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                 //reads fasta file and return number of seqs
 119                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 120
 121                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 122                 if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->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                 m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 171                 m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine();
 172                 printData(newFastaFile, newNamesFile);
 173
 174                 m->mothurOutEndLine();
 175                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 176                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 177                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 178                 m->mothurOutEndLine();
 179
 180                 return 0;
 181
 182         }
 183         catch(exception& e) {
 184                 m->errorOut(e, "PreClusterCommand", "execute");
 185                 exit(1);
 186         }
 187 }
 188
 189 /**************************************************************************************************/
 190
 191 //this requires the names and fasta file to be in the same order
 192
 193 int PreClusterCommand::readFASTA(){
 194         try {
 195                 //ifstream inNames;
 196                 ifstream inFasta;
 197
 198                 //openInputFile(namefile, inNames);
 199                 openInputFile(fastafile, inFasta);
 200
 201                 //string firstCol, secondCol, nameString;
 202                 length = 0;
 203
 204                 while (!inFasta.eof()) {
 205
 206                         //inNames >> firstCol >> secondCol;
 207                         //nameString = secondCol;
 208
 209                         //gobble(inNames);
 210                         //int size = 1;
 211                         //while (secondCol.find_first_of(',') != -1) {
 212                         //      size++;
 213                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 214                         //}
 215
 216                         Sequence seq(inFasta);  gobble(inFasta);
 217
 218                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 219                                 if (namefile != "") {
 220                                         itSize = sizes.find(seq.getName());
 221
 222                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 223                                         else{
 224                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 225                                                 alignSeqs.push_back(tempNode);
 226                                                 if (seq.getAligned().length() > length) {  length = alignSeqs[0].seq.getAligned().length();  }
 227                                         }
 228                                 }else { //no names file, you are identical to yourself
 229                                         seqPNode tempNode(1, seq, seq.getName());
 230                                         alignSeqs.push_back(tempNode);
 231                                         if (seq.getAligned().length() > length) {  length = alignSeqs[0].seq.getAligned().length();  }
 232                                 }
 233                         }
 234                 }
 235                 inFasta.close();
 236                 //inNames.close();
 237                 return alignSeqs.size();
 238         }
 239
 240         catch(exception& e) {
 241                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 242                 exit(1);
 243         }
 244 }
 245
 246 /**************************************************************************************************/
 247
 248 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 249         try {
 250                 int numBad = 0;
 251
 252                 for (int i = 0; i < seq1.length(); i++) {
 253                         //do they match
 254                         if (seq1[i] != seq2[i]) { numBad++; }
 255                         if (numBad > diffs) { return length;  } //to far to cluster
 256                 }
 257
 258                 return numBad;
 259         }
 260         catch(exception& e) {
 261                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 262                 exit(1);
 263         }
 264 }
 265
 266 /**************************************************************************************************/
 267
 268 void PreClusterCommand::printData(string newfasta, string newname){
 269         try {
 270                 ofstream outFasta;
 271                 ofstream outNames;
 272
 273                 openOutputFile(newfasta, outFasta);
 274                 openOutputFile(newname, outNames);
 275
 276
 277                 for (int i = 0; i < alignSeqs.size(); i++) {
 278                         if (alignSeqs[i].numIdentical != 0) {
 279                                 alignSeqs[i].seq.printSequence(outFasta);
 280                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 281                         }
 282                 }
 283
 284                 outFasta.close();
 285                 outNames.close();
 286
 287         }
 288         catch(exception& e) {
 289                 m->errorOut(e, "PreClusterCommand", "printData");
 290                 exit(1);
 291         }
 292 }
 293
 294 /**************************************************************************************************/
 295
 296 void PreClusterCommand::readNameFile(){
 297         try {
 298                 ifstream in;
 299                 openInputFile(namefile, in);
 300                 string firstCol, secondCol;
 301
 302                 while (!in.eof()) {
 303                         in >> firstCol >> secondCol; gobble(in);
 304                         names[firstCol] = secondCol;
 305                         int size = 1;
 306                         while (secondCol.find_first_of(',') != -1) {
 307                                 size++;
 308                                 secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 309                         }
 310                         sizes[firstCol] = size;
 311                 }
 312                 in.close();
 313         }
 314         catch(exception& e) {
 315                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 316                 exit(1);
 317         }
 318 }
 319
 320 /**************************************************************************************************/
 321
 322