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"};
  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
  33                         //check to make sure all parameters are valid for command
  34                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  35                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  36                         }
  37
  38                         //check for required parameters
  39                         fastafile = validParameter.validFile(parameters, "fasta", true);
  40                         if (fastafile == "not found") { mothurOut("fasta is a required parameter for the pre.cluster command."); mothurOutEndLine(); abort = true; }
  41                         else if (fastafile == "not open") { abort = true; }
  42
  43                         //check for optional parameter and set defaults
  44                         // ...at some point should added some additional type checking...
  45                         namefile = validParameter.validFile(parameters, "name", true);
  46
  47                         if (namefile == "not found") { namefile =  "";  }
  48                         else if (namefile == "not open") { abort = true; }
  49 //                      else {  readNameFile();  }
  50
  51                         string temp     = validParameter.validFile(parameters, "diffs", false);                         if(temp == "not found"){        temp = "1"; }
  52                         convert(temp, diffs);
  53                 }
  54
  55         }
  56         catch(exception& e) {
  57                 errorOut(e, "PreClusterCommand", "PreClusterCommand");
  58                 exit(1);
  59         }
  60 }
  61
  62 //**********************************************************************************************************************
  63 PreClusterCommand::~PreClusterCommand(){}
  64 //**********************************************************************************************************************
  65
  66 void PreClusterCommand::help(){
  67         try {
  68                 mothurOut("The pre.cluster command groups sequences that are within a given number of base mismatches.\n");
  69                 mothurOut("The pre.cluster command outputs a new fasta and name file.\n");
  70                 mothurOut("The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n");
  71                 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");
  72                 mothurOut("The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n");
  73                 mothurOut("The pre.cluster command should be in the following format: \n");
  74                 mothurOut("pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n");
  75                 mothurOut("Example pre.cluster(fasta=amazon.fasta, diffs=2).\n");
  76                 mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
  77         }
  78         catch(exception& e) {
  79                 errorOut(e, "PreClusterCommand", "help");
  80                 exit(1);
  81         }
  82 }
  83 //**********************************************************************************************************************
  84
  85 int PreClusterCommand::execute(){
  86         try {
  87
  88                 if (abort == true) { return 0; }
  89
  90                 //reads fasta file and return number of seqs
  91                 int numSeqs = readNamesFASTA(); //fills alignSeqs and makes all seqs active
  92
  93                 if (numSeqs == 0) { mothurOut("Error reading fasta file...please correct."); mothurOutEndLine(); return 0;  }
  94                 if (diffs > length) { mothurOut("Error: diffs is greater than your sequence length."); mothurOutEndLine(); return 0;  }
  95
  96                 //clear sizes since you only needed this info to build the alignSeqs seqPNode structs
  97 //              sizes.clear();
  98
  99                 //sort seqs by number of identical seqs
 100                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 101
 102                 //go through active list and cluster everthing you can, until all nodes are inactive
 103                 //taking advantage of the fact that maps are already sorted
 104 //              map<string, bool>::iterator itActive;
 105 //              map<string, bool>::iterator it2Active;
 106                 int count = 0;
 107
 108                 for (int i = 0; i < numSeqs; i++) {
 109
 110                         //are you active
 111                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 112
 113                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 114
 115                                 //try to merge it with all smaller seqs
 116                                 for (int j = i+1; j < numSeqs; j++) {
 117                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 118                                                 //are you within "diff" bases
 119                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 120
 121                                                 if (mismatch <= diffs) {
 122                                                         //merge
 123                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 124                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 125
 126                                                         alignSeqs[j].active = 0;
 127                                                         alignSeqs[j].numIdentical = 0;
 128                                                         count++;
 129                                                 }
 130                                         }//end if j active
 131                                 }//end if i != j
 132
 133                         //remove from active list
 134                                 alignSeqs[i].active = 0;
 135                         }//end if active i
 136                         if(i % 100 == 0)        { cout << i << '\t' << numSeqs - count << '\t' << count << endl;        }
 137                 }
 138
