phylotree.cpp

   1 /*
   2  *  doTaxonomy.cpp
   3  *
   4  *
   5  *  Created by Pat Schloss on 6/17/09.
   6  *  Copyright 2009 Patrick D. Schloss. All rights reserved.
   7  *
   8  */
   9
  10 #include "phylotree.h"
  11
  12 /**************************************************************************************************/
  13
  14 PhyloTree::PhyloTree(){
  15         try {
  16                 m = MothurOut::getInstance();
  17                 numNodes = 1;
  18                 numSeqs = 0;
  19                 tree.push_back(TaxNode("Root"));
  20                 tree[0].heirarchyID = "0";
  21                 maxLevel = 0;
  22                 calcTotals = true;
  23         }
  24         catch(exception& e) {
  25                 m->errorOut(e, "PhyloTree", "PhyloTree");
  26                 exit(1);
  27         }
  28 }
  29 /**************************************************************************************************/
  30
  31 PhyloTree::PhyloTree(ifstream& in, string filename){
  32         try {
  33                 m = MothurOut::getInstance();
  34                 calcTotals = false;
  35                 numNodes = 0;
  36                 numSeqs = 0;
  37
  38                 #ifdef USE_MPI
  39                         MPI_File inMPI;
  40                         MPI_Offset size;
  41                         MPI_Status status;
  42
  43                         char inFileName[1024];
  44                         strcpy(inFileName, filename.c_str());
  45
  46                         MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
  47                         MPI_File_get_size(inMPI, &size);
  48
  49                         char* buffer = new char[size];
  50                         MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
  51
  52                         string tempBuf = buffer;
  53                         if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
  54                         istringstream iss (tempBuf,istringstream::in);
  55                         delete buffer;
  56
  57                         //read version
  58                         m->getline(iss); m->gobble(iss);
  59
  60                         iss >> numNodes; m->gobble(iss);
  61
  62                         tree.resize(numNodes);
  63
  64                         for (int i = 0; i < tree.size(); i++) {
  65                                 iss >> tree[i].name >> tree[i].level >> tree[i].parent; m->gobble(iss);
  66                         }
  67
  68                         //read genus nodes
  69                         int numGenus = 0;
  70                         iss >> numGenus; m->gobble(iss);
  71
  72                         int gnode, gsize;
  73                         totals.clear();
  74                         for (int i = 0; i < numGenus; i++) {
  75                                 iss >> gnode >> gsize; m->gobble(iss);
  76
  77                                 uniqueTaxonomies[gnode] = gnode;
  78                                 totals.push_back(gsize);
  79                         }
  80
  81                         MPI_File_close(&inMPI);
  82
  83                 #else
  84                         //read version
  85                         string line = m->getline(in); m->gobble(in);
  86
  87                         in >> numNodes; m->gobble(in);
  88
  89                         tree.resize(numNodes);
  90
  91                         for (int i = 0; i < tree.size(); i++) {
  92                                 in >> tree[i].name >> tree[i].level >> tree[i].parent; m->gobble(in);
  93                         }
  94
  95                         //read genus nodes
  96                         int numGenus = 0;
  97                         in >> numGenus; m->gobble(in);
  98
  99                         int gnode, gsize;
 100                         totals.clear();
 101                         for (int i = 0; i < numGenus; i++) {
 102                                 in >> gnode >> gsize; m->gobble(in);
 103
 104                                 uniqueTaxonomies[gnode] = gnode;
 105                                 totals.push_back(gsize);
 106                         }
 107
 108                         in.close();
 109
 110                 #endif
 111
 112         }
 113         catch(exception& e) {
 114                 m->errorOut(e, "PhyloTree", "PhyloTree");
 115                 exit(1);
 116         }
 117 }
 118 /**************************************************************************************************/
 119
 120 PhyloTree::PhyloTree(string tfile){
 121         try {
 122                 m = MothurOut::getInstance();
 123                 numNodes = 1;
 124                 numSeqs = 0;
 125                 tree.push_back(TaxNode("Root"));
 126                 tree[0].heirarchyID = "0";
 127                 maxLevel = 0;
 128                 calcTotals = true;
 129                 string name, tax;
 130
 131
 132                 #ifdef USE_MPI
 133                         int pid, num, processors;
 134                         vector<unsigned long int> positions;
 135
 136                         MPI_Status status;
 137                         MPI_File inMPI;
 138                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 139                         MPI_Comm_size(MPI_COMM_WORLD, &processors);
 140
 141                         char inFileName[1024];
 142                         strcpy(inFileName, tfile.c_str());
 143
 144                         MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 145
 146                         if (pid == 0) {
 147                                 positions = m->setFilePosEachLine(tfile, num);
 148
 149                                 //send file positions to all processes
