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 #include "needlemanoverlap.hpp"
  12
  13 //**********************************************************************************************************************
  14 //sort by unaligned
  15 inline bool comparePriority(seqRNode first, seqRNode second) {
  16         bool better = false;
  17
  18         if (first.length > second.length) {
  19                 better = true;
  20         }else if (first.length == second.length) {
  21                 if (first.numIdentical > second.numIdentical) {
  22                         better = true;
  23                 }
  24         }
  25
  26         return better;
  27 }
  28 //**********************************************************************************************************************
  29 vector<string> ClusterFragmentsCommand::setParameters(){
  30         try {
  31                 CommandParameter pfasta("fasta", "InputTypes", "", "", "none", "none", "none",false,true); parameters.push_back(pfasta);
  32                 CommandParameter pname("name", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pname);
  33                 CommandParameter pdiffs("diffs", "Number", "", "0", "", "", "",false,false); parameters.push_back(pdiffs);
  34                 CommandParameter ppercent("percent", "Number", "", "0", "", "", "",false,false); parameters.push_back(ppercent);
  35                 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "",false,false); parameters.push_back(pinputdir);
  36                 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "",false,false); parameters.push_back(poutputdir);
  37
  38                 vector<string> myArray;
  39                 for (int i = 0; i < parameters.size(); i++) {   myArray.push_back(parameters[i].name);          }
  40                 return myArray;
  41         }
  42         catch(exception& e) {
  43                 m->errorOut(e, "ClusterFragmentsCommand", "setParameters");
  44                 exit(1);
  45         }
  46 }
  47 //**********************************************************************************************************************
  48 string ClusterFragmentsCommand::getHelpString(){
  49         try {
  50                 string helpString = "";
  51                 helpString += "The cluster.fragments command groups sequences that are part of a larger sequence.\n";
  52                 helpString += "The cluster.fragments command outputs a new fasta and name file.\n";
  53                 helpString += "The cluster.fragments command parameters are fasta, name, diffs and percent. The fasta parameter is required, unless you have a valid current file. \n";
  54                 helpString += "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";
  55                 helpString += "The diffs parameter allows you to set the number of differences allowed, default=0. \n";
  56                 helpString += "The percent parameter allows you to set percentage of differences allowed, default=0. percent=2 means if the number of difference is less than or equal to two percent of the length of the fragment, then cluster.\n";
  57                 helpString += "You may use diffs and percent at the same time to say something like: If the number or differences is greater than 1 or more than 2% of the fragment length, don't merge. \n";
  58                 helpString += "The cluster.fragments command should be in the following format: \n";
  59                 helpString += "cluster.fragments(fasta=yourFastaFile, names=yourNamesFile) \n";
  60                 helpString += "Example cluster.fragments(fasta=amazon.fasta).\n";
  61                 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
  62                 return helpString;
  63         }
  64         catch(exception& e) {
  65                 m->errorOut(e, "ClusterFragmentsCommand", "getHelpString");
  66                 exit(1);
  67         }
  68 }
  69 //**********************************************************************************************************************
  70 ClusterFragmentsCommand::ClusterFragmentsCommand(){
  71         try {
  72                 abort = true; calledHelp = true;
  73                 setParameters();
  74                 vector<string> tempOutNames;
  75                 outputTypes["fasta"] = tempOutNames;
  76                 outputTypes["name"] = tempOutNames;
  77         }
  78         catch(exception& e) {
  79                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
  80                 exit(1);
  81         }
  82 }
  83 //**********************************************************************************************************************
  84 ClusterFragmentsCommand::ClusterFragmentsCommand(string option) {
  85         try {
  86                 abort = false; calledHelp = false;
  87
  88                 //allow user to run help
  89                 if(option == "help") { help(); abort = true; calledHelp = true; }
  90                 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
  91
  92                 else {
  93                         vector<string> myArray = setParameters();
  94
  95                         OptionParser parser(option);
  96                         map<string, string> parameters = parser.getParameters();
  97
  98                         ValidParameters validParameter;
  99                         map<string, string>::iterator it;
 100
 101                         //check to make sure all parameters are valid for command
 102                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
 103                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
 104                         }
 105
 106                         //initialize outputTypes
 107                         vector<string> tempOutNames;
 108                         outputTypes["fasta"] = tempOutNames;
 109                         outputTypes["name"] = tempOutNames;
 110
 111                         //if the user changes the input directory command factory will send this info to us in the output parameter
 112                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 113                         if (inputDir == "not found"){   inputDir = "";          }
 114                         else {
 115                                 string path;
 116                                 it = parameters.find("fasta");
 117                                 //user has given a template file
 118                                 if(it != parameters.end()){
 119                                         path = m->hasPath(it->second);
 120                                         //if the user has not given a path then, add inputdir. else leave path alone.
 121                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 122                                 }
 123
 124                                 it = parameters.find("name");
 125                                 //user has given a template file
 126                                 if(it != parameters.end()){
 127                                         path = m->hasPath(it->second);
 128                                         //if the user has not given a path then, add inputdir. else leave path alone.
