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 pre.cluster 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                         delete parser;
 239
 240                         //run unique.seqs for deconvolute results
 241                         string inputString = "fasta=" + newFastaFile;
 242                         if (namefile != "") { inputString += ", name=" + newNamesFile; }
 243                         m->mothurOutEndLine();
 244                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 245                         m->mothurOut("Running command: unique.seqs(" + inputString + ")"); m->mothurOutEndLine();
 246
 247                         Command* uniqueCommand = new DeconvoluteCommand(inputString);
 248                         uniqueCommand->execute();
 249
 250                         map<string, vector<string> > filenames = uniqueCommand->getOutputFiles();
 251
 252                         delete uniqueCommand;
 253
 254                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 255
 256                         m->renameFile(filenames["fasta"][0], newFastaFile);
 257
 258                 }else {
 259                         if (namefile != "") { readNameFile(); }
 260
 261                         //reads fasta file and return number of seqs
 262                         int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 263
 264                         if (m->control_pressed) { return 0; }
 265
 266                         if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 267                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 268
 269                         int count = process();
 270
 271                         if (m->control_pressed) { return 0; }
 272
 273                         m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 274                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 275                         printData(newFastaFile, newNamesFile);
 276
 277                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 278                 }
 279
 280                 if (m->control_pressed) { m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 281
 282                 m->mothurOutEndLine();
 283                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 284                 m->mothurOut(newFastaFile); m->mothurOutEndLine();      outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 285                 m->mothurOut(newNamesFile); m->mothurOutEndLine();      outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile);
 286                 m->mothurOutEndLine();
 287
 288                 //set fasta file as new current fastafile
 289                 string current = "";
 290                 itTypes = outputTypes.find("fasta");
 291                 if (itTypes != outputTypes.end()) {
 292                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 293                 }
 294
 295                 itTypes = outputTypes.find("name");
 296                 if (itTypes != outputTypes.end()) {
 297                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 298                 }
 299
 300                 return 0;
 301
 302         }
 303         catch(exception& e) {
 304                 m->errorOut(e, "PreClusterCommand", "execute");
 305                 exit(1);
 306         }
 307 }
 308 /**************************************************************************************************/
 309 int PreClusterCommand::process(){
 310         try {
 311
 312                 //sort seqs by number of identical seqs
 313                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 314
 315                 int count = 0;
 316                 int numSeqs = alignSeqs.size();
 317
 318                 //think about running through twice...
 319                 for (int i = 0; i < numSeqs; i++) {
 320
 321                         //are you active
 322                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 323
 324                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 325
 326                                 //try to merge it with all smaller seqs
 327                                 for (int j = i+1; j < numSeqs; j++) {
 328
 329                                         if (m->control_pressed) { return 0; }
 330
 331                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 332                                                 //are you within "diff" bases
 333                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 334
 335                                                 if (mismatch <= diffs) {
 336                                                         //merge
 337                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 338                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 339
 340                                                         alignSeqs[j].active = 0;
 341                                                         alignSeqs[j].numIdentical = 0;
 342                                                         count++;
 343                                                 }
 344                                         }//end if j active
 345                                 }//end if i != j
 346
 347                                 //remove from active list
 348                                 alignSeqs[i].active = 0;
 349
 350                         }//end if active i
 351                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 352                 }
 353
 354                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 355
 356                 return count;
 357
 358         }
 359         catch(exception& e) {
 360                 m->errorOut(e, "PreClusterCommand", "process");
 361                 exit(1);
 362         }
 363 }
 364 /**************************************************************************************************/
 365 int PreClusterCommand::readFASTA(){
 366         try {
 367                 //ifstream inNames;
 368                 ifstream inFasta;
 369
 370                 //m->openInputFile(namefile, inNames);
 371                 m->openInputFile(fastafile, inFasta);
 372
 373                 //string firstCol, secondCol, nameString;
 374                 length = 0;
 375
 376                 while (!inFasta.eof()) {
 377
 378                         if (m->control_pressed) { inFasta.close(); return 0; }
 379
 380                         //inNames >> firstCol >> secondCol;
 381                         //nameString = secondCol;
 382
 383                         //m->gobble(inNames);
 384                         //int size = 1;
 385                         //while (secondCol.find_first_of(',') != -1) {
 386                         //      size++;
 387                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 388                         //}
 389