 150                                 for(int i = 1; i < processors; i++) {
 151                                         MPI_Send(&num, 1, MPI_INT, i, 2001, MPI_COMM_WORLD);
 152                                         MPI_Send(&positions[0], (num+1), MPI_LONG, i, 2001, MPI_COMM_WORLD);
 153                                 }
 154                         }else{
 155                                 MPI_Recv(&num, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 156                                 positions.resize(num+1);
 157                                 MPI_Recv(&positions[0], (num+1), MPI_LONG, 0, 2001, MPI_COMM_WORLD, &status);
 158                         }
 159
 160                         //read file
 161                         for(int i=0;i<num;i++){
 162                                 //read next sequence
 163                                 int length = positions[i+1] - positions[i];
 164                                 char* buf4 = new char[length];
 165
 166                                 MPI_File_read_at(inMPI, positions[i], buf4, length, MPI_CHAR, &status);
 167
 168                                 string tempBuf = buf4;
 169                                 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
 170                                 delete buf4;
 171
 172                                 istringstream iss (tempBuf,istringstream::in);
 173                                 iss >> name >> tax;
 174                                 addSeqToTree(name, tax);
 175                         }
 176
 177                         MPI_File_close(&inMPI);
 178                         MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 179
 180                 #else
 181                         ifstream in;
 182                         m->openInputFile(tfile, in);
 183
 184                         //read in users taxonomy file and add sequences to tree
 185                         while(!in.eof()){
 186                                 in >> name >> tax; m->gobble(in);
 187
 188                                 addSeqToTree(name, tax);
 189                         }
 190                         in.close();
 191                 #endif
 192
 193                 assignHeirarchyIDs(0);
 194
 195                 //create file for summary if needed
 196                 setUp(tfile);
 197         }
 198         catch(exception& e) {
 199                 m->errorOut(e, "PhyloTree", "PhyloTree");
 200                 exit(1);
 201         }
 202 }
 203
 204 /**************************************************************************************************/
 205
 206 string PhyloTree::getNextTaxon(string& heirarchy, string seqname){
 207         try {
 208                 string currentLevel = "";
 209                 if(heirarchy != ""){
 210                         int pos = heirarchy.find_first_of(';');
 211
 212                         if (pos == -1) { //you can't find another ;
 213                                 currentLevel = heirarchy;
 214                                 heirarchy = "";
 215                                 m->mothurOut(seqname + " is missing a ;, please check for other errors."); m->mothurOutEndLine();
 216                         }else{
 217                                 currentLevel=heirarchy.substr(0,pos);
 218                                 if (pos != (heirarchy.length()-1)) {  heirarchy=heirarchy.substr(pos+1);  }
 219                                 else { heirarchy = ""; }
 220                         }
 221
 222                 }
 223                 return currentLevel;
 224         }
 225         catch(exception& e) {
 226                 m->errorOut(e, "PhyloTree", "getNextTaxon");
 227                 exit(1);
 228         }
 229 }
 230
 231 /**************************************************************************************************/
 232
 233 int PhyloTree::addSeqToTree(string seqName, string seqTaxonomy){
 234         try {
 235
 236                 numSeqs++;
 237
 238                 map<string, int>::iterator childPointer;
 239
 240                 int currentNode = 0;
 241                 int level = 1;
 242
 243                 tree[0].accessions.push_back(seqName);
 244                 string taxon;// = getNextTaxon(seqTaxonomy);
 245
 246                 while(seqTaxonomy != ""){
 247
 248                         level++;
 249
 250                         if (m->control_pressed) { return 0; }
 251
 252                         //somehow the parent is getting one too many accnos
 253                         //use print to reassign the taxa id
 254                         taxon = getNextTaxon(seqTaxonomy, seqName);
 255
 256                         if (taxon == "") {  m->mothurOut(seqName + " has an error in the taxonomy.  This may be due to a ;;"); m->mothurOutEndLine(); if (currentNode != 0) {  uniqueTaxonomies[currentNode] = currentNode; } break;  }
 257
 258                         childPointer = tree[currentNode].children.find(taxon);
 259
 260                         if(childPointer != tree[currentNode].children.end()){   //if the node already exists, move on
 261                                 currentNode = childPointer->second;
 262                                 tree[currentNode].accessions.push_back(seqName);
 263                                 name2Taxonomy[seqName] = currentNode;
 264                         }
 265                         else{                                                                                   //otherwise, create it
 266                                 tree.push_back(TaxNode(taxon));
 267                                 numNodes++;
 268                                 tree[currentNode].children[taxon] = numNodes-1;
