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 vector<string> ClusterFragmentsCommand::getValidParameters(){
  29         try {
  30                 string AlignArray[] =  {"fasta","name","outputdir","inputdir"};
  31                 vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  32                 return myArray;
  33         }
  34         catch(exception& e) {
  35                 m->errorOut(e, "ClusterFragmentsCommand", "getValidParameters");
  36                 exit(1);
  37         }
  38 }
  39 //**********************************************************************************************************************
  40 ClusterFragmentsCommand::ClusterFragmentsCommand(){
  41         try {
  42                 abort = true;
  43                 //initialize outputTypes
  44                 vector<string> tempOutNames;
  45                 outputTypes["fasta"] = tempOutNames;
  46                 outputTypes["name"] = tempOutNames;
  47         }
  48         catch(exception& e) {
  49                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
  50                 exit(1);
  51         }
  52 }
  53 //**********************************************************************************************************************
  54 vector<string> ClusterFragmentsCommand::getRequiredParameters(){
  55         try {
  56                 string Array[] =  {"fasta"};
  57                 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  58                 return myArray;
  59         }
  60         catch(exception& e) {
  61                 m->errorOut(e, "ClusterFragmentsCommand", "getRequiredParameters");
  62                 exit(1);
  63         }
  64 }
  65 //**********************************************************************************************************************
  66 vector<string> ClusterFragmentsCommand::getRequiredFiles(){
  67         try {
  68                 vector<string> myArray;
  69                 return myArray;
  70         }
  71         catch(exception& e) {
  72                 m->errorOut(e, "ClusterFragmentsCommand", "getRequiredFiles");
  73                 exit(1);
  74         }
  75 }
  76 //**********************************************************************************************************************
  77 ClusterFragmentsCommand::ClusterFragmentsCommand(string option) {
  78         try {
  79                 abort = false;
  80
  81                 //allow user to run help
  82                 if(option == "help") { help(); abort = true; }
  83
  84                 else {
  85                         //valid paramters for this command
  86                         string Array[] =  {"fasta","name","outputdir","inputdir"};
  87                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  88
  89                         OptionParser parser(option);
  90                         map<string, string> parameters = parser.getParameters();
  91
  92                         ValidParameters validParameter;
  93                         map<string, string>::iterator it;
  94
  95                         //check to make sure all parameters are valid for command
  96                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  97                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  98                         }
  99
 100                         //initialize outputTypes
 101                         vector<string> tempOutNames;
 102                         outputTypes["fasta"] = tempOutNames;
 103                         outputTypes["name"] = tempOutNames;
 104
 105                         //if the user changes the input directory command factory will send this info to us in the output parameter
 106                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 107                         if (inputDir == "not found"){   inputDir = "";          }
 108                         else {
 109                                 string path;
 110                                 it = parameters.find("fasta");
 111                                 //user has given a template file
 112                                 if(it != parameters.end()){
 113                                         path = m->hasPath(it->second);
 114                                         //if the user has not given a path then, add inputdir. else leave path alone.
 115                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 116                                 }
 117
 118                                 it = parameters.find("name");
 119                                 //user has given a template file
 120                                 if(it != parameters.end()){
 121                                         path = m->hasPath(it->second);
 122                                         //if the user has not given a path then, add inputdir. else leave path alone.
 123                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 124                                 }
 125                         }
 126
 127                         //check for required parameters
 128                         fastafile = validParameter.validFile(parameters, "fasta", true);
 129                         if (fastafile == "not found") { m->mothurOut("fasta is a required parameter for the cluster.fragments command."); m->mothurOutEndLine(); abort = true; }
 130                         else if (fastafile == "not open") { abort = true; }
 131
 132                         //if the user changes the output directory command factory will send this info to us in the output parameter
 133                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = m->hasPath(fastafile);      }
 134
 135                         //check for optional parameter and set defaults
 136                         // ...at some point should added some additional type checking...
 137                         namefile = validParameter.validFile(parameters, "name", true);
 138                         if (namefile == "not found") { namefile =  "";  }
 139                         else if (namefile == "not open") { abort = true; }
 140                         else {  readNameFile();  }
 141                 }
 142
 143         }
 144         catch(exception& e) {
 145                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
 146                 exit(1);
 147         }
 148 }
 149
 150 //**********************************************************************************************************************
 151 ClusterFragmentsCommand::~ClusterFragmentsCommand(){}
 152 //**********************************************************************************************************************
 153 void ClusterFragmentsCommand::help(){
 154         try {
 155                 m->mothurOut("The cluster.fragments command groups sequences that are part of a larger sequence.\n");
 156                 m->mothurOut("The cluster.fragments command outputs a new fasta and name file.\n");
 157                 m->mothurOut("The cluster.fragments command parameters are fasta and name. The fasta parameter is required. \n");
 158                 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");
 159                 m->mothurOut("The cluster.fragments command should be in the following format: \n");
 160                 m->mothurOut("cluster.fragments(fasta=yourFastaFile, names=yourNamesFile) \n");
 161                 m->mothurOut("Example cluster.fragments(fasta=amazon.fasta).\n");
 162                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
 163         }
 164         catch(exception& e) {
 165                 m->errorOut(e, "ClusterFragmentsCommand", "help");
 166                 exit(1);
 167         }
 168 }
 169 //**********************************************************************************************************************
 170 int ClusterFragmentsCommand::execute(){
 171         try {
 172
 173                 if (abort == true) { return 0; }
 174
 175                 int start = time(NULL);
 176
 177                 //reads fasta file and return number of seqs
 178                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 179
 180                 if (m->control_pressed) { return 0; }
 181
 182                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 183
 184                 //sort seqs by length of unaligned sequence
 185                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 186
 187                 int count = 0;
 188
 189                 //think about running through twice...
