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