 269                                 tree[numNodes-1].parent = currentNode;
 270
 271                                 currentNode = tree[currentNode].children[taxon];
 272                                 tree[currentNode].accessions.push_back(seqName);
 273                                 name2Taxonomy[seqName] = currentNode;
 274                         }
 275
 276                         if (seqTaxonomy == "") {   uniqueTaxonomies[currentNode] = currentNode; }
 277                 }
 278         }
 279         catch(exception& e) {
 280                 m->errorOut(e, "PhyloTree", "addSeqToTree");
 281                 exit(1);
 282         }
 283 }
 284 /**************************************************************************************************/
 285 vector<int> PhyloTree::getGenusNodes()  {
 286         try {
 287                 genusIndex.clear();
 288                 //generate genusIndexes
 289                 map<int, int>::iterator it2;
 290                 for (it2=uniqueTaxonomies.begin(); it2!=uniqueTaxonomies.end(); it2++) {  genusIndex.push_back(it2->first);     }
 291
 292                 return genusIndex;
 293         }
 294         catch(exception& e) {
 295                 m->errorOut(e, "PhyloTree", "getGenusNodes");
 296                 exit(1);
 297         }
 298 }
 299 /**************************************************************************************************/
 300 vector<int> PhyloTree::getGenusTotals() {
 301         try {
 302
 303                 if (calcTotals) {
 304                         totals.clear();
 305                         //reset counts because we are on a new word
 306                         for (int j = 0; j < genusIndex.size(); j++) {
 307                                 totals.push_back(tree[genusIndex[j]].accessions.size());
 308                         }
 309                         return totals;
 310                 }else{
 311                         return totals;
 312                 }
 313
 314         }
 315         catch(exception& e) {
 316                 m->errorOut(e, "PhyloTree", "getGenusNodes");
 317                 exit(1);
 318         }
 319 }
 320 /**************************************************************************************************/
 321
 322 void PhyloTree::assignHeirarchyIDs(int index){
 323         try {
 324                 map<string,int>::iterator it;
 325                 int counter = 1;
 326
 327                 for(it=tree[index].children.begin();it!=tree[index].children.end();it++){
 328                         tree[it->second].heirarchyID = tree[index].heirarchyID + '.' + toString(counter);
 329                         counter++;
 330                         tree[it->second].level = tree[index].level + 1;
 331
 332                         //save maxLevel for binning the unclassified seqs
 333                         if (tree[it->second].level > maxLevel) { maxLevel = tree[it->second].level; }
 334
 335                         assignHeirarchyIDs(it->second);
 336                 }
 337         }
 338         catch(exception& e) {
 339                 m->errorOut(e, "PhyloTree", "assignHeirarchyIDs");
 340                 exit(1);
 341         }
 342 }
 343 /**************************************************************************************************/
 344 void PhyloTree::setUp(string tfile){
 345         try{
 346                 string taxFileNameTest = tfile.substr(0,tfile.find_last_of(".")+1) + "tree.sum";
 347
 348                 #ifdef USE_MPI
 349                         int pid;
 350                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 351
 352                         if (pid == 0) {  binUnclassified(taxFileNameTest);  }
 353
 354                 #else
 355                         binUnclassified(taxFileNameTest);
 356                 #endif
 357         }
 358         catch(exception& e) {
 359                 m->errorOut(e, "PhyloTree", "setUp");
 360                 exit(1);
 361         }
 362 }
 363 /**************************************************************************************************/
 364 void PhyloTree::binUnclassified(string file){
 365         try {
 366
 367                 ofstream out;
 368                 m->openOutputFile(file, out);
 369
 370                 map<string, int>::iterator itBin;
 371                 map<string, int>::iterator childPointer;
 372
 373                 vector<TaxNode> copy = tree;
 374
 375                 //fill out tree
 376                 fillOutTree(0, copy);
 377
 378                 //get leaf nodes that may need extension
 379                 for (int i = 0; i < copy.size(); i++) {
 380
 381                         if (copy[i].children.size() == 0) {
 382                                 leafNodes[i] = i;
 383                         }
 384                 }
 385
 386                 int copyNodes = copy.size();
 387
 388                 //go through the seqs and if a sequence finest taxon is not the same level as the most finely defined taxon then classify it as unclassified where necessary
 389                 map<int, int>::iterator itLeaf;
 390                 for (itLeaf = leafNodes.begin(); itLeaf != leafNodes.end(); itLeaf++) {
 391
 392                         if (m->control_pressed) {  out.close(); break;  }
 393
 394                         int level = copy[itLeaf->second].level;
 395                         int currentNode = itLeaf->second;
 396
 397                         //this sequence is unclassified at some levels
 398                         while(level < maxLevel){
 399
 400                                 level++;
 401
 402                                 string taxon = "unclassified";
 403
 404                                 //does the parent have a child names 'unclassified'?
