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 long> 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                 return 0;
 280         }
 281         catch(exception& e) {
 282                 m->errorOut(e, "PhyloTree", "addSeqToTree");
 283                 exit(1);
 284         }
 285 }
 286 /**************************************************************************************************/
 287 vector<int> PhyloTree::getGenusNodes()  {
 288         try {
 289                 genusIndex.clear();
 290                 //generate genusIndexes
 291                 map<int, int>::iterator it2;
 292                 for (it2=uniqueTaxonomies.begin(); it2!=uniqueTaxonomies.end(); it2++) {  genusIndex.push_back(it2->first);     }
 293
 294                 return genusIndex;
 295         }
 296         catch(exception& e) {
 297                 m->errorOut(e, "PhyloTree", "getGenusNodes");
 298                 exit(1);
 299         }
 300 }
 301 /**************************************************************************************************/
 302 vector<int> PhyloTree::getGenusTotals() {
 303         try {
 304
 305                 if (calcTotals) {
 306                         totals.clear();
 307                         //reset counts because we are on a new word
 308                         for (int j = 0; j < genusIndex.size(); j++) {
 309                                 totals.push_back(tree[genusIndex[j]].accessions.size());
 310                         }
 311                         return totals;
 312                 }else{
 313                         return totals;
 314                 }
 315
 316         }
 317         catch(exception& e) {
 318                 m->errorOut(e, "PhyloTree", "getGenusNodes");
 319                 exit(1);
 320         }
 321 }
 322 /**************************************************************************************************/
 323
 324 void PhyloTree::assignHeirarchyIDs(int index){
 325         try {
 326                 map<string,int>::iterator it;
 327                 int counter = 1;
 328
 329                 for(it=tree[index].children.begin();it!=tree[index].children.end();it++){
 330                         tree[it->second].heirarchyID = tree[index].heirarchyID + '.' + toString(counter);
 331                         counter++;
 332                         tree[it->second].level = tree[index].level + 1;
 333
 334                         //save maxLevel for binning the unclassified seqs
 335                         if (tree[it->second].level > maxLevel) { maxLevel = tree[it->second].level; }
 336
 337                         assignHeirarchyIDs(it->second);
 338                 }
 339         }
 340         catch(exception& e) {
 341                 m->errorOut(e, "PhyloTree", "assignHeirarchyIDs");
 342                 exit(1);
 343         }
 344 }
 345 /**************************************************************************************************/
 346 void PhyloTree::setUp(string tfile){
 347         try{
 348                 string taxFileNameTest = tfile.substr(0,tfile.find_last_of(".")+1) + "tree.sum";
 349
 350                 #ifdef USE_MPI
 351                         int pid;
 352                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 353
 354                         if (pid == 0) {  binUnclassified(taxFileNameTest);  }
 355
 356                 #else
 357                         binUnclassified(taxFileNameTest);
 358                 #endif
 359         }
 360         catch(exception& e) {
 361                 m->errorOut(e, "PhyloTree", "setUp");
 362                 exit(1);
 363         }
 364 }
 365 /**************************************************************************************************/
 366 void PhyloTree::binUnclassified(string file){
 367         try {
 368
 369                 ofstream out;
 370                 m->openOutputFile(file, out);
 371
 372                 map<string, int>::iterator itBin;
 373                 map<string, int>::iterator childPointer;
 374
 375                 vector<TaxNode> copy = tree;
 376
 377                 //fill out tree
 378                 fillOutTree(0, copy);
 379
 380                 //get leaf nodes that may need extension
 381                 for (int i = 0; i < copy.size(); i++) {
 382
 383                         if (copy[i].children.size() == 0) {
 384                                 leafNodes[i] = i;
 385                         }
 386                 }
 387
 388                 int copyNodes = copy.size();
 389
 390                 //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
 391                 map<int, int>::iterator itLeaf;
 392                 for (itLeaf = leafNodes.begin(); itLeaf != leafNodes.end(); itLeaf++) {
 393
 394                         if (m->control_pressed) {  out.close(); break;  }
 395
 396                         int level = copy[itLeaf->second].level;
 397                         int currentNode = itLeaf->second;
 398
 399                         //this sequence is unclassified at some levels
 400                         while(level < maxLevel){
 401
 402                                 level++;
 403
 404                                 string taxon = "unclassified";
 405
 406                                 //does the parent have a child names 'unclassified'?