 139                 string newFastaFile = getRootName(fastafile) + "precluster" + getExtension(fastafile);
 140                 string newNamesFile = getRootName(fastafile) + "precluster.names";
 141
 142
 143                 mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); mothurOutEndLine();
 144                 mothurOut("pre.cluster removed " + toString(count) + " sequences."); mothurOutEndLine();
 145                 printData(newFastaFile, newNamesFile);
 146
 147                 return 0;
 148
 149         }
 150         catch(exception& e) {
 151                 errorOut(e, "PreClusterCommand", "execute");
 152                 exit(1);
 153         }
 154 }
 155 /**************************************************************************************************/
 156 int PreClusterCommand::readFASTA(){
 157         try {
 158 //              ifstream inFasta;
 159 //              openInputFile(fastafile, inFasta);
 160 //              length = 0;
 161 ////            map<string, string>::iterator it;
 162 //
 163 //              while (!inFasta.eof()) {
 164 //                      Sequence temp(inFasta);  //read seq
 165 //
 166 //                      if (temp.getName() != "") {
 167 //                              if (namefile != "") {
 168 //                                      //make sure fasta and name files match
 169 //                                      it = names.find(temp.getName());
 170 //                                      if (it == names.end()) { mothurOut(temp.getName() + " is not in your names file, please correct."); mothurOutEndLine(); exit(1); }
 171 //                              }else { sizes[temp.getName()] = 1; }
 172 //
 173 //                              seqPNode tempNode(sizes[temp.getName()], temp);
 174 //                              alignSeqs.push_back(tempNode);
 175 //                              active[temp.getName()] = true;
 176 //                      }
 177 //                      gobble(inFasta);
 178 //              }
 179 //              inFasta.close();
 180 //
 181 //              if (alignSeqs.size() != 0) {  length = alignSeqs[0].seq.getAligned().length();  }
 182 //
 183                 return alignSeqs.size();
 184         }
 185         catch(exception& e) {
 186                 errorOut(e, "PreClusterCommand", "readFASTA");
 187                 exit(1);
 188         }
 189 }
 190 /**************************************************************************************************/
 191
 192 int PreClusterCommand::readNamesFASTA(){
 193         try {
 194                 ifstream inNames;
 195                 ifstream inFasta;
 196
 197                 openInputFile(namefile, inNames);
 198                 openInputFile(fastafile, inFasta);
 199
 200                 string firstCol, secondCol, nameString;
 201                 length = 0;
 202
 203                 while (inFasta && inNames) {
 204
 205                         inNames >> firstCol >> secondCol;
 206                         nameString = secondCol;
 207
 208                         gobble(inNames);
 209                         int size = 1;
 210                         while (secondCol.find_first_of(',') != -1) {
 211                                 size++;
 212                                 secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 213                         }
 214                         Sequence seq(inFasta);
 215                         if (seq.getName() != firstCol) { mothurOut(seq.getName() + " is not in your names file, please correct."); mothurOutEndLine(); exit(1); }
 216                         else{
 217                                 seqPNode tempNode(size, seq, nameString);
 218                                 alignSeqs.push_back(tempNode);
 219                                 if (seq.getAligned().length() > length) {  length = alignSeqs[0].seq.getAligned().length();  }
 220                         }
 221                 }
 222                 inFasta.close();
 223                 inNames.close();
 224                 return alignSeqs.size();
 225         }
 226
 227         catch(exception& e) {
 228                 errorOut(e, "PreClusterCommand", "readNamesFASTA");
 229                 exit(1);
 230         }
 231 }
 232
 233 /**************************************************************************************************/
 234
 235 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 236         try {
 237                 int numBad = 0;
 238
 239                 for (int i = 0; i < seq1.length(); i++) {
 240                         //do they match
 241                         if (seq1[i] != seq2[i]) { numBad++; }
 242                         if (numBad > diffs) { return length;  } //to far to cluster
 243                 }
 244
 245                 return numBad;
 246         }
 247         catch(exception& e) {
 248                 errorOut(e, "PreClusterCommand", "calcMisMatches");
 249                 exit(1);
 250         }
 251 }
 252
 253 /**************************************************************************************************/
 254
 255 void PreClusterCommand::printData(string newfasta, string newname){
 256         try {
 257                 ofstream outFasta;
 258                 ofstream outNames;
 259
 260                 openOutputFile(newfasta, outFasta);
 261                 openOutputFile(newname, outNames);
 262
 263
 264                 for (int i = 0; i < alignSeqs.size(); i++) {
 265                         if (alignSeqs[i].numIdentical != 0) {
 266                                 alignSeqs[i].seq.printSequence(outFasta);
 267                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 268                         }
 269                 }
 270
 271                 outFasta.close();
 272                 outNames.close();
 273
 274         }
 275         catch(exception& e) {
 276                 errorOut(e, "PreClusterCommand", "printData");
 277                 exit(1);
 278         }
 279 }
 280
 281 /**************************************************************************************************/
 282
 283