5 * Created by Sarah Westcott on 7/9/09.
6 * Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
10 #include "bellerophon.h"
11 #include "eachgapdist.h"
12 #include "ignoregaps.h"
13 #include "onegapdist.h"
16 //***************************************************************************************************************
18 Bellerophon::Bellerophon(string name) {
23 errorOut(e, "Bellerophon", "Bellerophon");
28 //***************************************************************************************************************
29 void Bellerophon::print(ostream& out) {
32 out << "Name\tScore\tLeft\tRight\t" << endl;
33 //output prefenence structure to .chimeras file
34 for (int i = 0; i < pref.size(); i++) {
35 out << pref[i].name << '\t' << setprecision(3) << pref[i].score[0] << '\t' << pref[i].leftParent[0] << '\t' << pref[i].rightParent[0] << endl;
37 //calc # of seqs with preference above 1.0
38 if (pref[i].score[0] > 1.0) {
40 mothurOut(pref[i].name + " is a suspected chimera at breakpoint " + toString(pref[i].midpoint)); mothurOutEndLine();
41 mothurOut("It's score is " + toString(pref[i].score[0]) + " with suspected left parent " + pref[i].leftParent[0] + " and right parent " + pref[i].rightParent[0]); mothurOutEndLine();
45 //output results to screen
47 mothurOut("Sequence with preference score above 1.0: " + toString(above1)); mothurOutEndLine();
50 mothurOut("Minimum:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
51 spot = pref.size() * 0.975;
52 mothurOut("2.5%-tile:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
53 spot = pref.size() * 0.75;
54 mothurOut("25%-tile:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
55 spot = pref.size() * 0.50;
56 mothurOut("Median: \t" + toString(pref[spot].score[0])); mothurOutEndLine();
57 spot = pref.size() * 0.25;
58 mothurOut("75%-tile:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
59 spot = pref.size() * 0.025;
60 mothurOut("97.5%-tile:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
62 mothurOut("Maximum:\t" + toString(pref[spot].score[0])); mothurOutEndLine();
66 errorOut(e, "Bellerophon", "print");
71 //********************************************************************************************************************
72 //sorts highest score to lowest
73 inline bool comparePref(Preference left, Preference right){
74 return (left.score[0] > right.score[0]);
77 //***************************************************************************************************************
78 int Bellerophon::getChimeras() {
83 string optionString = "fasta=" + fastafile + ", soft=50";
84 filterSeqs = new FilterSeqsCommand(optionString);
85 filterSeqs->execute();
88 //reset fastafile to filtered file
89 fastafile = getRootName(fastafile) + "filter.fasta";
92 distCalculator = new eachGapDist();
95 seqs = readSeqs(fastafile);
97 if (unaligned) { mothurOut("Your sequences need to be aligned when you use the bellerophon method."); mothurOutEndLine(); return 1; }
99 int numSeqs = seqs.size();
101 if (numSeqs == 0) { mothurOut("Error in reading you sequences."); mothurOutEndLine(); exit(1); }
103 //set default window to 25% of sequence length
104 string seq0 = seqs[0]->getAligned();
105 if (window == 0) { window = seq0.length() / 4; }
106 else if (window > (seq0.length() / 2)) {
107 mothurOut("Your sequence length is = " + toString(seq0.length()) + ". You have selected a window size greater than the length of half your aligned sequence. I will run it with a window size of " + toString((seq0.length() / 2))); mothurOutEndLine();
108 window = (seq0.length() / 2);
111 if (increment > (seqs[0]->getAlignLength() - (2*window))) {
112 if (increment != 10) {
114 mothurOut("You have selected a increment that is too large. I will use the default."); mothurOutEndLine();
116 if (increment > (seqs[0]->getAlignLength() - (2*window))) { increment = 0; }
118 }else{ increment = 0; }
121 if (increment == 0) { iters = 1; }
122 else { iters = ((seqs[0]->getAlignLength() - (2*window)) / increment); }
125 pref.resize(numSeqs);
127 for (int i = 0; i < numSeqs; i++ ) {
128 pref[i].leftParent.resize(2); pref[i].rightParent.resize(2); pref[i].score.resize(2); pref[i].closestLeft.resize(2); pref[i].closestRight.resize(3);
129 pref[i].name = seqs[i]->getName();
130 pref[i].score[0] = 0.0; pref[i].score[1] = 0.0;
131 pref[i].closestLeft[0] = 100000.0; pref[i].closestLeft[1] = 100000.0;
132 pref[i].closestRight[0] = 100000.0; pref[i].closestRight[1] = 100000.0;
135 int midpoint = window;
137 while (count < iters) {
