unifracweightedcommand.cpp

   1 /*
   2  *  unifracweightedcommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Sarah Westcott on 2/9/09.
   6  *  Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
   7  *
   8  */
   9
  10 #include "unifracweightedcommand.h"
  11
  12 /***********************************************************/
  13 UnifracWeightedCommand::UnifracWeightedCommand(string option) {
  14         try {
  15                 globaldata = GlobalData::getInstance();
  16                 abort = false;
  17                 Groups.clear();
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string Array[] =  {"groups","iters","distance","random","processors","outputdir","inputdir"};
  25                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  26
  27                         OptionParser parser(option);
  28                         map<string,string> parameters=parser.getParameters();
  29
  30                         ValidParameters validParameter;
  31
  32                         //check to make sure all parameters are valid for command
  33                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  34                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  35                         }
  36
  37                         if (globaldata->gTree.size() == 0) {//no trees were read
  38                                 m->mothurOut("You must execute the read.tree command, before you may execute the unifrac.weighted command."); m->mothurOutEndLine(); abort = true;  }
  39
  40                         //if the user changes the output directory command factory will send this info to us in the output parameter
  41                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
  42                                 outputDir = "";
  43                                 outputDir += m->hasPath(globaldata->inputFileName); //if user entered a file with a path then preserve it
  44                         }
  45
  46                         //check for optional parameter and set defaults
  47                         // ...at some point should added some additional type checking...
  48                         groups = validParameter.validFile(parameters, "groups", false);
  49                         if (groups == "not found") { groups = ""; }
  50                         else {
  51                                 m->splitAtDash(groups, Groups);
  52                                 globaldata->Groups = Groups;
  53                         }
  54
  55                         itersString = validParameter.validFile(parameters, "iters", false);                     if (itersString == "not found") { itersString = "1000"; }
  56                         convert(itersString, iters);
  57
  58                         string temp = validParameter.validFile(parameters, "distance", false);                  if (temp == "not found") { temp = "false"; }
  59                         phylip = m->isTrue(temp);
  60
  61                         temp = validParameter.validFile(parameters, "random", false);                                   if (temp == "not found") { temp = "F"; }
  62                         random = m->isTrue(temp);
  63
  64                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
  65                         convert(temp, processors);
  66
  67                         if (!random) {  iters = 0;  } //turn off random calcs
  68
  69
  70                         if (abort == false) {
  71                                 T = globaldata->gTree;
  72                                 tmap = globaldata->gTreemap;
  73                                 sumFile = outputDir + m->getSimpleName(globaldata->getTreeFile()) + ".wsummary";
  74                                 m->openOutputFile(sumFile, outSum);
  75                                 outputNames.push_back(sumFile);
  76
  77                                 util = new SharedUtil();
  78                                 string s; //to make work with setgroups
  79                                 util->setGroups(globaldata->Groups, tmap->namesOfGroups, s, numGroups, "weighted");     //sets the groups the user wants to analyze
  80                                 util->getCombos(groupComb, globaldata->Groups, numComp);
  81
  82                                 weighted = new Weighted(tmap);
  83
  84                         }
  85                 }
  86
  87
  88         }
  89         catch(exception& e) {
  90                 m->errorOut(e, "UnifracWeightedCommand", "UnifracWeightedCommand");
  91                 exit(1);
  92         }
  93 }
  94 //**********************************************************************************************************************
  95
  96 void UnifracWeightedCommand::help(){
  97         try {
  98                 m->mothurOut("The unifrac.weighted command can only be executed after a successful read.tree command.\n");
  99                 m->mothurOut("The unifrac.weighted command parameters are groups, iters, distance and random.  No parameters are required.\n");
 100                 m->mothurOut("The groups parameter allows you to specify which of the groups in your groupfile you would like analyzed.  You must enter at least 2 valid groups.\n");
 101                 m->mothurOut("The group names are separated by dashes.  The iters parameter allows you to specify how many random trees you would like compared to your tree.\n");
 102                 m->mothurOut("The distance parameter allows you to create a distance file from the results. The default is false.\n");
 103                 m->mothurOut("The random parameter allows you to shut off the comparison to random trees. The default is false, meaning don't compare your trees with randomly generated trees.\n");
 104                 m->mothurOut("The unifrac.weighted command should be in the following format: unifrac.weighted(groups=yourGroups, iters=yourIters).\n");
 105                 m->mothurOut("Example unifrac.weighted(groups=A-B-C, iters=500).\n");
 106                 m->mothurOut("The default value for groups is all the groups in your groupfile, and iters is 1000.\n");
 107                 m->mothurOut("The unifrac.weighted command output two files: .weighted and .wsummary their descriptions are in the manual.\n");
 108                 m->mothurOut("Note: No spaces between parameter labels (i.e. groups), '=' and parameters (i.e.yourGroups).\n\n");
 109         }
 110         catch(exception& e) {
 111                 m->errorOut(e, "UnifracWeightedCommand", "help");
 112                 exit(1);
 113         }
 114 }
 115
 116 /***********************************************************/
 117 int UnifracWeightedCommand::execute() {
 118         try {
 119
 120                 if (abort == true) { return 0; }
 121
 122                 int start = time(NULL);
 123
 124                 //get weighted for users tree
 125                 userData.resize(numComp,0);  //data[0] = weightedscore AB, data[1] = weightedscore AC...
