clusterfragmentscommand.cpp

   1 /*
   2  *  ryanscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 9/23/10.
   6  *  Copyright 2010 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "clusterfragmentscommand.h"
  11
  12 //**********************************************************************************************************************
  13 //sort by unaligned
  14 inline bool comparePriority(seqRNode first, seqRNode second) {
  15         bool better = false;
  16
  17         if (first.length > second.length) {
  18                 better = true;
  19         }else if (first.length == second.length) {
  20                 if (first.numIdentical > second.numIdentical) {
  21                         better = true;
  22                 }
  23         }
  24
  25         return better;
  26 }
  27 //**********************************************************************************************************************
  28 ClusterFragmentsCommand::ClusterFragmentsCommand(string option) {
  29         try {
  30                 abort = false;
  31
  32                 //allow user to run help
  33                 if(option == "help") { help(); abort = true; }
  34
  35                 else {
  36                         //valid paramters for this command
  37                         string Array[] =  {"fasta","name","outputdir","inputdir"};
  38                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  39
  40                         OptionParser parser(option);
  41                         map<string, string> parameters = parser.getParameters();
  42
  43                         ValidParameters validParameter;
  44                         map<string, string>::iterator it;
  45
  46                         //check to make sure all parameters are valid for command
  47                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  48                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  49                         }
  50
  51                         //if the user changes the input directory command factory will send this info to us in the output parameter
  52                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
  53                         if (inputDir == "not found"){   inputDir = "";          }
  54                         else {
  55                                 string path;
  56                                 it = parameters.find("fasta");
  57                                 //user has given a template file
  58                                 if(it != parameters.end()){
  59                                         path = m->hasPath(it->second);
  60                                         //if the user has not given a path then, add inputdir. else leave path alone.
  61                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
  62                                 }
  63
  64                                 it = parameters.find("name");
  65                                 //user has given a template file
  66                                 if(it != parameters.end()){
  67                                         path = m->hasPath(it->second);
  68                                         //if the user has not given a path then, add inputdir. else leave path alone.
  69                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
  70                                 }
  71                         }
  72
  73                         //check for required parameters
  74                         fastafile = validParameter.validFile(parameters, "fasta", true);
  75                         if (fastafile == "not found") { m->mothurOut("fasta is a required parameter for the cluster.fragments command."); m->mothurOutEndLine(); abort = true; }
  76                         else if (fastafile == "not open") { abort = true; }
  77
  78                         //if the user changes the output directory command factory will send this info to us in the output parameter
  79                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = m->hasPath(fastafile);      }
  80
  81                         //check for optional parameter and set defaults
  82                         // ...at some point should added some additional type checking...
  83                         namefile = validParameter.validFile(parameters, "name", true);
  84                         if (namefile == "not found") { namefile =  "";  }
  85                         else if (namefile == "not open") { abort = true; }
  86                         else {  readNameFile();  }
  87                 }
  88
  89         }
  90         catch(exception& e) {
  91                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
  92                 exit(1);
  93         }
  94 }
  95
  96 //**********************************************************************************************************************
  97 ClusterFragmentsCommand::~ClusterFragmentsCommand(){}
  98 //**********************************************************************************************************************
  99 void ClusterFragmentsCommand::help(){
 100         try {
 101                 m->mothurOut("The cluster.fragments command groups sequences that are part of a larger sequence.\n");
 102                 m->mothurOut("The cluster.fragments command outputs a new fasta and name file.\n");
 103                 m->mothurOut("The cluster.fragments command parameters are fasta and name. The fasta parameter is required. \n");
 104                 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");
 105                 m->mothurOut("The cluster.fragments command should be in the following format: \n");
 106                 m->mothurOut("cluster.fragments(fasta=yourFastaFile, names=yourNamesFile) \n");
 107                 m->mothurOut("Example cluster.fragments(fasta=amazon.fasta).\n");
 108                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
 109         }
 110         catch(exception& e) {
 111                 m->errorOut(e, "ClusterFragmentsCommand", "help");
 112                 exit(1);
 113         }
 114 }
 115 //**********************************************************************************************************************
 116 int ClusterFragmentsCommand::execute(){
 117         try {
 118
 119                 if (abort == true) { return 0; }
 120
 121                 int start = time(NULL);
 122
 123                 //reads fasta file and return number of seqs
 124                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 125
 126                 if (m->control_pressed) { return 0; }
 127
 128                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 129
 130                 //sort seqs by length of unaligned sequence
 131                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 132
 133                 int count = 0;
 134
 135                 //think about running through twice...