139 //create 2 vectors of sequences, 1 for left side and one for right side
140 vector<Sequence> left; vector<Sequence> right;
142 for (int i = 0; i < seqs.size(); i++) {
143 //cout << "midpoint = " << midpoint << "\twindow = " << window << endl;
144 //cout << "whole = " << seqs[i]->getAligned().length() << endl;
146 string seqLeft = seqs[i]->getAligned().substr(midpoint-window, window);
148 tempLeft.setName(seqs[i]->getName());
149 tempLeft.setAligned(seqLeft);
150 left.push_back(tempLeft);
151 //cout << "left = " << tempLeft.getAligned().length() << endl;
153 string seqRight = seqs[i]->getAligned().substr(midpoint, window);
155 tempRight.setName(seqs[i]->getName());
156 tempRight.setAligned(seqRight);
157 right.push_back(tempRight);
158 //cout << "right = " << seqRight.length() << endl;
161 //adjust midpoint by increment
162 midpoint += increment;
165 //this should be parallelized
166 //perference = sum of (| distance of my left to sequence j's left - distance of my right to sequence j's right | )
167 //create a matrix containing the distance from left to left and right to right
168 //calculate distances
169 SparseMatrix* SparseLeft = new SparseMatrix();
170 SparseMatrix* SparseRight = new SparseMatrix();
172 createSparseMatrix(0, left.size(), SparseLeft, left);
173 createSparseMatrix(0, right.size(), SparseRight, right);
175 vector<SeqMap> distMapRight;
176 vector<SeqMap> distMapLeft;
178 // Create a data structure to quickly access the distance information.
179 //this is from thallingers reimplementation on get.oturep
180 // It consists of a vector of distance maps, where each map contains
181 // all distances of a certain sequence. Vector and maps are accessed
182 // via the index of a sequence in the distance matrix
183 distMapRight = vector<SeqMap>(numSeqs);
184 distMapLeft = vector<SeqMap>(numSeqs);
185 //cout << "left" << endl << endl;
186 for (MatData currentCell = SparseLeft->begin(); currentCell != SparseLeft->end(); currentCell++) {
187 distMapLeft[currentCell->row][currentCell->column] = currentCell->dist;
188 //cout << " i = " << currentCell->row << " j = " << currentCell->column << " dist = " << currentCell->dist << endl;
190 //cout << "right" << endl << endl;
191 for (MatData currentCell = SparseRight->begin(); currentCell != SparseRight->end(); currentCell++) {
192 distMapRight[currentCell->row][currentCell->column] = currentCell->dist;
193 //cout << " i = " << currentCell->row << " j = " << currentCell->column << " dist = " << currentCell->dist << endl;
199 //fill preference structure
200 generatePreferences(distMapLeft, distMapRight, midpoint);
206 delete distCalculator;
208 //rank preference score to eachother
210 float expectedPercent = 1 / (float) (pref.size());
212 for (int i = 0; i < pref.size(); i++) { dme += pref[i].score[0]; }
214 for (int i = 0; i < pref.size(); i++) {
216 //gives the actual percentage of the dme this seq adds
217 pref[i].score[0] = pref[i].score[0] / dme;
219 //how much higher or lower is this than expected
220 pref[i].score[0] = pref[i].score[0] / expectedPercent;
224 //sort Preferences highest to lowest
225 sort(pref.begin(), pref.end(), comparePref);
230 catch(exception& e) {
231 errorOut(e, "Bellerophon", "getChimeras");
236 /***************************************************************************************************************/
237 int Bellerophon::createSparseMatrix(int startSeq, int endSeq, SparseMatrix* sparse, vector<Sequence> s){
240 for(int i=startSeq; i<endSeq; i++){
242 for(int j=0;j<i;j++){
244 distCalculator->calcDist(s[i], s[j]);
245 float dist = distCalculator->getDist();
247 PCell temp(i, j, dist);
248 sparse->addCell(temp);
255 catch(exception& e) {
256 errorOut(e, "Bellerophon", "createSparseMatrix");
260 /***************************************************************************************************************/
261 void Bellerophon::generatePreferences(vector<SeqMap> left, vector<SeqMap> right, int mid){
265 SeqMap::iterator itR;
266 SeqMap::iterator itL;
269 for (int i = 0; i < pref.size(); i++) {
270 pref[i].score[1] = 0.0;
271 pref[i].closestLeft[1] = 100000.0;
272 pref[i].closestRight[1] = 100000.0;
273 pref[i].leftParent[1] = "";
274 pref[i].rightParent[1] = "";
277 for (int i = 0; i < left.size(); i++) {
279 SeqMap currentLeft = left[i]; //example i = 3; currentLeft is a map of 0 to the distance of sequence 3 to sequence 0,
280 // 1 to the distance of sequence 3 to sequence 1,
281 // 2 to the distance of sequence 3 to sequence 2.