 126                 randomData.resize(numComp,0); //data[0] = weightedscore AB, data[1] = weightedscore AC...
 127
 128                 //get weighted scores for users trees
 129                 for (int i = 0; i < T.size(); i++) {
 130
 131                         if (m->control_pressed) { outSum.close(); for (int i = 0; i < outputNames.size(); i++) {        remove(outputNames[i].c_str());  } return 0; }
 132
 133                         counter = 0;
 134                         rScores.resize(numComp);  //data[0] = weightedscore AB, data[1] = weightedscore AC...
 135                         uScores.resize(numComp);  //data[0] = weightedscore AB, data[1] = weightedscore AC...
 136
 137                         if (random) {
 138                                 output = new ColumnFile(outputDir + m->getSimpleName(globaldata->getTreeFile())  + toString(i+1) + ".weighted", itersString);
 139                                 outputNames.push_back(outputDir + m->getSimpleName(globaldata->getTreeFile())  + toString(i+1) + ".weighted");
 140                         }
 141
 142                         userData = weighted->getValues(T[i], processors, outputDir);  //userData[0] = weightedscore
 143
 144                         if (m->control_pressed) { if (random) { delete output; } outSum.close(); for (int i = 0; i < outputNames.size(); i++) { remove(outputNames[i].c_str());  } return 0; }
 145
 146                         //save users score
 147                         for (int s=0; s<numComp; s++) {
 148                                 //add users score to vector of user scores
 149                                 uScores[s].push_back(userData[s]);
 150
 151                                 //save users tree score for summary file
 152                                 utreeScores.push_back(userData[s]);
 153                         }
 154
 155                         if (random) {
 156                                 vector<double> sums = weighted->getBranchLengthSums(T[i]);
 157
 158                                 //calculate number of comparisons i.e. with groups A,B,C = AB, AC, BC = 3;
 159                                 vector< vector<string> > namesOfGroupCombos;
 160                                 for (int a=0; a<numGroups; a++) {
 161                                         for (int l = 0; l < a; l++) {
 162                                                 vector<string> groups; groups.push_back(globaldata->Groups[a]); groups.push_back(globaldata->Groups[l]);
 163                                                 namesOfGroupCombos.push_back(groups);
 164                                         }
 165                                 }
 166
 167                                 lines.clear();
 168
 169                                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 170                                         if(processors != 1){
 171                                                 int numPairs = namesOfGroupCombos.size();
 172                                                 int numPairsPerProcessor = numPairs / processors;
 173
 174                                                 for (int i = 0; i < processors; i++) {
 175                                                         int startPos = i * numPairsPerProcessor;
 176                                                         if(i == processors - 1){
 177                                                                 numPairsPerProcessor = numPairs - i * numPairsPerProcessor;
 178                                                         }
 179                                                         lines.push_back(linePair(startPos, numPairsPerProcessor));
 180                                                 }
 181                                         }
 182                                 #endif
 183
 184
 185                                 //get scores for random trees
 186                                 for (int j = 0; j < iters; j++) {
 187
 188                                         #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 189                                                 if(processors == 1){
 190                                                         driver(T[i],  namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums, rScores);
 191                                                 }else{
 192                                                         createProcesses(T[i],  namesOfGroupCombos, sums, rScores);
 193                                                 }
 194                                         #else
 195                                                 driver(T[i], namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums, rScores);
 196                                         #endif
 197
 198                                         if (m->control_pressed) { delete output; outSum.close(); for (int i = 0; i < outputNames.size(); i++) { remove(outputNames[i].c_str());  } return 0; }
 199
 200                                         //report progress
 201                                         m->mothurOut("Iter: " + toString(j+1)); m->mothurOutEndLine();
 202                                 }
 203                                 lines.clear();
 204
 205                                 //find the signifigance of the score for summary file
 206                                 for (int f = 0; f < numComp; f++) {
 207                                         //sort random scores
 208                                         sort(rScores[f].begin(), rScores[f].end());
 209
 210                                         //the index of the score higher than yours is returned
 211                                         //so if you have 1000 random trees the index returned is 100
 212                                         //then there are 900 trees with a score greater then you.
