preclustercommand.cpp

   1 /*
   2  *  preclustercommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 12/21/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "preclustercommand.h"
  11 #include "sequenceparser.h"
  12 #include "deconvolutecommand.h"
  13
  14 //**********************************************************************************************************************
  15 inline bool comparePriority(seqPNode first, seqPNode second) {  return (first.numIdentical > second.numIdentical); }
  16
  17 //**********************************************************************************************************************
  18 vector<string> PreClusterCommand::setParameters(){
  19         try {
  20                 CommandParameter pfasta("fasta", "InputTypes", "", "", "none", "none", "none",false,true); parameters.push_back(pfasta);
  21                 CommandParameter pname("name", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pname);
  22                 CommandParameter pgroup("group", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pgroup);
  23                 CommandParameter pdiffs("diffs", "Number", "", "0", "", "", "",false,false); parameters.push_back(pdiffs);
  24                 CommandParameter pbygroup("bygroup", "Boolean", "", "T", "", "", "",false,false); parameters.push_back(pbygroup);
  25                 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "",false,false); parameters.push_back(pinputdir);
  26                 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "",false,false); parameters.push_back(poutputdir);
  27
  28                 vector<string> myArray;
  29                 for (int i = 0; i < parameters.size(); i++) {   myArray.push_back(parameters[i].name);          }
  30                 return myArray;
  31         }
  32         catch(exception& e) {
  33                 m->errorOut(e, "PreClusterCommand", "setParameters");
  34                 exit(1);
  35         }
  36 }
  37 //**********************************************************************************************************************
  38 string PreClusterCommand::getHelpString(){
  39         try {
  40                 string helpString = "";
  41                 helpString += "The pre.cluster command groups sequences that are within a given number of base mismatches.\n";
  42                 helpString += "The pre.cluster command outputs a new fasta and name file.\n";
  43                 helpString += "The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n";
  44                 helpString += "The names parameter allows you to give a list of seqs that are identical. This file is 2 columns, first column is name or representative sequence, second column is a list of its identical sequences separated by commas.\n";
  45                 helpString += "The group parameter allows you to provide a group file. \n";
  46                 helpString += "The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n";
  47                 helpString += "The bygroup parameter allows you to indicate you whether you want to cluster sequences only within each group, default=T, when a groupfile is given. \n";
  48                 helpString += "The pre.cluster command should be in the following format: \n";
  49                 helpString += "pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n";
  50                 helpString += "Example pre.cluster(fasta=amazon.fasta, diffs=2).\n";
  51                 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
  52                 return helpString;
  53         }
  54         catch(exception& e) {
  55                 m->errorOut(e, "PreClusterCommand", "getHelpString");
  56                 exit(1);
  57         }
  58 }
  59
  60 //**********************************************************************************************************************
  61 PreClusterCommand::PreClusterCommand(){
  62         try {
  63                 abort = true; calledHelp = true;
  64                 setParameters();
  65                 vector<string> tempOutNames;
  66                 outputTypes["fasta"] = tempOutNames;
  67                 outputTypes["name"] = tempOutNames;
  68         }
  69         catch(exception& e) {
  70                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
  71                 exit(1);
  72         }
  73 }
  74 //**********************************************************************************************************************
  75
  76 PreClusterCommand::PreClusterCommand(string option) {
  77         try {
  78                 abort = false; calledHelp = false;
  79
  80                 //allow user to run help
  81                 if(option == "help") { help(); abort = true; calledHelp = true; }
  82                 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
  83
  84                 else {
  85                         vector<string> myArray = setParameters();
  86
  87                         OptionParser parser(option);
  88                         map<string, string> parameters = parser.getParameters();
  89
  90                         ValidParameters validParameter;
  91                         map<string, string>::iterator it;
  92
  93                         //check to make sure all parameters are valid for command
  94                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  95                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  96                         }
  97
  98                         //initialize outputTypes
  99                         vector<string> tempOutNames;
 100                         outputTypes["fasta"] = tempOutNames;
 101                         outputTypes["name"] = tempOutNames;
 102
 103                         //if the user changes the input directory command factory will send this info to us in the output parameter
 104                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 105                         if (inputDir == "not found"){   inputDir = "";          }
 106                         else {
 107                                 string path;
 108                                 it = parameters.find("fasta");
 109                                 //user has given a template file
 110                                 if(it != parameters.end()){
 111                                         path = m->hasPath(it->second);
 112                                         //if the user has not given a path then, add inputdir. else leave path alone.
