5 * Created by Sarah Westcott on 2/9/09.
6 * Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
12 /**************************************************************************************************/
14 EstOutput Weighted::getValues(Tree* t, int p, string o) {
16 globaldata = GlobalData::getInstance();
18 data.clear(); //clear out old values
24 numGroups = globaldata->Groups.size();
26 if (m->control_pressed) { return data; }
28 //calculate number of comparisons i.e. with groups A,B,C = AB, AC, BC = 3;
29 vector< vector<string> > namesOfGroupCombos;
30 for (int i=0; i<numGroups; i++) {
31 for (int l = 0; l < i; l++) {
32 //initialize weighted scores
33 //WScore[globaldata->Groups[i]+globaldata->Groups[l]] = 0.0;
34 vector<string> groups; groups.push_back(globaldata->Groups[i]); groups.push_back(globaldata->Groups[l]);
35 namesOfGroupCombos.push_back(groups);
39 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
41 data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size());
43 int numPairs = namesOfGroupCombos.size();
45 int numPairsPerProcessor = numPairs / processors;
47 for (int i = 0; i < processors; i++) {
48 int startPos = i * numPairsPerProcessor;
49 if(i == processors - 1){
50 numPairsPerProcessor = numPairs - i * numPairsPerProcessor;
52 lines.push_back(linePair(startPos, numPairsPerProcessor));
55 data = createProcesses(t, namesOfGroupCombos);
60 data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size());
66 m->errorOut(e, "Weighted", "getValues");
70 /**************************************************************************************************/
72 EstOutput Weighted::createProcesses(Tree* t, vector< vector<string> > namesOfGroupCombos) {
74 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
76 vector<int> processIDS;
80 //loop through and create all the processes you want
81 while (process != processors) {
85 processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later
90 Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num);
92 m->mothurOut("Merging results."); m->mothurOutEndLine();
94 //pass numSeqs to parent
97 string tempFile = outputDir + toString(getpid()) + ".weighted.results.temp";
99 m->openOutputFile(tempFile, out);
101 out << Myresults.size() << endl;
102 for (int i = 0; i < Myresults.size(); i++) { out << Myresults[i] << '\t'; } out << endl;
107 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
108 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
113 results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num);
115 //force parent to wait until all the processes are done
116 for (int i=0;i<(processors-1);i++) {
117 int temp = processIDS[i];
121 if (m->control_pressed) { return results; }
123 //get data created by processes
124 for (int i=0;i<(processors-1);i++) {
126 string s = outputDir + toString(processIDS[i]) + ".weighted.results.temp";
127 m->openInputFile(s, in);
132 in >> num; m->gobble(in);
134 if (m->control_pressed) { break; }
137 for (int j = 0; j < num; j++) {
139 results.push_back(w);
147 m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine();
152 catch(exception& e) {
153 m->errorOut(e, "Weighted", "createProcesses");
157 /**************************************************************************************************/
158 EstOutput Weighted::driver(Tree* t, vector< vector<string> > namesOfGroupCombos, int start, int num) {
164 for (int h = start; h < (start+num); h++) {
166 if (m->control_pressed) { return results; }
168 //initialize weighted score
169 string groupA = namesOfGroupCombos[h][0];
170 string groupB = namesOfGroupCombos[h][1];
172 set<int> validBranches;
173 WScore[groupA+groupB] = 0.0;
174 D.push_back(0.0000); //initialize a spot in D for each combination
176 //adding the wieghted sums from group i
177 for (int j = 0; j < t->groupNodeInfo[groupA].size(); j++) { //the leaf nodes that have seqs from group i
178 map<string, int>::iterator it = t->tree[t->groupNodeInfo[groupA][j]].pcount.find(groupA);
179 int numSeqsInGroupI = it->second;
181 double sum = getLengthToRoot(t, t->groupNodeInfo[groupA][j], groupA, groupB);
182 double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groupA]);
184 D[count] += weightedSum;
187 //adding the wieghted sums from group l
188 for (int j = 0; j < t->groupNodeInfo[groupB].size(); j++) { //the leaf nodes that have seqs from group l
189 map<string, int>::iterator it = t->tree[t->groupNodeInfo[groupB][j]].pcount.find(groupB);
190 int numSeqsInGroupL = it->second;
192 double sum = getLengthToRoot(t, t->groupNodeInfo[groupB][j], groupA, groupB);
193 double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groupB]);
195 D[count] += weightedSum;
201 //calculate u for the group comb
203 for (int h = start; h < (start+num); h++) {
205 m->mothurOut("Processing combo: " + toString(h)); m->mothurOutEndLine();
207 string groupA = namesOfGroupCombos[h][0];
208 string groupB = namesOfGroupCombos[h][1];
210 //calculate u for the group comb
211 for(int i=0;i<t->getNumNodes();i++){
213 if (m->control_pressed) { return data; }
216 //int pcountSize = 0;
217 //does this node have descendants from groupA
218 it = t->tree[i].pcount.find(groupA);
219 //if it does u = # of its descendants with a certain group / total number in tree with a certain group
220 if (it != t->tree[i].pcount.end()) {
221 u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA];