 213                                         //giving you a signifigance of 0.900
 214                                         int index = findIndex(userData[f], f);    if (index == -1) { m->mothurOut("error in UnifracWeightedCommand"); m->mothurOutEndLine(); exit(1); } //error code
 215
 216                                         //the signifigance is the number of trees with the users score or higher
 217                                         WScoreSig.push_back((iters-index)/(float)iters);
 218                                 }
 219
 220                                 //out << "Tree# " << i << endl;
 221                                 calculateFreqsCumuls();
 222                                 printWeightedFile();
 223
 224                                 delete output;
 225
 226                         }
 227
 228                         //clear data
 229                         rScores.clear();
 230                         uScores.clear();
 231                         validScores.clear();
 232                 }
 233
 234
 235                 if (m->control_pressed) { outSum.close(); for (int i = 0; i < outputNames.size(); i++) {        remove(outputNames[i].c_str());  } return 0;  }
 236
 237                 printWSummaryFile();
 238
 239                 if (phylip) {   createPhylipFile();             }
 240
 241                 //clear out users groups
 242                 globaldata->Groups.clear();
 243
 244
 245                 if (m->control_pressed) {
 246                         for (int i = 0; i < outputNames.size(); i++) {  remove(outputNames[i].c_str());  }
 247                         return 0;
 248                 }
 249
 250                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to run unifrac.weighted."); m->mothurOutEndLine();
 251
 252                 m->mothurOutEndLine();
 253                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 254                 for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 255                 m->mothurOutEndLine();
 256
 257                 return 0;
 258
 259         }
 260         catch(exception& e) {
 261                 m->errorOut(e, "UnifracWeightedCommand", "execute");
 262                 exit(1);
 263         }
 264 }
 265 /**************************************************************************************************/
 266
 267 int UnifracWeightedCommand::createProcesses(Tree* t, vector< vector<string> > namesOfGroupCombos, vector<double>& sums, vector< vector<double> >& scores) {
 268         try {
 269 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 270                 int process = 1;
 271                 int num = 0;
 272                 vector<int> processIDS;
 273
 274                 EstOutput results;
 275
 276                 //loop through and create all the processes you want
 277                 while (process != processors) {
 278                         int pid = fork();
 279
 280                         if (pid > 0) {
 281                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 282                                 process++;
 283                         }else if (pid == 0){
 284                                 driver(t, namesOfGroupCombos, lines[process].start, lines[process].num, sums, scores);
 285
 286                                 //pass numSeqs to parent
 287                                 ofstream out;
 288                                 string tempFile = outputDir + toString(getpid()) + ".weightedcommand.results.temp";
 289                                 m->openOutputFile(tempFile, out);
 290                                 for (int i = lines[process].start; i < (lines[process].start + lines[process].num); i++) { out << scores[i][(scores[i].size()-1)] << '\t';  } out << endl;
 291                                 out.close();
 292
 293                                 exit(0);
 294                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 295                 }
 296
 297                 driver(t, namesOfGroupCombos, lines[0].start, lines[0].num, sums, scores);
 298
 299                 //force parent to wait until all the processes are done
 300                 for (int i=0;i<(processors-1);i++) {
 301                         int temp = processIDS[i];
 302                         wait(&temp);
 303                 }
 304
 305                 //get data created by processes
 306                 for (int i=0;i<(processors-1);i++) {
 307
 308                         ifstream in;
 309                         string s = outputDir + toString(processIDS[i]) + ".weightedcommand.results.temp";
 310                         m->openInputFile(s, in);
 311
 312                         double tempScore;
 313                         for (int j = lines[(i+1)].start; j < (lines[(i+1)].start + lines[(i+1)].num); j++) { in >> tempScore; scores[j].push_back(tempScore); }
 314                         in.close();
 315                         remove(s.c_str());
 316                 }
 317
 318                 return 0;
 319 #endif
 320         }
 321         catch(exception& e) {
 322                 m->errorOut(e, "UnifracWeightedCommand", "createProcesses");
 323                 exit(1);
 324         }
 325 }
 326
 327 /**************************************************************************************************/
 328 int UnifracWeightedCommand::driver(Tree* t, vector< vector<string> > namesOfGroupCombos, int start, int num, vector<double>& sums, vector< vector<double> >& scores) {
 329  try {
 330                 Tree* randT = new Tree();
 331
 332                 for (int h = start; h < (start+num); h++) {
 333
 334                         if (m->control_pressed) { return 0; }
 335
 336                         //initialize weighted score
 337                         string groupA = namesOfGroupCombos[h][0];
 338                         string groupB = namesOfGroupCombos[h][1];
 339
 340                         //copy T[i]'s info.