 113                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 114                                 }
 115
 116                                 it = parameters.find("name");
 117                                 //user has given a template file
 118                                 if(it != parameters.end()){
 119                                         path = m->hasPath(it->second);
 120                                         //if the user has not given a path then, add inputdir. else leave path alone.
 121                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 122                                 }
 123
 124                                 it = parameters.find("group");
 125                                 //user has given a template file
 126                                 if(it != parameters.end()){
 127                                         path = m->hasPath(it->second);
 128                                         //if the user has not given a path then, add inputdir. else leave path alone.
 129                                         if (path == "") {       parameters["group"] = inputDir + it->second;            }
 130                                 }
 131                         }
 132
 133                         //check for required parameters
 134                         fastafile = validParameter.validFile(parameters, "fasta", true);
 135                         if (fastafile == "not found") {
 136                                 fastafile = m->getFastaFile();
 137                                 if (fastafile != "") { m->mothurOut("Using " + fastafile + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
 138                                 else {  m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
 139                         }
 140                         else if (fastafile == "not open") { abort = true; }
 141                         else { m->setFastaFile(fastafile); }
 142
 143                         //if the user changes the output directory command factory will send this info to us in the output parameter
 144                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
 145                                 outputDir = "";
 146                                 outputDir += m->hasPath(fastafile); //if user entered a file with a path then preserve it
 147                         }
 148
 149                         //check for optional parameter and set defaults
 150                         // ...at some point should added some additional type checking...
 151                         namefile = validParameter.validFile(parameters, "name", true);
 152                         if (namefile == "not found") { namefile =  "";  }
 153                         else if (namefile == "not open") { abort = true; }
 154                         else {  m->setNameFile(namefile); }
 155
 156                         groupfile = validParameter.validFile(parameters, "group", true);
 157                         if (groupfile == "not found") { groupfile =  "";  }
 158                         else if (groupfile == "not open") { abort = true; groupfile =  ""; }
 159                         else {   m->setGroupFile(groupfile); }
 160
 161                         string temp     = validParameter.validFile(parameters, "diffs", false);         if(temp == "not found"){        temp = "1"; }
 162                         convert(temp, diffs);
 163
 164                         temp = validParameter.validFile(parameters, "bygroup", false);
 165                         if (temp == "not found") {
 166                                 if (groupfile != "") { temp = "T"; }
 167                                 else { temp = "F"; }
 168                         }
 169                         bygroup = m->isTrue(temp);
 170
 171                         if (bygroup && (groupfile == "")) { m->mothurOut("You cannot set bygroup=T, unless you provide a groupfile."); m->mothurOutEndLine(); abort=true; }
 172
 173                 }
 174
 175         }
 176         catch(exception& e) {
 177                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
 178                 exit(1);
 179         }
 180 }
 181 //**********************************************************************************************************************
 182
 183 int PreClusterCommand::execute(){
 184         try {
 185
 186                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 187
 188                 int start = time(NULL);
 189
 190                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 191                 string newFastaFile = fileroot + "precluster" + m->getExtension(fastafile);
 192                 string newNamesFile = fileroot + "precluster.names";
 193
 194                 if (bygroup) {
 195                         //clear out old files
 196                         ofstream outFasta; m->openOutputFile(newFastaFile, outFasta); outFasta.close();
 197                         ofstream outNames; m->openOutputFile(newNamesFile, outNames);  outNames.close();
 198
 199                         //parse fasta and name file by group
 200                         SequenceParser* parser;
 201                         if (namefile != "") { parser = new SequenceParser(groupfile, fastafile, namefile);      }
 202                         else                            { parser = new SequenceParser(groupfile, fastafile);                    }
 203
 204                         vector<string> groups = parser->getNamesOfGroups();
 205
 206                         //precluster each group
 207                         for (int i = 0; i < groups.size(); i++) {
 208
 209                                 start = time(NULL);
 210
 211                                 if (m->control_pressed) {  delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 212
 213                                 m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
 214
 215                                 map<string, string> thisNameMap;
 216                                 if (namefile != "") { thisNameMap = parser->getNameMap(groups[i]); }
 217                                 vector<Sequence> thisSeqs = parser->getSeqs(groups[i]);
 218
 219                                 //fill alignSeqs with this groups info.