 405                                 childPointer = copy[currentNode].children.find(taxon);
 406
 407                                 if(childPointer != copy[currentNode].children.end()){   //if the node already exists, move on
 408                                         currentNode = childPointer->second; //currentNode becomes 'unclassified'
 409                                 }
 410                                 else{                                                                                   //otherwise, create it
 411                                         copy.push_back(TaxNode(taxon));
 412                                         copyNodes++;
 413                                         copy[currentNode].children[taxon] = copyNodes-1;
 414                                         copy[copyNodes-1].parent = currentNode;
 415                                         copy[copyNodes-1].level = copy[currentNode].level + 1;
 416
 417                                         currentNode = copy[currentNode].children[taxon];
 418                                 }
 419                         }
 420                 }
 421
 422                 if (!m->control_pressed) {
 423                         //print copy tree
 424                         print(out, copy);
 425                 }
 426
 427         }
 428         catch(exception& e) {
 429                 m->errorOut(e, "PhyloTree", "binUnclassified");
 430                 exit(1);
 431         }
 432 }
 433 /**************************************************************************************************/
 434 void PhyloTree::fillOutTree(int index, vector<TaxNode>& copy) {
 435         try {
 436
 437                 map<string,int>::iterator it;
 438
 439                 it = copy[index].children.find("unclassified");
 440                 if (it == copy[index].children.end()) { //no unclassified at this level
 441                         string taxon = "unclassified";
 442                         copy.push_back(TaxNode(taxon));
 443                         copy[index].children[taxon] = copy.size()-1;
 444                         copy[copy.size()-1].parent = index;
 445                         copy[copy.size()-1].level = copy[index].level + 1;
 446                 }
 447
 448                 if (tree[index].level < maxLevel) {
 449                         for(it=tree[index].children.begin();it!=tree[index].children.end();it++){ //check your children
 450                                 fillOutTree(it->second, copy);
 451                         }
 452                 }
 453
 454         }
 455         catch(exception& e) {
 456                 m->errorOut(e, "PhyloTree", "fillOutTree");
 457                 exit(1);
 458         }
 459 }
 460 /**************************************************************************************************/
 461 string PhyloTree::getFullTaxonomy(string seqName) {
 462         try {
 463                 string tax = "";
 464
 465                 int currentNode = name2Taxonomy[seqName];
 466
 467                 while (tree[currentNode].parent != -1) {
 468                         tax = tree[currentNode].name + ";" + tax;
 469                         currentNode = tree[currentNode].parent;
 470                 }
 471
 472                 return tax;
 473         }
 474         catch(exception& e) {
 475                 m->errorOut(e, "PhyloTree", "getFullTaxonomy");
 476                 exit(1);
 477         }
 478 }
 479 /**************************************************************************************************/
 480
 481 void PhyloTree::print(ofstream& out, vector<TaxNode>& copy){
 482         try {
 483
 484                 //output mothur version
 485                 out << "#" << m->getVersion() << endl;
 486
 487                 out << copy.size() << endl;
 488
 489                 out << maxLevel << endl;
 490
 491                 for (int i = 0; i < copy.size(); i++) {
 492
 493                         out << copy[i].level << '\t'<< copy[i].name << '\t' << copy[i].children.size() << '\t';
 494
 495                         map<string,int>::iterator it;
 496                         for(it=copy[i].children.begin();it!=copy[i].children.end();it++){
 497                                 out << it->first << '\t' << it->second << '\t';
 498                         }
 499                         out << endl;
 500                 }
 501
 502                 out.close();
 503         }
 504         catch(exception& e) {
 505                 m->errorOut(e, "PhyloTree", "print");
 506                 exit(1);
 507         }
 508 }
 509 /**************************************************************************************************/
 510 void PhyloTree::printTreeNodes(string treefilename) {
 511         try {
 512
 513                 #ifdef USE_MPI
 514                         int pid;
 515                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 516
 517                         if (pid == 0) {
 518
 519                 #endif
 520
 521                         ofstream outTree;
 522                         m->openOutputFile(treefilename, outTree);
 523
 524                         //output mothur version
 525                         outTree << "#" << m->getVersion() << endl;
 526
 527                         //print treenodes
