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
  12 //**********************************************************************************************************************
  13 inline bool comparePriority(seqPNode first, seqPNode second) {  return (first.numIdentical > second.numIdentical); }
  14 //**********************************************************************************************************************
  15 vector<string> PreClusterCommand::getValidParameters(){
  16         try {
  17                 string Array[] =  {"fasta", "name", "diffs", "outputdir","inputdir"};
  18                 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  19                 return myArray;
  20         }
  21         catch(exception& e) {
  22                 m->errorOut(e, "PreClusterCommand", "getValidParameters");
  23                 exit(1);
  24         }
  25 }
  26 //**********************************************************************************************************************
  27 PreClusterCommand::PreClusterCommand(){
  28         try {
  29                 abort = true; calledHelp = true;
  30                 vector<string> tempOutNames;
  31                 outputTypes["fasta"] = tempOutNames;
  32                 outputTypes["name"] = tempOutNames;
  33         }
  34         catch(exception& e) {
  35                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
  36                 exit(1);
  37         }
  38 }
  39 //**********************************************************************************************************************
  40 vector<string> PreClusterCommand::getRequiredParameters(){
  41         try {
  42                 string Array[] =  {"fasta"};
  43                 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  44                 return myArray;
  45         }
  46         catch(exception& e) {
  47                 m->errorOut(e, "PreClusterCommand", "getRequiredParameters");
  48                 exit(1);
  49         }
  50 }
  51 //**********************************************************************************************************************
  52 vector<string> PreClusterCommand::getRequiredFiles(){
  53         try {
  54                 vector<string> myArray;
  55                 return myArray;
  56         }
  57         catch(exception& e) {
  58                 m->errorOut(e, "PreClusterCommand", "getRequiredFiles");
  59                 exit(1);
  60         }
  61 }
  62 //**********************************************************************************************************************
  63
  64 PreClusterCommand::PreClusterCommand(string option) {
  65         try {
  66                 abort = false; calledHelp = false;
  67
  68                 //allow user to run help
  69                 if(option == "help") { help(); abort = true; calledHelp = true; }
  70
  71                 else {
  72                         //valid paramters for this command
  73                         string Array[] =  {"fasta", "name", "diffs", "outputdir","inputdir"};
  74                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  75
  76                         OptionParser parser(option);
  77                         map<string, string> parameters = parser.getParameters();
  78
  79                         ValidParameters validParameter;
  80                         map<string, string>::iterator it;
  81
  82                         //check to make sure all parameters are valid for command
  83                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  84                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  85                         }
  86
  87                         //initialize outputTypes
  88                         vector<string> tempOutNames;
  89                         outputTypes["fasta"] = tempOutNames;
  90                         outputTypes["name"] = tempOutNames;
  91
  92                         //if the user changes the input directory command factory will send this info to us in the output parameter
  93                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
  94                         if (inputDir == "not found"){   inputDir = "";          }
  95                         else {
  96                                 string path;
  97                                 it = parameters.find("fasta");
  98                                 //user has given a template file
  99                                 if(it != parameters.end()){
 100                                         path = m->hasPath(it->second);
 101                                         //if the user has not given a path then, add inputdir. else leave path alone.
 102                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 103                                 }
 104
 105                                 it = parameters.find("name");
 106                                 //user has given a template file
 107                                 if(it != parameters.end()){
 108                                         path = m->hasPath(it->second);
 109                                         //if the user has not given a path then, add inputdir. else leave path alone.
 110                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 111                                 }
 112                         }
 113
 114                         //check for required parameters
 115                         fastafile = validParameter.validFile(parameters, "fasta", true);
 116                         if (fastafile == "not found") { m->mothurOut("fasta is a required parameter for the pre.cluster command."); m->mothurOutEndLine(); abort = true; }
 117                         else if (fastafile == "not open") { abort = true; }
 118
 119                         //if the user changes the output directory command factory will send this info to us in the output parameter
 120                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
 121                                 outputDir = "";
 122                                 outputDir += m->hasPath(fastafile); //if user entered a file with a path then preserve it
 123                         }
 124
 125                         //check for optional parameter and set defaults
 126                         // ...at some point should added some additional type checking...