 136                 for (int i = 0; i < numSeqs; i++) {
 137
 138                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 139
 140                                 string iBases = alignSeqs[i].seq.getUnaligned();
 141
 142                                 //try to merge it with all smaller seqs
 143                                 for (int j = i+1; j < numSeqs; j++) {
 144
 145                                         if (m->control_pressed) { return 0; }
 146
 147                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 148
 149                                                 string jBases = alignSeqs[j].seq.getUnaligned();
 150
 151                                                 int pos = iBases.find(jBases);
 152
 153                                                 if (pos != string::npos) {
 154                                                         //merge
 155                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 156                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 157
 158                                                         alignSeqs[j].active = 0;
 159                                                         alignSeqs[j].numIdentical = 0;
 160                                                         count++;
 161                                                 }
 162                                         }//end if j active
 163                                 }//end if i != j
 164
 165                                 //remove from active list
 166                                 alignSeqs[i].active = 0;
 167
 168                         }//end if active i
 169                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 170                 }
 171
 172                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 173
 174
 175                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 176
 177                 string newFastaFile = fileroot + "fragclust.fasta";
 178                 string newNamesFile = fileroot + "names";
 179
 180                 if (m->control_pressed) { return 0; }
 181
 182                 m->mothurOutEndLine();
 183                 m->mothurOut("Total number of sequences before cluster.fragments was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 184                 m->mothurOut("cluster.fragments removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 185
 186                 printData(newFastaFile, newNamesFile);
 187
 188                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 189
 190                 if (m->control_pressed) { remove(newFastaFile.c_str()); remove(newNamesFile.c_str()); return 0; }
 191
 192                 m->mothurOutEndLine();
 193                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 194                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 195                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 196                 m->mothurOutEndLine();
 197
 198                 return 0;
 199
 200         }
 201         catch(exception& e) {
 202                 m->errorOut(e, "ClusterFragmentsCommand", "execute");
 203                 exit(1);
 204         }
 205 }
 206
 207 /**************************************************************************************************/
 208 int ClusterFragmentsCommand::readFASTA(){
 209         try {
 210
 211                 ifstream inFasta;
 212                 m->openInputFile(fastafile, inFasta);
 213
 214                 while (!inFasta.eof()) {
 215
 216                         if (m->control_pressed) { inFasta.close(); return 0; }
 217
 218                         Sequence seq(inFasta);  m->gobble(inFasta);
 219
 220                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 221                                 if (namefile != "") {
 222                                         itSize = sizes.find(seq.getName());
 223
 224                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 225                                         else{
 226                                                 seqRNode tempNode(itSize->second, seq, names[seq.getName()], seq.getUnaligned().length());
 227                                                 alignSeqs.push_back(tempNode);
 228                                         }
 229                                 }else { //no names file, you are identical to yourself
 230                                         seqRNode tempNode(1, seq, seq.getName(), seq.getUnaligned().length());
 231                                         alignSeqs.push_back(tempNode);
 232                                 }
 233                         }
 234                 }
 235
 236                 inFasta.close();
 237                 return alignSeqs.size();
 238         }
 239
 240         catch(exception& e) {
 241                 m->errorOut(e, "ClusterFragmentsCommand", "readFASTA");
 242                 exit(1);
 243         }
 244 }
 245 /**************************************************************************************************/
 246 void ClusterFragmentsCommand::printData(string newfasta, string newname){
 247         try {
 248                 ofstream outFasta;
 249                 ofstream outNames;
 250
 251                 m->openOutputFile(newfasta, outFasta);
 252                 m->openOutputFile(newname, outNames);
 253
 254                 for (int i = 0; i < alignSeqs.size(); i++) {
 255                         if (alignSeqs[i].numIdentical != 0) {
 256                                 alignSeqs[i].seq.printSequence(outFasta);
 257                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 258                         }
 259                 }
 260
 261                 outFasta.close();
 262                 outNames.close();
 263         }
 264         catch(exception& e) {
 265                 m->errorOut(e, "ClusterFragmentsCommand", "printData");
 266                 exit(1);
 267         }
 268 }
 269 /**************************************************************************************************/
 270
 271 void ClusterFragmentsCommand::readNameFile(){
 272         try {
 273                 ifstream in;
 274                 m->openInputFile(namefile, in);
 275                 string firstCol, secondCol;
 276
 277                 while (!in.eof()) {
 278                         in >> firstCol >> secondCol; m->gobble(in);
 279                         names[firstCol] = secondCol;
 280                         int size = 1;
 281
 282                         for(int i=0;i<secondCol.size();i++){
 283                                 if(secondCol[i] == ','){        size++; }
 284                         }
 285                         sizes[firstCol] = size;
 286                 }
 287                 in.close();
 288         }
 289         catch(exception& e) {
 290                 m->errorOut(e, "ClusterFragmentsCommand", "readNameFile");
 291                 exit(1);
 292         }
 293 }
 294
 295 /**************************************************************************************************/
 296