 407                                 childPointer = copy[currentNode].children.find(taxon);
 408
 409                                 if(childPointer != copy[currentNode].children.end()){   //if the node already exists, move on
 410                                         currentNode = childPointer->second; //currentNode becomes 'unclassified'
 411                                 }
 412                                 else{                                                                                   //otherwise, create it
 413                                         copy.push_back(TaxNode(taxon));
 414                                         copyNodes++;
 415                                         copy[currentNode].children[taxon] = copyNodes-1;
 416                                         copy[copyNodes-1].parent = currentNode;
 417                                         copy[copyNodes-1].level = copy[currentNode].level + 1;
 418
 419                                         currentNode = copy[currentNode].children[taxon];
 420                                 }
 421                         }
 422                 }
 423
 424                 if (!m->control_pressed) {
 425                         //print copy tree
 426                         print(out, copy);
 427                 }
 428
 429         }
 430         catch(exception& e) {
 431                 m->errorOut(e, "PhyloTree", "binUnclassified");
 432                 exit(1);
 433         }
 434 }
 435 /**************************************************************************************************/
 436 void PhyloTree::fillOutTree(int index, vector<TaxNode>& copy) {
 437         try {
 438
 439                 map<string,int>::iterator it;
 440
 441                 it = copy[index].children.find("unclassified");
 442                 if (it == copy[index].children.end()) { //no unclassified at this level
 443                         string taxon = "unclassified";
 444                         copy.push_back(TaxNode(taxon));
 445                         copy[index].children[taxon] = copy.size()-1;
 446                         copy[copy.size()-1].parent = index;
 447                         copy[copy.size()-1].level = copy[index].level + 1;
 448                 }
 449
 450                 if (tree[index].level < maxLevel) {
 451                         for(it=tree[index].children.begin();it!=tree[index].children.end();it++){ //check your children
 452                                 fillOutTree(it->second, copy);
 453                         }
 454                 }
 455
 456         }
 457         catch(exception& e) {
 458                 m->errorOut(e, "PhyloTree", "fillOutTree");
 459                 exit(1);
 460         }
 461 }
 462 /**************************************************************************************************/
 463 string PhyloTree::getFullTaxonomy(string seqName) {
 464         try {
 465                 string tax = "";
 466
 467                 int currentNode = name2Taxonomy[seqName];
 468
 469                 while (tree[currentNode].parent != -1) {
 470                         tax = tree[currentNode].name + ";" + tax;
 471                         currentNode = tree[currentNode].parent;
 472                 }
 473
 474                 return tax;
 475         }
 476         catch(exception& e) {
 477                 m->errorOut(e, "PhyloTree", "getFullTaxonomy");
 478                 exit(1);
 479         }
 480 }
 481 /**************************************************************************************************/
 482
 483 void PhyloTree::print(ofstream& out, vector<TaxNode>& copy){
 484         try {
 485
 486                 //output mothur version
 487                 out << "#" << m->getVersion() << endl;
 488
 489                 out << copy.size() << endl;
 490
 491                 out << maxLevel << endl;
 492
 493                 for (int i = 0; i < copy.size(); i++) {
 494
 495                         out << copy[i].level << '\t'<< copy[i].name << '\t' << copy[i].children.size() << '\t';
 496
 497                         map<string,int>::iterator it;
 498                         for(it=copy[i].children.begin();it!=copy[i].children.end();it++){
 499                                 out << it->first << '\t' << it->second << '\t';
 500                         }
 501                         out << endl;
 502                 }
 503
 504                 out.close();
 505         }
 506         catch(exception& e) {
 507                 m->errorOut(e, "PhyloTree", "print");
 508                 exit(1);
 509         }
 510 }
 511 /**************************************************************************************************/
 512 void PhyloTree::printTreeNodes(string treefilename) {
 513         try {
 514
 515                 #ifdef USE_MPI
 516                         int pid;
 517                         MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 518
 519                         if (pid == 0) {
 520
 521                 #endif
 522
 523                         ofstream outTree;
 524                         m->openOutputFile(treefilename, outTree);
 525
 526                         //output mothur version
 527                         outTree << "#" << m->getVersion() << endl;
 528
 529                         //print treenodes
 530                         outTree << tree.size() << endl;