225 //does this node have descendants from group l
226 it = t->tree[i].pcount.find(groupB);
227 //if it does subtract their percentage from u
228 if (it != t->tree[i].pcount.end()) {
229 u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB];
232 //if this is not the root then add it
233 if (rootForGrouping[namesOfGroupCombos[h]].count(i) == 0) {
234 u = abs(u * t->tree[i].getBranchLength());
235 WScore[(groupA+groupB)] += u;
241 /********************************************************/
242 //calculate weighted score for the group combination
245 for (int h = start; h < (start+num); h++) {
246 UN = (WScore[namesOfGroupCombos[h][0]+namesOfGroupCombos[h][1]] / D[count]);
248 if (isnan(UN) || isinf(UN)) { UN = 0; }
249 results.push_back(UN);
255 catch(exception& e) {
256 m->errorOut(e, "Weighted", "driver");
260 /**************************************************************************************************/
261 EstOutput Weighted::getValues(Tree* t, string groupA, string groupB) {
263 globaldata = GlobalData::getInstance();
265 data.clear(); //clear out old values
267 if (m->control_pressed) { return data; }
269 //initialize weighted score
270 WScore[(groupA+groupB)] = 0.0;
272 set<int> validBranches;
274 vector<string> groups; groups.push_back(groupA); groups.push_back(groupB);
276 //adding the wieghted sums from group i
277 for (int j = 0; j < t->groupNodeInfo[groups[0]].size(); j++) { //the leaf nodes that have seqs from group i
278 map<string, int>::iterator it = t->tree[t->groupNodeInfo[groups[0]][j]].pcount.find(groups[0]);
279 int numSeqsInGroupI = it->second;
281 double sum = getLengthToRoot(t, t->groupNodeInfo[groups[0]][j], groups[0], groups[1]);
282 double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groups[0]]);
287 //adding the wieghted sums from group l
288 for (int j = 0; j < t->groupNodeInfo[groups[1]].size(); j++) { //the leaf nodes that have seqs from group l
289 map<string, int>::iterator it = t->tree[t->groupNodeInfo[groups[1]][j]].pcount.find(groups[1]);
290 int numSeqsInGroupL = it->second;
292 double sum = getLengthToRoot(t, t->groupNodeInfo[groups[1]][j], groups[0], groups[1]);
293 double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groups[1]]);
298 //calculate u for the group comb
299 for(int i=0;i<t->getNumNodes();i++){
301 if (m->control_pressed) { return data; }
304 //int pcountSize = 0;
305 //does this node have descendants from groupA
306 it = t->tree[i].pcount.find(groupA);
307 //if it does u = # of its descendants with a certain group / total number in tree with a certain group
308 if (it != t->tree[i].pcount.end()) {
309 u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA];
313 //does this node have descendants from group l
314 it = t->tree[i].pcount.find(groupB);
315 //if it does subtract their percentage from u
316 if (it != t->tree[i].pcount.end()) {
317 u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB];
320 //if this is not the root then add it
321 if (rootForGrouping[groups].count(i) == 0) {
322 u = abs(u * t->tree[i].getBranchLength());
323 WScore[(groupA+groupB)] += u;
326 /********************************************************/
328 //calculate weighted score for the group combination
330 UN = (WScore[(groupA+groupB)] / D);
332 if (isnan(UN) || isinf(UN)) { UN = 0; }
337 catch(exception& e) {
338 m->errorOut(e, "Weighted", "getValues");
342 /**************************************************************************************************/
343 double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) {
350 if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); }
351 double tempTotal = 0.0;
352 index = t->tree[index].getParent();
354 vector<string> grouping; grouping.push_back(groupA); grouping.push_back(groupB);
356 rootForGrouping[grouping].insert(index);
358 //while you aren't at root
359 while(t->tree[index].getParent() != -1){
361 if (m->control_pressed) { return sum; }
363 //am I the root for this grouping? if so I want to stop "early"
364 //does my sibling have descendants from the users groups?
365 int parent = t->tree[index].getParent();
366 int lc = t->tree[parent].getLChild();
367 int rc = t->tree[parent].getRChild();
370 if (lc == index) { sib = rc; }
372 map<string, int>::iterator itGroup;
374 itGroup = t->tree[sib].pcount.find(groupA);
375 if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; }
376 itGroup = t->tree[sib].pcount.find(groupB);
377 if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; }
379 //if yes, I am not the root so add me
380 if (pcountSize != 0) {
381 if (t->tree[index].getBranchLength() != -1) {
382 sum += abs(t->tree[index].getBranchLength()) + tempTotal;
388 rootForGrouping[grouping].clear();
389 rootForGrouping[grouping].insert(parent);
390 }else { //if no, I may be the root so add my br to tempTotal until I am proven innocent
391 if (t->tree[index].getBranchLength() != -1) {
392 tempTotal += abs(t->tree[index].getBranchLength());
399 //get all nodes above the root to add so we don't add their u values above
400 index = *(rootForGrouping[grouping].begin());
401 while(t->tree[index].getParent() != -1){
402 int parent = t->tree[index].getParent();
403 rootForGrouping[grouping].insert(parent);
409 catch(exception& e) {
410 m->errorOut(e, "Weighted", "getBranchLengthSums");
414 /**************************************************************************************************/