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