 531                         for (int i = 0; i < tree.size(); i++) {
 532                                 outTree << tree[i].name << '\t' << tree[i].level << '\t' << tree[i].parent << endl;
 533                         }
 534
 535                         //print genus nodes
 536                         outTree << endl << uniqueTaxonomies.size() << endl;
 537                         map<int, int>::iterator it2;
 538                         for (it2=uniqueTaxonomies.begin(); it2!=uniqueTaxonomies.end(); it2++) {  outTree << it2->first << '\t' << tree[it2->first].accessions.size() << endl;  }
 539                         outTree << endl;
 540
 541                         outTree.close();
 542
 543                 #ifdef USE_MPI
 544                         }
 545                 #endif
 546
 547
 548         }
 549         catch(exception& e) {
 550                 m->errorOut(e, "PhyloTree", "printTreeNodes");
 551                 exit(1);
 552         }
 553 }
 554 /**************************************************************************************************/
 555 TaxNode PhyloTree::get(int i ){
 556         try {
 557                 if (i < tree.size()) {  return tree[i];  }
 558                 else {  cout << i << '\t' << tree.size() << endl ; m->mothurOut("Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1); }
 559         }
 560         catch(exception& e) {
 561                 m->errorOut(e, "PhyloTree", "get");
 562                 exit(1);
 563         }
 564 }
 565 /**************************************************************************************************/
 566 TaxNode PhyloTree::get(string seqName){
 567         try {
 568                 map<string, int>::iterator itFind = name2Taxonomy.find(seqName);
 569
 570                 if (itFind != name2Taxonomy.end()) {  return tree[name2Taxonomy[seqName]];  }
 571                 else { m->mothurOut("Cannot find " + seqName + ". Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1);}
 572         }
 573         catch(exception& e) {
 574                 m->errorOut(e, "PhyloTree", "get");
 575                 exit(1);
 576         }
 577 }
 578 /**************************************************************************************************/
 579 string PhyloTree::getName(int i ){
 580         try {
 581                 if (i < tree.size()) {  return tree[i].name;     }
 582                 else { m->mothurOut("Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1); }
 583         }
 584         catch(exception& e) {
 585                 m->errorOut(e, "PhyloTree", "get");
 586                 exit(1);
 587         }
 588 }
 589 /**************************************************************************************************/
 590 int PhyloTree::getIndex(string seqName){
 591         try {
 592                 map<string, int>::iterator itFind = name2Taxonomy.find(seqName);
 593
 594                 if (itFind != name2Taxonomy.end()) {  return name2Taxonomy[seqName];  }
 595                 else { m->mothurOut("Cannot find " + seqName + ". Mismatch with taxonomy and template files. Cannot continue."); m->mothurOutEndLine(); exit(1);}
 596         }
 597         catch(exception& e) {
 598                 m->errorOut(e, "PhyloTree", "get");
 599                 exit(1);
 600         }
 601 }
 602 /**************************************************************************************************/
 603 bool PhyloTree::ErrorCheck(vector<string> templateFileNames){
 604         try {
 605
 606                 bool okay = true;
 607
 608                 map<string, int>::iterator itFind;
 609                 map<string, int> taxonomyFileNames = name2Taxonomy;
 610
 611                 for (int i = 0; i < templateFileNames.size(); i++) {
 612                         itFind = taxonomyFileNames.find(templateFileNames[i]);
 613
 614                         if (itFind != taxonomyFileNames.end()) { //found it so erase it
 615                                 taxonomyFileNames.erase(itFind);
 616                         }else {
 617                                 m->mothurOut(templateFileNames[i] + " is in your template file and is not in your taxonomy file. Please correct."); m->mothurOutEndLine();
 618                                 okay = false;
 619                         }
 620
 621                         //templateFileNames.erase(templateFileNames.begin()+i);
 622                         //i--;
 623                 }
 624                 templateFileNames.clear();
 625
 626                 if (taxonomyFileNames.size() > 0) { //there are names in tax file that are not in template
 627                         okay = false;
 628
 629                         for (itFind = taxonomyFileNames.begin(); itFind != taxonomyFileNames.end(); itFind++) {
 630                                 m->mothurOut(itFind->first + " is in your taxonomy file and is not in your template file. Please correct."); m->mothurOutEndLine();
 631                         }
 632                 }
 633
 634                 return okay;
 635         }
 636         catch(exception& e) {
 637                 m->errorOut(e, "PhyloTree", "ErrorCheck");
 638                 exit(1);
 639         }
 640 }
 641 /**************************************************************************************************/
 642
 643
 644
 645