 190                 for (int i = 0; i < numSeqs; i++) {
 191
 192                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 193
 194                                 string iBases = alignSeqs[i].seq.getUnaligned();
 195
 196                                 //try to merge it with all smaller seqs
 197                                 for (int j = i+1; j < numSeqs; j++) {
 198
 199                                         if (m->control_pressed) { return 0; }
 200
 201                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 202
 203                                                 string jBases = alignSeqs[j].seq.getUnaligned();
 204
 205                                                 int pos = iBases.find(jBases);
 206
 207                                                 if (pos != string::npos) {
 208                                                         //merge
 209                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 210                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 211
 212                                                         alignSeqs[j].active = 0;
 213                                                         alignSeqs[j].numIdentical = 0;
 214                                                         count++;
 215                                                 }
 216                                         }//end if j active
 217                                 }//end if i != j
 218
 219                                 //remove from active list
 220                                 alignSeqs[i].active = 0;
 221
 222                         }//end if active i
 223                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 224                 }
 225
 226                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 227
 228
 229                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 230
 231                 string newFastaFile = fileroot + "fragclust.fasta";
 232                 string newNamesFile = fileroot + "names";
 233
 234                 if (m->control_pressed) { return 0; }
 235
 236                 m->mothurOutEndLine();
 237                 m->mothurOut("Total number of sequences before cluster.fragments was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 238                 m->mothurOut("cluster.fragments removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 239
 240                 printData(newFastaFile, newNamesFile);
 241
 242                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 243
 244                 if (m->control_pressed) { remove(newFastaFile.c_str()); remove(newNamesFile.c_str()); return 0; }
 245
 246                 m->mothurOutEndLine();
 247                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 248                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 249                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 250                 outputNames.push_back(newFastaFile);  outputNames.push_back(newNamesFile); outputTypes["fasta"].push_back(newFastaFile); outputTypes["name"].push_back(newNamesFile);
 251                 m->mothurOutEndLine();
 252
 253                 return 0;
 254
 255         }
 256         catch(exception& e) {
 257                 m->errorOut(e, "ClusterFragmentsCommand", "execute");
 258                 exit(1);
 259         }
 260 }
 261
 262 /**************************************************************************************************/
 263 int ClusterFragmentsCommand::readFASTA(){
 264         try {
 265
 266                 ifstream inFasta;
 267                 m->openInputFile(fastafile, inFasta);
 268
 269                 while (!inFasta.eof()) {
 270
 271                         if (m->control_pressed) { inFasta.close(); return 0; }
 272
 273                         Sequence seq(inFasta);  m->gobble(inFasta);
 274
 275                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 276                                 if (namefile != "") {
 277                                         itSize = sizes.find(seq.getName());
 278
 279                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 280                                         else{
 281                                                 seqRNode tempNode(itSize->second, seq, names[seq.getName()], seq.getUnaligned().length());
 282                                                 alignSeqs.push_back(tempNode);
 283                                         }
 284                                 }else { //no names file, you are identical to yourself
 285                                         seqRNode tempNode(1, seq, seq.getName(), seq.getUnaligned().length());
 286                                         alignSeqs.push_back(tempNode);
 287                                 }
 288                         }
 289                 }
 290
 291                 inFasta.close();
 292                 return alignSeqs.size();
 293         }
 294
 295         catch(exception& e) {
 296                 m->errorOut(e, "ClusterFragmentsCommand", "readFASTA");
 297                 exit(1);
 298         }
 299 }
 300 /**************************************************************************************************/
 301 void ClusterFragmentsCommand::printData(string newfasta, string newname){
 302         try {
 303                 ofstream outFasta;
 304                 ofstream outNames;
 305
 306                 m->openOutputFile(newfasta, outFasta);
 307                 m->openOutputFile(newname, outNames);
 308
 309                 for (int i = 0; i < alignSeqs.size(); i++) {
 310                         if (alignSeqs[i].numIdentical != 0) {
 311                                 alignSeqs[i].seq.printSequence(outFasta);
 312                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 313                         }
 314                 }
 315
 316                 outFasta.close();
 317                 outNames.close();
 318         }
 319         catch(exception& e) {
 320                 m->errorOut(e, "ClusterFragmentsCommand", "printData");
 321                 exit(1);
 322         }
 323 }
 324 /**************************************************************************************************/
 325
 326 void ClusterFragmentsCommand::readNameFile(){
 327         try {
 328                 ifstream in;
 329                 m->openInputFile(namefile, in);
 330                 string firstCol, secondCol;
 331
 332                 while (!in.eof()) {
 333                         in >> firstCol >> secondCol; m->gobble(in);
 334                         names[firstCol] = secondCol;
 335                         int size = 1;
 336
 337                         for(int i=0;i<secondCol.size();i++){
 338                                 if(secondCol[i] == ','){        size++; }
 339                         }
 340                         sizes[firstCol] = size;
 341                 }
 342                 in.close();
 343         }
 344         catch(exception& e) {
 345                 m->errorOut(e, "ClusterFragmentsCommand", "readNameFile");
 346                 exit(1);
 347         }
 348 }
 349
 350 /**************************************************************************************************/
 351