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() << '\t' << length << endl;
-
+//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;
+
//check window size
if (length < (2*windowSize+windowStep)) {
- 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."); mothurOutEndLine();
+ 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 breakpoint = i;
int leftLength = breakpoint + 1;
int rightLength = length - leftLength;
//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 << 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) {
member.qra = QRA;
member.qlb = QLB;
member.winLStart = 0;
- member.winLEnd = breakpoint;
+ member.winLEnd = breakpoint;
member.winRStart = breakpoint+1;
member.winREnd = length-1;
member.querySeq = *(q);
}
catch(exception& e) {
- errorOut(e, "Slayer", "runBellerophon");
+ m->errorOut(e, "Slayer", "runBellerophon");
exit(1);
}
}
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)) {
- snps member;
- member.queryChar = Q;
- member.parentAChar = A;
- member.parentBChar = B;
+//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)
+ &&
+ //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)
+ &&
+ //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)
+ )
+ {
- data.push_back(member);
+ 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);
+ }
}
}
}
catch(exception& e) {
- errorOut(e, "Slayer", "getSNPS");
+ m->errorOut(e, "Slayer", "getSNPS");
exit(1);
}
}
/***********************************************************************/
-void 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) {
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()/10 +0.5));
- int numRight = max(1, int(right.size()/10 + 0.5));
-
- for (int i = 0; i < 100; i++) {
+ 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++) {
//random sampling with replacement.
+ if (m->control_pressed) { return 0; }
+
vector<snps> selectedLeft;
+
for (int j = 0; j < numLeft; j++) {
int index = int(rand() % left.size());
selectedLeft.push_back(left[index]);
}
-
+
vector<snps> selectedRight;
for (int j = 0; j < numRight; j++) {
int index = int(rand() % right.size());
float QLB = snpQB(selectedLeft);
float QRB = snpQB(selectedRight);
-
+
//in original - not used - not sure why?
//float ALB = snpAB(selectedLeft);
//float ARB = snpAB(selectedRight);
if ((QLB > QLA) && (QRA > QRB)) {
count_B++;
}
-
+
+//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;
}
- BSA = (float) count_A;
- BSB = (float) count_B;
+
+
+
+ BSA = ((float) count_A / (float) iters) * 100;
+ BSB = ((float) count_B / (float) iters) * 100;
+//cout << "bsa = " << BSA << " bsb = " << BSB << endl;
+
+ return 0;
}
catch(exception& e) {
- errorOut(e, "Slayer", "bootstrapSNPS");
+ m->errorOut(e, "Slayer", "bootstrapSNPS");
exit(1);
}
}
}
}
- float percentID = (numIdentical / data.size()) * 100;
+ float percentID = (numIdentical / (float) data.size()) * 100;
return percentID;
}
catch(exception& e) {
- errorOut(e, "Slayer", "snpQA");
+ m->errorOut(e, "Slayer", "snpQA");
exit(1);
}
}
}
}
- float percentID = (numIdentical / data.size()) * 100;
+ float percentID = (numIdentical / (float) data.size()) * 100;
return percentID;
}
catch(exception& e) {
- errorOut(e, "Slayer", "snpQB");
+ m->errorOut(e, "Slayer", "snpQB");
exit(1);
}
}
}
}
- float percentID = (numIdentical / data.size()) * 100;
+ float percentID = (numIdentical / (float) data.size()) * 100;
return percentID;
}
catch(exception& e) {
- errorOut(e, "Slayer", "snpAB");
+ m->errorOut(e, "Slayer", "snpAB");
exit(1);
}
}
try {
int total = 0;
int matches = 0;
-
+
for (int i = left; i <= right; i++) {
total++;
if (queryFrag[i] == parent[i]) {
return percentID;
}
catch(exception& e) {
- errorOut(e, "Slayer", "computePercentID");
+ m->errorOut(e, "Slayer", "computePercentID");
exit(1);
}
}
/***********************************************************************/
//remove columns that contain any gaps
-void Slayer::verticalFilter(vector<Sequence*> seqs) {
+map<int, int> Slayer::verticalFilter(vector<Sequence*> seqs) {
try {
+ //find baseSpots
+ baseSpots.clear();
+ baseSpots.resize(3); //query, parentA, parentB
+
vector<int> gaps; gaps.resize(seqs[0]->getAligned().length(), 0);
string filterString = (string(seqs[0]->getAligned().length(), '1'));
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]++; }
+ if ((seqAligned[j] == '-') || (seqAligned[j] == '.') || (toupper(seqAligned[j]) == 'N')) { gaps[j]++; }
}
}
- //zero out spot where all sequences have blanks
+ //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++){
if(gaps[i] != 0) { filterString[i] = '0'; numColRemoved++; }
+ else {
+ maskMap[count] = i;
+ count++;
+ }
}
//for each sequence
string seqAligned = seqs[i]->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++; }
+
//if this spot is not a gap
- if (filterString[j] == '1') { newAligned += seqAligned[j]; }
+ if (filterString[j] == '1') {
+ newAligned += seqAligned[j];
+ baseSpots[i][count] = baseCount;
+ count++;
+ }
}
seqs[i]->setAligned(newAligned);
}
+
+ return maskMap;
}
catch(exception& e) {
- errorOut(e, "Slayer", "verticalFilter");
+ m->errorOut(e, "Slayer", "verticalFilter");
exit(1);
}
}