282 SeqMap currentRight = right[i]; // same as left but with distances on the right side.
284 for (int j = 0; j < i; j++) {
286 itL = currentLeft.find(j);
287 itR = currentRight.find(j);
288 //cout << " i = " << i << " j = " << j << " distLeft = " << itL->second << endl;
289 //cout << " i = " << i << " j = " << j << " distright = " << itR->second << endl;
291 //if you can find this entry update the preferences
292 if ((itL != currentLeft.end()) && (itR != currentRight.end())) {
295 pref[i].score[1] += abs((itL->second - itR->second));
296 pref[j].score[1] += abs((itL->second - itR->second));
297 //cout << "left " << i << " " << j << " = " << itL->second << " right " << i << " " << j << " = " << itR->second << endl;
298 //cout << "abs = " << abs((itL->second - itR->second)) << endl;
299 //cout << i << " score = " << pref[i].score[1] << endl;
300 //cout << j << " score = " << pref[j].score[1] << endl;
302 pref[i].score[1] += abs((sqrt(itL->second) - sqrt(itR->second)));
303 pref[j].score[1] += abs((sqrt(itL->second) - sqrt(itR->second)));
304 //cout << "left " << i << " " << j << " = " << itL->second << " right " << i << " " << j << " = " << itR->second << endl;
305 //cout << "abs = " << abs((sqrt(itL->second) - sqrt(itR->second))) << endl;
306 //cout << i << " score = " << pref[i].score[1] << endl;
307 //cout << j << " score = " << pref[j].score[1] << endl;
309 //cout << "pref[" << i << "].closestLeft[1] = " << pref[i].closestLeft[1] << " parent = " << pref[i].leftParent[1] << endl;
310 //are you the closest left sequence
311 if (itL->second < pref[i].closestLeft[1]) {
313 pref[i].closestLeft[1] = itL->second;
314 pref[i].leftParent[1] = seqs[j]->getName();
315 //cout << "updating closest left to " << pref[i].leftParent[1] << endl;
317 //cout << "pref[" << j << "].closestLeft[1] = " << pref[j].closestLeft[1] << " parent = " << pref[j].leftParent[1] << endl;
318 if (itL->second < pref[j].closestLeft[1]) {
319 pref[j].closestLeft[1] = itL->second;
320 pref[j].leftParent[1] = seqs[i]->getName();
321 //cout << "updating closest left to " << pref[j].leftParent[1] << endl;
324 //are you the closest right sequence
325 if (itR->second < pref[i].closestRight[1]) {
326 pref[i].closestRight[1] = itR->second;
327 pref[i].rightParent[1] = seqs[j]->getName();
329 if (itR->second < pref[j].closestRight[1]) {
330 pref[j].closestRight[1] = itR->second;
331 pref[j].rightParent[1] = seqs[i]->getName();
343 for (int i = 0; i < pref.size(); i++) { dme += pref[i].score[1]; if (pref[i].score[1] == 0.0) { count0++; } }
345 float expectedPercent = 1 / (float) (pref.size() - count0);
346 //cout << endl << "dme = " << dme << endl;
347 //recalculate prefernences based on dme
348 for (int i = 0; i < pref.size(); i++) {
349 //cout << "unadjusted pref " << i << " = " << pref[i].score[1] << endl;
350 // gives the actual percentage of the dme this seq adds
351 pref[i].score[1] = pref[i].score[1] / dme;
353 //how much higher or lower is this than expected
354 pref[i].score[1] = pref[i].score[1] / expectedPercent;
356 //pref[i].score[1] = dme / (dme - 2 * pref[i].score[1]);
358 //so a non chimeric sequence would be around 1, and a chimeric would be signifigantly higher.
359 //cout << "adjusted pref " << i << " = " << pref[i].score[1] << endl;
362 //is this score bigger then the last score
363 for (int i = 0; i < pref.size(); i++) {
365 //update biggest score
366 if (pref[i].score[1] > pref[i].score[0]) {
367 pref[i].score[0] = pref[i].score[1];
368 pref[i].leftParent[0] = pref[i].leftParent[1];
369 pref[i].rightParent[0] = pref[i].rightParent[1];
370 pref[i].closestLeft[0] = pref[i].closestLeft[1];
371 pref[i].closestRight[0] = pref[i].closestRight[1];
372 pref[i].midpoint = mid;
378 catch(exception& e) {
379 errorOut(e, "Bellerophon", "generatePreferences");
383 /**************************************************************************************************/