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") { m->mothurOut("fasta is a required parameter for the pre.cluster command."); m->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                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
  86                 exit(1);
  87         }
  88 }
  89
  90 //**********************************************************************************************************************
  91 PreClusterCommand::~PreClusterCommand(){}
  92 //**********************************************************************************************************************
  93
  94 void PreClusterCommand::help(){
  95         try {
  96                 m->mothurOut("The pre.cluster command groups sequences that are within a given number of base mismatches.\n");
  97                 m->mothurOut("The pre.cluster command outputs a new fasta and name file.\n");
  98                 m->mothurOut("The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n");
  99                 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");
 100                 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");
 101                 m->mothurOut("The pre.cluster command should be in the following format: \n");
 102                 m->mothurOut("pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n");
 103                 m->mothurOut("Example pre.cluster(fasta=amazon.fasta, diffs=2).\n");
 104                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
 105         }
 106         catch(exception& e) {
 107                 m->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                 int start = time(NULL);
 119
 120                 //reads fasta file and return number of seqs
 121                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 122
 123                 if (m->control_pressed) { return 0; }
 124
 125                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 126                 if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 127
 128                 string fileroot = outputDir + getRootName(getSimpleName(fastafile));
 129                 string newFastaFile = fileroot + "precluster" + getExtension(fastafile);
 130                 string newNamesFile = fileroot + "precluster.names";
 131                 ofstream outFasta;
 132                 ofstream outNames;
 133
 134                 openOutputFile(newFastaFile, outFasta);
 135                 openOutputFile(newNamesFile, outNames);
 136
 137                 //sort seqs by number of identical seqs
 138                 alignSeqs.sort(comparePriority);
 139
 140                 int count = 0;
 141                 int i = 0;
 142                 //think about running through twice...
 143                 list<seqPNode>::iterator itList;
 144                 list<seqPNode>::iterator itList2;
 145                 for (itList = alignSeqs.begin(); itList != alignSeqs.end();) {
 146
 147                         //try to merge it with all smaller seqs
 148                         for (itList2 = alignSeqs.begin(); itList2 != alignSeqs.end();) {
 149
 150                                 if (m->control_pressed) { outFasta.close(); outNames.close(); remove(newFastaFile.c_str()); remove(newNamesFile.c_str());  return 0; }
 151
 152
 153                                 if (itList->seq.getName() != itList2->seq.getName()) { //you don't want to merge with yourself
 154                                         //are you within "diff" bases
 155
 156                                         int mismatch = calcMisMatches((*itList).seq.getAligned(), (*itList2).seq.getAligned());
 157
 158                                         if (mismatch <= diffs) {
 159                                                 //merge
 160                                                 (*itList).names += ',' + (*itList2).names;
 161                                                 (*itList).numIdentical += (*itList2).numIdentical;
 162
 163                                                 itList2 = alignSeqs.erase(itList2); //itList2--;
 164                                                 count++;
 165                                         }else{ itList2++; }
 166                                 }else{ itList2++; }
 167
 168                         }
 169
 170                         //ouptut this sequence
 171                         printData(outFasta, outNames, (*itList));
 172
 173                         //remove sequence
 174                         itList = alignSeqs.erase(itList);
 175
 176                         i++;
 177
 178                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 179                 }
 180
 181                 if(i % 100 != 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 182
 183                 outFasta.close();
 184                 outNames.close();
 185
 186                 if (m->control_pressed) {  remove(newFastaFile.c_str()); remove(newNamesFile.c_str());  return 0; }
 187
 188                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences.");
 189                 m->mothurOut("Total number of sequences before precluster was " + toString(numSeqs) + "."); m->mothurOutEndLine();
 190                 m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine();
 191
 192                 if (m->control_pressed) { remove(newFastaFile.c_str()); remove(newNamesFile.c_str()); return 0; }
 193
 194                 m->mothurOutEndLine();
 195                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 196                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 197                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 198                 m->mothurOutEndLine();
 199
 200                 return 0;
 201
 202         }
 203         catch(exception& e) {
 204                 m->errorOut(e, "PreClusterCommand", "execute");
 205                 exit(1);
 206         }
 207 }
 208
 209 /**************************************************************************************************/
 210
 211 //this requires the names and fasta file to be in the same order
 212
 213 int PreClusterCommand::readFASTA(){
 214         try {
 215                 //ifstream inNames;
 216                 ifstream inFasta;
 217
 218                 //openInputFile(namefile, inNames);
 219                 openInputFile(fastafile, inFasta);
 220
 221                 //string firstCol, secondCol, nameString;
 222                 length = 0;
 223
 224                 while (!inFasta.eof()) {
 225
 226                         if (m->control_pressed) { inFasta.close(); return 0; }
 227
 228                         //inNames >> firstCol >> secondCol;
 229                         //nameString = secondCol;
 230
 231                         //gobble(inNames);
 232                         //int size = 1;
 233                         //while (secondCol.find_first_of(',') != -1) {
 234                         //      size++;
 235                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 236                         //}
 237
 238                         Sequence seq(inFasta);  gobble(inFasta);
 239
 240                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 241                                 if (namefile != "") {
 242                                         itSize = sizes.find(seq.getName());
 243
 244                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 245                                         else{
 246                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 247                                                 alignSeqs.push_back(tempNode);
 248                                                 if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 249                                         }
 250                                 }else { //no names file, you are identical to yourself
 251                                         seqPNode tempNode(1, seq, seq.getName());
 252                                         alignSeqs.push_back(tempNode);
 253                                         if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 254                                 }
 255                         }
 256                 }
 257                 inFasta.close();
 258                 //inNames.close();
 259                 return alignSeqs.size();
 260         }
 261
 262         catch(exception& e) {
 263                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 264                 exit(1);
 265         }
 266 }
 267
 268 /**************************************************************************************************/
 269
 270 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 271         try {
 272                 int numBad = 0;
 273
 274                 for (int i = 0; i < seq1.length(); i++) {
 275                         //do they match
 276                         if (seq1[i] != seq2[i]) { numBad++; }
 277                         if (numBad > diffs) { return length;  } //to far to cluster
 278                 }
 279
 280                 return numBad;
 281         }
 282         catch(exception& e) {
 283                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 284                 exit(1);
 285         }
 286 }
 287
 288 /**************************************************************************************************/
 289
 290 void PreClusterCommand::printData(ofstream& outFasta, ofstream& outNames, seqPNode thisSeq){
 291         try {
 292                 thisSeq.seq.printSequence(outFasta);
 293                 outNames << thisSeq.seq.getName() << '\t' << thisSeq.names << endl;
 294         }
 295         catch(exception& e) {
 296                 m->errorOut(e, "PreClusterCommand", "printData");
 297                 exit(1);
 298         }
 299 }
 300 /**************************************************************************************************/
 301
 302 void PreClusterCommand::readNameFile(){
 303         try {
 304                 ifstream in;
 305                 openInputFile(namefile, in);
 306                 string firstCol, secondCol;
 307
 308                 while (!in.eof()) {
 309                         in >> firstCol >> secondCol; gobble(in);
 310                         names[firstCol] = secondCol;
 311                         int size = 1;
 312
 313                         for(int i=0;i<secondCol.size();i++){
 314                                 if(secondCol[i] == ','){        size++; }
 315                         }
 316                         sizes[firstCol] = size;
 317                 }
 318                 in.close();
 319         }
 320         catch(exception& e) {
 321                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 322                 exit(1);
 323         }
 324 }
 325
 326 /**************************************************************************************************/
 327
 328