 341                         randT->getCopy(t);
 342
 343                         //create a random tree with same topology as T[i], but different labels
 344                         randT->assembleRandomUnifracTree(groupA, groupB);
 345
 346                         if (m->control_pressed) { delete randT;  return 0;  }
 347
 348                         //get wscore of random tree
 349                         EstOutput randomData = weighted->getValues(randT, groupA, groupB, sums);
 350
 351                         if (m->control_pressed) { delete randT;  return 0;  }
 352
 353                         //save scores
 354                         scores[h].push_back(randomData[0]);
 355                 }
 356
 357                 delete randT;
 358
 359                 return 0;
 360
 361         }
 362         catch(exception& e) {
 363                 m->errorOut(e, "UnifracWeightedCommand", "driver");
 364                 exit(1);
 365         }
 366 }
 367 /***********************************************************/
 368 void UnifracWeightedCommand::printWeightedFile() {
 369         try {
 370                 vector<double> data;
 371                 vector<string> tags;
 372                 tags.push_back("Score"); tags.push_back("RandFreq"); tags.push_back("RandCumul");
 373
 374                 for(int a = 0; a < numComp; a++) {
 375                         output->initFile(groupComb[a], tags);
 376                         //print each line
 377                         for (map<float,float>::iterator it = validScores.begin(); it != validScores.end(); it++) {
 378                                 data.push_back(it->first);  data.push_back(rScoreFreq[a][it->first]); data.push_back(rCumul[a][it->first]);
 379                                 output->output(data);
 380                                 data.clear();
 381                         }
 382                         output->resetFile();
 383                 }
 384         }
 385         catch(exception& e) {
 386                 m->errorOut(e, "UnifracWeightedCommand", "printWeightedFile");
 387                 exit(1);
 388         }
 389 }
 390
 391
 392 /***********************************************************/
 393 void UnifracWeightedCommand::printWSummaryFile() {
 394         try {
 395                 //column headers
 396                 outSum << "Tree#" << '\t' << "Groups" << '\t' << "WScore" << '\t' << "WSig" <<  endl;
 397                 m->mothurOut("Tree#\tGroups\tWScore\tWSig"); m->mothurOutEndLine();
 398
 399                 //format output
 400                 outSum.setf(ios::fixed, ios::floatfield); outSum.setf(ios::showpoint);
 401
 402                 //print each line
 403                 int count = 0;
 404                 for (int i = 0; i < T.size(); i++) {
 405                         for (int j = 0; j < numComp; j++) {
 406                                 if (random) {
 407                                         if (WScoreSig[count] > (1/(float)iters)) {
 408                                                 outSum << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << setprecision(itersString.length()) << WScoreSig[count] << endl;
 409                                                 cout << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << setprecision(itersString.length()) << WScoreSig[count] << endl;
 410                                                 m->mothurOutJustToLog(toString(i+1) +"\t" + groupComb[j] +"\t" + toString(utreeScores[count]) +"\t" +  toString(WScoreSig[count]) + "\n");
 411                                         }else{
 412                                                 outSum << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << setprecision(itersString.length()) << "<" << (1/float(iters)) << endl;
 413                                                 cout << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << setprecision(itersString.length()) << "<" << (1/float(iters)) << endl;
 414                                                 m->mothurOutJustToLog(toString(i+1) +"\t" + groupComb[j] +"\t" + toString(utreeScores[count]) +"\t<" +  toString((1/float(iters))) + "\n");
 415                                         }
 416                                 }else{
 417                                         outSum << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << "0.00" << endl;
 418                                         cout << setprecision(6) << i+1 << '\t' << groupComb[j] << '\t' << utreeScores[count] << '\t' << "0.00" << endl;
 419                                         m->mothurOutJustToLog(toString(i+1) +"\t" + groupComb[j] +"\t" + toString(utreeScores[count]) +"\t0.00\n");
 420                                 }
 421                                 count++;
 422                         }
 423                 }
 424                 outSum.close();
 425         }
 426         catch(exception& e) {
 427                 m->errorOut(e, "UnifracWeightedCommand", "printWSummaryFile");
 428                 exit(1);
 429         }
 430 }
 431 /***********************************************************/
 432 void UnifracWeightedCommand::createPhylipFile() {
 433         try {
 434                 int count = 0;
 435                 //for each tree
 436                 for (int i = 0; i < T.size(); i++) {
 437
 438                         string phylipFileName = outputDir + m->getSimpleName(globaldata->getTreeFile())  + toString(i+1) + ".weighted.dist";
