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