 220                                 int numSeqs = loadSeqs(thisNameMap, thisSeqs);
 221
 222                                 if (m->control_pressed) {  delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 223
 224                                 if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 225
 226                                 int count = process();
 227
 228                                 if (m->control_pressed) {  delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 229
 230                                 m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 231                                 m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 232                                 printData(newFastaFile, newNamesFile);
 233
 234                                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 235
 236                         }
 237
 238                         //run unique.seqs for deconvolute results
 239                         string inputString = "fasta=" + newFastaFile;
 240                         if (namefile != "") { inputString += ", name=" + newNamesFile; }
 241                         m->mothurOutEndLine();
 242                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 243                         m->mothurOut("Running command: unique.seqs(" + inputString + ")"); m->mothurOutEndLine();
 244
 245                         Command* uniqueCommand = new DeconvoluteCommand(inputString);
 246                         uniqueCommand->execute();
 247
 248                         map<string, vector<string> > filenames = uniqueCommand->getOutputFiles();
 249
 250                         delete uniqueCommand;
 251
 252                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 253
 254                         newNamesFile = filenames["name"][0];
 255                         newFastaFile = filenames["fasta"][0];
 256
 257                 }else {
 258                         if (namefile != "") { readNameFile(); }
 259
 260                         //reads fasta file and return number of seqs
 261                         int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 262
 263                         if (m->control_pressed) { return 0; }
 264
 265                         if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 266                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 267
 268                         int count = process();
 269
 270                         if (m->control_pressed) { return 0; }
 271
 272                         m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 273                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 274                         printData(newFastaFile, newNamesFile);
 275
 276                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 277                 }
 278
 279                 if (m->control_pressed) { m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 280
 281                 m->mothurOutEndLine();
 282                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 283                 m->mothurOut(newFastaFile); m->mothurOutEndLine();      outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 284                 m->mothurOut(newNamesFile); m->mothurOutEndLine();      outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile);
 285                 m->mothurOutEndLine();
 286
 287                 //set fasta file as new current fastafile
 288                 string current = "";
 289                 itTypes = outputTypes.find("fasta");
 290                 if (itTypes != outputTypes.end()) {
 291                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 292                 }
 293
 294                 itTypes = outputTypes.find("name");
 295                 if (itTypes != outputTypes.end()) {
 296                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 297                 }
 298
 299                 return 0;
 300
 301         }
 302         catch(exception& e) {
 303                 m->errorOut(e, "PreClusterCommand", "execute");
 304                 exit(1);
 305         }
 306 }
 307 /**************************************************************************************************/
 308 int PreClusterCommand::process(){
 309         try {
 310
 311                 //sort seqs by number of identical seqs
 312                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 313
 314                 int count = 0;
 315                 int numSeqs = alignSeqs.size();
 316
 317                 //think about running through twice...
 318                 for (int i = 0; i < numSeqs; i++) {
 319
 320                         //are you active
 321                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 322
 323                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 324
 325                                 //try to merge it with all smaller seqs
 326                                 for (int j = i+1; j < numSeqs; j++) {
 327
 328                                         if (m->control_pressed) { return 0; }
 329
 330                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 331                                                 //are you within "diff" bases
 332                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 333
 334                                                 if (mismatch <= diffs) {
 335                                                         //merge
 336                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 337                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 338
 339                                                         alignSeqs[j].active = 0;
 340                                                         alignSeqs[j].numIdentical = 0;
 341                                                         count++;
 342                                                 }
 343                                         }//end if j active
 344                                 }//end if i != j
 345
 346                                 //remove from active list
 347                                 alignSeqs[i].active = 0;
 348
 349                         }//end if active i
 350                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 351                 }
 352
 353                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 354
 355                 return count;
 356
 357         }
 358         catch(exception& e) {
 359                 m->errorOut(e, "PreClusterCommand", "process");
 360                 exit(1);
 361         }
 362 }
 363 /**************************************************************************************************/
 364 int PreClusterCommand::readFASTA(){
 365         try {
 366                 //ifstream inNames;
 367                 ifstream inFasta;
 368
 369                 //m->openInputFile(namefile, inNames);
 370                 m->openInputFile(fastafile, inFasta);
 371
 372                 //string firstCol, secondCol, nameString;
 373                 length = 0;
 374
 375                 while (!inFasta.eof()) {
 376
 377                         if (m->control_pressed) { inFasta.close(); return 0; }
 378
 379                         //inNames >> firstCol >> secondCol;