 127                         namefile = validParameter.validFile(parameters, "name", true);
 128                         if (namefile == "not found") { namefile =  "";  }
 129                         else if (namefile == "not open") { abort = true; }
 130                         else {  readNameFile();  }
 131
 132                         string temp     = validParameter.validFile(parameters, "diffs", false);                         if(temp == "not found"){        temp = "1"; }
 133                         convert(temp, diffs);
 134                 }
 135
 136         }
 137         catch(exception& e) {
 138                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
 139                 exit(1);
 140         }
 141 }
 142
 143 //**********************************************************************************************************************
 144 PreClusterCommand::~PreClusterCommand(){}
 145 //**********************************************************************************************************************
 146
 147 void PreClusterCommand::help(){
 148         try {
 149                 m->mothurOut("The pre.cluster command groups sequences that are within a given number of base mismatches.\n");
 150                 m->mothurOut("The pre.cluster command outputs a new fasta and name file.\n");
 151                 m->mothurOut("The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n");
 152                 m->mothurOut("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");
 153                 m->mothurOut("The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n");
 154                 m->mothurOut("The pre.cluster command should be in the following format: \n");
 155                 m->mothurOut("pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n");
 156                 m->mothurOut("Example pre.cluster(fasta=amazon.fasta, diffs=2).\n");
 157                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
 158         }
 159         catch(exception& e) {
 160                 m->errorOut(e, "PreClusterCommand", "help");
 161                 exit(1);
 162         }
 163 }
 164 //**********************************************************************************************************************
 165
 166 int PreClusterCommand::execute(){
 167         try {
 168
 169                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 170
 171                 int start = time(NULL);
 172
 173                 //reads fasta file and return number of seqs
 174                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 175
 176                 if (m->control_pressed) { return 0; }
 177
 178                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 179                 if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 180
 181                 //clear sizes since you only needed this info to build the alignSeqs seqPNode structs
 182 //              sizes.clear();
 183
 184                 //sort seqs by number of identical seqs
 185                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 186
 187                 int count = 0;
 188
 189                 //think about running through twice...
 190                 for (int i = 0; i < numSeqs; i++) {
 191
 192                         //are you active
 193                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 194
 195                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 196
 197                                 //try to merge it with all smaller seqs
 198                                 for (int j = i+1; j < numSeqs; j++) {
 199
 200                                         if (m->control_pressed) { return 0; }
 201
 202                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 203                                                 //are you within "diff" bases
 204                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 205
 206                                                 if (mismatch <= diffs) {
 207                                                         //merge
 208                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 209                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 210
 211                                                         alignSeqs[j].active = 0;
 212                                                         alignSeqs[j].numIdentical = 0;
 213                                                         count++;
 214                                                 }
 215                                         }//end if j active
 216                                 }//end if i != j
 217
 218                         //remove from active list
 219                                 alignSeqs[i].active = 0;
 220
 221                         }//end if active i
 222                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 223                 }
 224
 225                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 226
 227
 228                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 229
 230                 string newFastaFile = fileroot + "precluster" + m->getExtension(fastafile);
 231                 string newNamesFile = fileroot + "precluster.names";
 232
 233                 if (m->control_pressed) { return 0; }
 234
 235                 m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 236                 m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 237                 printData(newFastaFile, newNamesFile);
 238
 239                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 240
 241                 if (m->control_pressed) { remove(newFastaFile.c_str()); remove(newNamesFile.c_str()); return 0; }
 242
 243                 m->mothurOutEndLine();
 244                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 245                 m->mothurOut(newFastaFile); m->mothurOutEndLine();      outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 246                 m->mothurOut(newNamesFile); m->mothurOutEndLine();      outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile);
 247                 m->mothurOutEndLine();
 248