 439                         outputNames.push_back(phylipFileName);
 440                         ofstream out;
 441                         m->openOutputFile(phylipFileName, out);
 442
 443                         //output numSeqs
 444                         out << globaldata->Groups.size() << endl;
 445
 446                         //make matrix with scores in it
 447                         vector< vector<float> > dists;  dists.resize(globaldata->Groups.size());
 448                         for (int i = 0; i < globaldata->Groups.size(); i++) {
 449                                 dists[i].resize(globaldata->Groups.size(), 0.0);
 450                         }
 451
 452                         //flip it so you can print it
 453                         for (int r=0; r<globaldata->Groups.size(); r++) {
 454                                 for (int l = r+1; l < globaldata->Groups.size(); l++) {
 455                                         dists[r][l] = utreeScores[count];
 456                                         dists[l][r] = utreeScores[count];
 457                                         count++;
 458                                 }
 459                         }
 460
 461                         //output to file
 462                         for (int r=0; r<globaldata->Groups.size(); r++) {
 463                                 //output name
 464                                 string name = globaldata->Groups[r];
 465                                 if (name.length() < 10) { //pad with spaces to make compatible
 466                                         while (name.length() < 10) {  name += " ";  }
 467                                 }
 468                                 out << name << '\t';
 469
 470                                 //output distances
 471                                 for (int l = 0; l < r; l++) {   out  << dists[r][l] << '\t';  }
 472                                 out << endl;
 473                         }
 474                         out.close();
 475                 }
 476         }
 477         catch(exception& e) {
 478                 m->errorOut(e, "UnifracWeightedCommand", "createPhylipFile");
 479                 exit(1);
 480         }
 481 }
 482 /***********************************************************/
 483 int UnifracWeightedCommand::findIndex(float score, int index) {
 484         try{
 485                 for (int i = 0; i < rScores[index].size(); i++) {
 486                         if (rScores[index][i] >= score) {       return i;       }
 487                 }
 488                 return rScores[index].size();
 489         }
 490         catch(exception& e) {
 491                 m->errorOut(e, "UnifracWeightedCommand", "findIndex");
 492                 exit(1);
 493         }
 494 }
 495
 496 /***********************************************************/
 497
 498 void UnifracWeightedCommand::calculateFreqsCumuls() {
 499         try {
 500                 //clear out old tree values
 501                 rScoreFreq.clear();
 502                 rScoreFreq.resize(numComp);
 503                 rCumul.clear();
 504                 rCumul.resize(numComp);
 505                 validScores.clear();
 506
 507                 //calculate frequency
 508                 for (int f = 0; f < numComp; f++) {
 509                         for (int i = 0; i < rScores[f].size(); i++) { //looks like 0,0,1,1,1,2,4,7...  you want to make a map that say rScoreFreq[0] = 2, rScoreFreq[1] = 3...
 510                                 validScores[rScores[f][i]] = rScores[f][i];
 511                                 map<float,float>::iterator it = rScoreFreq[f].find(rScores[f][i]);
 512                                 if (it != rScoreFreq[f].end()) {
 513                                         rScoreFreq[f][rScores[f][i]]++;
 514                                 }else{
 515                                         rScoreFreq[f][rScores[f][i]] = 1;
 516                                 }
 517                         }
 518                 }
 519
 520                 //calculate rcumul
 521                 for(int a = 0; a < numComp; a++) {
 522                         float rcumul = 1.0000;
 523                         //this loop fills the cumulative maps and put 0.0000 in the score freq map to make it easier to print.
 524                         for (map<float,float>::iterator it = validScores.begin(); it != validScores.end(); it++) {
 525                                 //make rscoreFreq map and rCumul
 526                                 map<float,float>::iterator it2 = rScoreFreq[a].find(it->first);
 527                                 rCumul[a][it->first] = rcumul;
 528                                 //get percentage of random trees with that info
 529                                 if (it2 != rScoreFreq[a].end()) {  rScoreFreq[a][it->first] /= iters; rcumul-= it2->second;  }
 530                                 else { rScoreFreq[a][it->first] = 0.0000; } //no random trees with that score
 531                         }
 532                 }
 533
 534         }
 535         catch(exception& e) {
 536                 m->errorOut(e, "UnifracWeightedCommand", "calculateFreqsCums");
 537                 exit(1);
 538         }
 539 }
 540
 541 /***********************************************************/
 542
 543
 544
 545
 546