 380                         //nameString = secondCol;
 381
 382                         //m->gobble(inNames);
 383                         //int size = 1;
 384                         //while (secondCol.find_first_of(',') != -1) {
 385                         //      size++;
 386                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 387                         //}
 388
 389                         Sequence seq(inFasta);  m->gobble(inFasta);
 390
 391                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 392                                 if (namefile != "") {
 393                                         itSize = sizes.find(seq.getName());
 394
 395                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 396                                         else{
 397                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 398                                                 alignSeqs.push_back(tempNode);
 399                                                 if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 400                                         }
 401                                 }else { //no names file, you are identical to yourself
 402                                         seqPNode tempNode(1, seq, seq.getName());
 403                                         alignSeqs.push_back(tempNode);
 404                                         if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 405                                 }
 406                         }
 407                 }
 408                 inFasta.close();
 409                 //inNames.close();
 410                 return alignSeqs.size();
 411         }
 412
 413         catch(exception& e) {
 414                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 415                 exit(1);
 416         }
 417 }
 418 /**************************************************************************************************/
 419 int PreClusterCommand::loadSeqs(map<string, string>& thisName, vector<Sequence>& thisSeqs){
 420         try {
 421                 length = 0;
 422                 alignSeqs.clear();
 423                 map<string, string>::iterator it;
 424                 bool error = false;
 425
 426                 for (int i = 0; i < thisSeqs.size(); i++) {
 427
 428                         if (m->control_pressed) { return 0; }
 429
 430                         if (namefile != "") {
 431                                 it = thisName.find(thisSeqs[i].getName());
 432
 433                                 //should never be true since parser checks for this
 434                                 if (it == thisName.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); error = true; }
 435                                 else{
 436                                         //get number of reps
 437                                         int numReps = 0;
 438                                         for(int j=0;j<(it->second).length();j++){
 439                                                 if((it->second)[j] == ','){     numReps++;      }
 440                                         }
 441
 442                                         seqPNode tempNode(numReps, thisSeqs[i], it->second);
 443                                         alignSeqs.push_back(tempNode);
 444                                         if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 445                                 }
 446                         }else { //no names file, you are identical to yourself
 447                                 seqPNode tempNode(1, thisSeqs[i], thisSeqs[i].getName());
 448                                 alignSeqs.push_back(tempNode);
 449                                 if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 450                         }
 451                 }
 452
 453                 //sanity check
 454                 if (error) { m->control_pressed = true; }
 455
 456                 thisSeqs.clear();
 457
 458                 return alignSeqs.size();
 459         }
 460
 461         catch(exception& e) {
 462                 m->errorOut(e, "PreClusterCommand", "loadSeqs");
 463                 exit(1);
 464         }
 465 }
 466
 467 /**************************************************************************************************/
 468
 469 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 470         try {
 471                 int numBad = 0;
 472
 473                 for (int i = 0; i < seq1.length(); i++) {
 474                         //do they match
 475                         if (seq1[i] != seq2[i]) { numBad++; }
 476                         if (numBad > diffs) { return length;  } //to far to cluster
 477                 }
 478
 479                 return numBad;
 480         }
 481         catch(exception& e) {
 482                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 483                 exit(1);
 484         }
 485 }
 486
 487 /**************************************************************************************************/
 488
 489 void PreClusterCommand::printData(string newfasta, string newname){
 490         try {
 491                 ofstream outFasta;
 492                 ofstream outNames;
 493
 494                 if (bygroup) {
 495                         m->openOutputFileAppend(newfasta, outFasta);
 496                         m->openOutputFileAppend(newname, outNames);
 497                 }else {
 498                         m->openOutputFile(newfasta, outFasta);
 499                         m->openOutputFile(newname, outNames);
 500                 }
 501
 502                 for (int i = 0; i < alignSeqs.size(); i++) {
 503                         if (alignSeqs[i].numIdentical != 0) {
 504                                 alignSeqs[i].seq.printSequence(outFasta);
 505                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 506                         }
 507                 }
 508
 509                 outFasta.close();
 510                 outNames.close();
 511
 512         }
 513         catch(exception& e) {
 514                 m->errorOut(e, "PreClusterCommand", "printData");
 515                 exit(1);
 516         }
 517 }
 518 /**************************************************************************************************/
 519
 520 void PreClusterCommand::readNameFile(){
 521         try {
 522                 ifstream in;
 523                 m->openInputFile(namefile, in);
 524                 string firstCol, secondCol;
 525
 526                 while (!in.eof()) {
 527                         in >> firstCol >> secondCol; m->gobble(in);
 528                         names[firstCol] = secondCol;
 529                         int size = 1;
 530
 531                         for(int i=0;i<secondCol.size();i++){
 532                                 if(secondCol[i] == ','){        size++; }
 533                         }
 534                         sizes[firstCol] = size;
 535                 }
 536                 in.close();
 537         }
 538         catch(exception& e) {
 539                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 540                 exit(1);
 541         }
 542 }
 543
 544 /**************************************************************************************************/
 545
 546