 249                 //set fasta file as new current fastafile
 250                 string current = "";
 251                 itTypes = outputTypes.find("fasta");
 252                 if (itTypes != outputTypes.end()) {
 253                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 254                 }
 255
 256                 itTypes = outputTypes.find("name");
 257                 if (itTypes != outputTypes.end()) {
 258                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 259                 }
 260
 261                 return 0;
 262
 263         }
 264         catch(exception& e) {
 265                 m->errorOut(e, "PreClusterCommand", "execute");
 266                 exit(1);
 267         }
 268 }
 269
 270 /**************************************************************************************************/
 271
 272 //this requires the names and fasta file to be in the same order
 273
 274 int PreClusterCommand::readFASTA(){
 275         try {
 276                 //ifstream inNames;
 277                 ifstream inFasta;
 278
 279                 //m->openInputFile(namefile, inNames);
 280                 m->openInputFile(fastafile, inFasta);
 281
 282                 //string firstCol, secondCol, nameString;
 283                 length = 0;
 284
 285                 while (!inFasta.eof()) {
 286
 287                         if (m->control_pressed) { inFasta.close(); return 0; }
 288
 289                         //inNames >> firstCol >> secondCol;
 290                         //nameString = secondCol;
 291
 292                         //m->gobble(inNames);
 293                         //int size = 1;
 294                         //while (secondCol.find_first_of(',') != -1) {
 295                         //      size++;
 296                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 297                         //}
 298
 299                         Sequence seq(inFasta);  m->gobble(inFasta);
 300
 301                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 302                                 if (namefile != "") {
 303                                         itSize = sizes.find(seq.getName());
 304
 305                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 306                                         else{
 307                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 308                                                 alignSeqs.push_back(tempNode);
 309                                                 if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 310                                         }
 311                                 }else { //no names file, you are identical to yourself
 312                                         seqPNode tempNode(1, seq, seq.getName());
 313                                         alignSeqs.push_back(tempNode);
 314                                         if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 315                                 }
 316                         }
 317                 }
 318                 inFasta.close();
 319                 //inNames.close();
 320                 return alignSeqs.size();
 321         }
 322
 323         catch(exception& e) {
 324                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 325                 exit(1);
 326         }
 327 }
 328
 329 /**************************************************************************************************/
 330
 331 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 332         try {
 333                 int numBad = 0;
 334
 335                 for (int i = 0; i < seq1.length(); i++) {
 336                         //do they match
 337                         if (seq1[i] != seq2[i]) { numBad++; }
 338                         if (numBad > diffs) { return length;  } //to far to cluster
 339                 }
 340
 341                 return numBad;
 342         }
 343         catch(exception& e) {
 344                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 345                 exit(1);
 346         }
 347 }
 348
 349 /**************************************************************************************************/
 350
 351 void PreClusterCommand::printData(string newfasta, string newname){
 352         try {
 353                 ofstream outFasta;
 354                 ofstream outNames;
 355
 356                 m->openOutputFile(newfasta, outFasta);
 357                 m->openOutputFile(newname, outNames);
 358
 359
 360                 for (int i = 0; i < alignSeqs.size(); i++) {
 361                         if (alignSeqs[i].numIdentical != 0) {
 362                                 alignSeqs[i].seq.printSequence(outFasta);
 363                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 364                         }
 365                 }
 366
 367                 outFasta.close();
 368                 outNames.close();
 369
 370         }
 371         catch(exception& e) {
 372                 m->errorOut(e, "PreClusterCommand", "printData");
 373                 exit(1);
 374         }
 375 }
 376 /**************************************************************************************************/
 377
 378 void PreClusterCommand::readNameFile(){
 379         try {
 380                 ifstream in;
 381                 m->openInputFile(namefile, in);
 382                 string firstCol, secondCol;
 383
 384                 while (!in.eof()) {
 385                         in >> firstCol >> secondCol; m->gobble(in);
 386                         names[firstCol] = secondCol;
 387                         int size = 1;
 388
 389                         for(int i=0;i<secondCol.size();i++){
 390                                 if(secondCol[i] == ','){        size++; }
 391                         }
 392                         sizes[firstCol] = size;
 393                 }
 394                 in.close();
 395         }
 396         catch(exception& e) {
 397                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 398                 exit(1);
 399         }
 400 }
 401
 402 /**************************************************************************************************/
 403
 404