 528                         outTree << tree.size() << endl;
 529                         for (int i = 0; i < tree.size(); i++) {
 530                                 outTree << tree[i].name << '\t' << tree[i].level << '\t' << tree[i].parent << endl;
 531                         }
 532
 533                         //print genus nodes
 534                         outTree << endl << uniqueTaxonomies.size() << endl;
 535                         map<int, int>::iterator it2;
 536                         for (it2=uniqueTaxonomies.begin(); it2!=uniqueTaxonomies.end(); it2++) {  outTree << it2->first << '\t' << tree[it2->first].accessions.size() << endl;  }
 537                         outTree << endl;
 538
 539                         outTree.close();
 540
 541                 #ifdef USE_MPI
 542                         }
 543                 #endif
 544
 545
 546         }
 547         catch(exception& e) {
 548                 m->errorOut(e, "PhyloTree", "printTreeNodes");
 549                 exit(1);
 550         }
 551 }
 552 /**************************************************************************************************/
 553 TaxNode PhyloTree::get(int i ){
 554         try {
 555                 if (i < tree.size()) {  return tree[i];  }
 556                 else {  cout << i << '\t' << tree.size() << endl ; m->mothurOut("Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1); }
 557         }
 558         catch(exception& e) {
 559                 m->errorOut(e, "PhyloTree", "get");
 560                 exit(1);
 561         }
 562 }
 563 /**************************************************************************************************/
 564 TaxNode PhyloTree::get(string seqName){
 565         try {
 566                 map<string, int>::iterator itFind = name2Taxonomy.find(seqName);
 567
 568                 if (itFind != name2Taxonomy.end()) {  return tree[name2Taxonomy[seqName]];  }
 569                 else { m->mothurOut("Cannot find " + seqName + ". Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1);}
 570         }
 571         catch(exception& e) {
 572                 m->errorOut(e, "PhyloTree", "get");
 573                 exit(1);
 574         }
 575 }
 576 /**************************************************************************************************/
 577 string PhyloTree::getName(int i ){
 578         try {
 579                 if (i < tree.size()) {  return tree[i].name;     }
 580                 else { m->mothurOut("Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1); }
 581         }
 582         catch(exception& e) {
 583                 m->errorOut(e, "PhyloTree", "get");
 584                 exit(1);
 585         }
 586 }
 587 /**************************************************************************************************/
 588 int PhyloTree::getIndex(string seqName){
 589         try {
 590                 map<string, int>::iterator itFind = name2Taxonomy.find(seqName);
 591
 592                 if (itFind != name2Taxonomy.end()) {  return name2Taxonomy[seqName];  }
 593                 else { m->mothurOut("Cannot find " + seqName + ". Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1);}
 594         }
 595         catch(exception& e) {
 596                 m->errorOut(e, "PhyloTree", "get");
 597                 exit(1);
 598         }
 599 }
 600 /**************************************************************************************************/
 601 bool PhyloTree::ErrorCheck(vector<string> templateFileNames){
 602         try {
 603
 604                 bool okay = true;
 605
 606                 map<string, int>::iterator itFind;
 607                 map<string, int> taxonomyFileNames = name2Taxonomy;
 608
 609                 for (int i = 0; i < templateFileNames.size(); i++) {
 610                         itFind = taxonomyFileNames.find(templateFileNames[i]);
 611
 612                         if (itFind != taxonomyFileNames.end()) { //found it so erase it
 613                                 taxonomyFileNames.erase(itFind);
 614                         }else {
 615                                 m->mothurOut(templateFileNames[i] + " is in your template file and is not in your taxonomy file. Please correct."); m->mothurOutEndLine();
 616                                 okay = false;
 617                         }
 618
 619                         templateFileNames.erase(templateFileNames.begin()+i);
 620                         i--;
 621                 }
 622
 623                 if (taxonomyFileNames.size() > 0) { //there are names in tax file that are not in template
 624                         okay = false;
 625
 626                         for (itFind = taxonomyFileNames.begin(); itFind != taxonomyFileNames.end(); itFind++) {
 627                                 m->mothurOut(itFind->first + " is in your taxonomy file and is not in your template file. Please correct."); m->mothurOutEndLine();
 628                         }
 629                 }
 630
 631                 return okay;
 632         }
 633         catch(exception& e) {
 634                 m->errorOut(e, "PhyloTree", "ErrorCheck");
 635                 exit(1);
 636         }
 637 }
 638 /**************************************************************************************************/
 639
 640
 641
 642