 390                         Sequence seq(inFasta);  m->gobble(inFasta);
 391
 392                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 393                                 if (namefile != "") {
 394                                         itSize = sizes.find(seq.getName());
 395
 396                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 397                                         else{
 398                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 399                                                 alignSeqs.push_back(tempNode);
 400                                                 if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 401                                         }
 402                                 }else { //no names file, you are identical to yourself
 403                                         seqPNode tempNode(1, seq, seq.getName());
 404                                         alignSeqs.push_back(tempNode);
 405                                         if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 406                                 }
 407                         }
 408                 }
 409                 inFasta.close();
 410                 //inNames.close();
 411                 return alignSeqs.size();
 412         }
 413
 414         catch(exception& e) {
 415                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 416                 exit(1);
 417         }
 418 }
 419 /**************************************************************************************************/
 420 int PreClusterCommand::loadSeqs(map<string, string>& thisName, vector<Sequence>& thisSeqs){
 421         try {
 422                 length = 0;
 423                 alignSeqs.clear();
 424                 map<string, string>::iterator it;
 425                 bool error = false;
 426
 427                 for (int i = 0; i < thisSeqs.size(); i++) {
 428
 429                         if (m->control_pressed) { return 0; }
 430
 431                         if (namefile != "") {
 432                                 it = thisName.find(thisSeqs[i].getName());
 433
 434                                 //should never be true since parser checks for this
 435                                 if (it == thisName.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); error = true; }
 436                                 else{
 437                                         //get number of reps
 438                                         int numReps = 1;
 439                                         for(int j=0;j<(it->second).length();j++){
 440                                                 if((it->second)[j] == ','){     numReps++;      }
 441                                         }
 442
 443                                         seqPNode tempNode(numReps, thisSeqs[i], it->second);
 444                                         alignSeqs.push_back(tempNode);
 445                                         if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 446                                 }
 447                         }else { //no names file, you are identical to yourself
 448                                 seqPNode tempNode(1, thisSeqs[i], thisSeqs[i].getName());
 449                                 alignSeqs.push_back(tempNode);
 450                                 if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 451                         }
 452                 }
 453
 454                 //sanity check
 455                 if (error) { m->control_pressed = true; }
 456
 457                 thisSeqs.clear();
 458
 459                 return alignSeqs.size();
 460         }
 461
 462         catch(exception& e) {
 463                 m->errorOut(e, "PreClusterCommand", "loadSeqs");
 464                 exit(1);
 465         }
 466 }
 467
 468 /**************************************************************************************************/
 469
 470 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 471         try {
 472                 int numBad = 0;
 473
 474                 for (int i = 0; i < seq1.length(); i++) {
 475                         //do they match
 476                         if (seq1[i] != seq2[i]) { numBad++; }
 477                         if (numBad > diffs) { return length;  } //to far to cluster
 478                 }
 479
 480                 return numBad;
 481         }
 482         catch(exception& e) {
 483                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 484                 exit(1);
 485         }
 486 }
 487
 488 /**************************************************************************************************/
 489
 490 void PreClusterCommand::printData(string newfasta, string newname){
 491         try {
 492                 ofstream outFasta;
 493                 ofstream outNames;
 494
 495                 if (bygroup) {
 496                         m->openOutputFileAppend(newfasta, outFasta);
 497                         m->openOutputFileAppend(newname, outNames);
 498                 }else {
 499                         m->openOutputFile(newfasta, outFasta);
 500                         m->openOutputFile(newname, outNames);
 501                 }
 502
 503                 for (int i = 0; i < alignSeqs.size(); i++) {
 504                         if (alignSeqs[i].numIdentical != 0) {
 505                                 alignSeqs[i].seq.printSequence(outFasta);
 506                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 507                         }
 508                 }
 509
 510                 outFasta.close();
 511                 outNames.close();
 512
 513         }
 514         catch(exception& e) {
 515                 m->errorOut(e, "PreClusterCommand", "printData");
 516                 exit(1);
 517         }
 518 }
 519 /**************************************************************************************************/
 520
 521 void PreClusterCommand::readNameFile(){
 522         try {
 523                 ifstream in;
 524                 m->openInputFile(namefile, in);
 525                 string firstCol, secondCol;
 526
 527                 while (!in.eof()) {
 528                         in >> firstCol >> secondCol; m->gobble(in);
 529                         names[firstCol] = secondCol;
 530                         int size = 1;
 531
 532                         for(int i=0;i<secondCol.size();i++){
 533                                 if(secondCol[i] == ','){        size++; }
 534                         }
 535                         sizes[firstCol] = size;
 536                 }
 537                 in.close();
 538         }
 539         catch(exception& e) {
 540                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 541                 exit(1);
 542         }
 543 }
 544
 545 /**************************************************************************************************/
 546
 547