 129                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 130                                 }
 131                         }
 132
 133                         //check for required parameters
 134                         fastafile = validParameter.validFile(parameters, "fasta", true);
 135                         if (fastafile == "not found") {
 136                                 fastafile = m->getFastaFile();
 137                                 if (fastafile != "") { m->mothurOut("Using " + fastafile + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
 138                                 else {  m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
 139                         }
 140                         else if (fastafile == "not open") { fastafile = ""; abort = true; }
 141                         else { m->setFastaFile(fastafile); }
 142
 143                         //if the user changes the output directory command factory will send this info to us in the output parameter
 144                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = m->hasPath(fastafile);      }
 145
 146                         //check for optional parameter and set defaults
 147                         // ...at some point should added some additional type checking...
 148                         namefile = validParameter.validFile(parameters, "name", true);
 149                         if (namefile == "not found") { namefile =  "";  }
 150                         else if (namefile == "not open") { namefile = ""; abort = true; }
 151                         else {  readNameFile(); m->setNameFile(namefile); }
 152
 153                         string temp;
 154                         temp = validParameter.validFile(parameters, "diffs", false);            if (temp == "not found"){       temp = "0";                             }
 155                         m->mothurConvert(temp, diffs);
 156
 157                         temp = validParameter.validFile(parameters, "percent", false);          if (temp == "not found"){       temp = "0";                             }
 158                         m->mothurConvert(temp, percent);
 159
 160                         if (namefile == "") {
 161                                 vector<string> files; files.push_back(fastafile);
 162                                 parser.getNameFile(files);
 163                         }
 164
 165                 }
 166
 167         }
 168         catch(exception& e) {
 169                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
 170                 exit(1);
 171         }
 172 }
 173 //**********************************************************************************************************************
 174 int ClusterFragmentsCommand::execute(){
 175         try {
 176
 177                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 178
 179                 int start = time(NULL);
 180
 181                 //reads fasta file and return number of seqs
 182                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 183
 184                 if (m->control_pressed) { return 0; }
 185
 186                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 187
 188                 //sort seqs by length of unaligned sequence
 189                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 190
 191                 int count = 0;
 192
 193                 //think about running through twice...
 194                 for (int i = 0; i < numSeqs; i++) {
 195
 196                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 197
 198                                 string iBases = alignSeqs[i].seq.getUnaligned();
 199
 200                                 //try to merge it with all smaller seqs
 201                                 for (int j = i+1; j < numSeqs; j++) {
 202
 203                                         if (m->control_pressed) { return 0; }
 204
 205                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 206
 207                                                 string jBases = alignSeqs[j].seq.getUnaligned();
 208
 209                                                 if (isFragment(iBases, jBases)) {
 210                                                         //merge
 211                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 212                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 213
 214                                                         alignSeqs[j].active = 0;
 215                                                         alignSeqs[j].numIdentical = 0;
 216                                                         count++;
 217                                                 }
 218                                         }//end if j active
 219                                 }//end if i != j
 220
 221                                 //remove from active list
 222                                 alignSeqs[i].active = 0;
 223
 224                         }//end if active i
 225                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 226                 }
 227
 228                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 229
 230
 231                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 232
 233                 string newFastaFile = fileroot + "fragclust.fasta";
 234                 string newNamesFile = fileroot + "fragclust.names";
 235
 236                 if (m->control_pressed) { return 0; }
 237
 238                 m->mothurOutEndLine();
 239                 m->mothurOut("Total number of sequences before cluster.fragments was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 240                 m->mothurOut("cluster.fragments removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 241
 242                 printData(newFastaFile, newNamesFile);
 243
 244                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 245
 246                 if (m->control_pressed) { m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 247
 248                 m->mothurOutEndLine();
 249                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 250                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 251                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 252                 outputNames.push_back(newFastaFile);  outputNames.push_back(newNamesFile); outputTypes["fasta"].push_back(newFastaFile); outputTypes["name"].push_back(newNamesFile);
 253                 m->mothurOutEndLine();
 254
 255                 //set fasta file as new current fastafile
 256                 string current = "";
 257                 itTypes = outputTypes.find("fasta");
 258                 if (itTypes != outputTypes.end()) {
 259                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 260                 }
 261
 262                 itTypes = outputTypes.find("name");
 263                 if (itTypes != outputTypes.end()) {
 264                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 265                 }
 266
 267                 return 0;
 268
 269         }
 270         catch(exception& e) {
 271                 m->errorOut(e, "ClusterFragmentsCommand", "execute");
 272                 exit(1);
 273         }
 274 }
 275 //***************************************************************************************************************
 276 bool ClusterFragmentsCommand::isFragment(string seq1, string seq2){
 277         try {
 278                 bool fragment = false;
 279
 280                 //exact match
 281                 int pos = seq1.find(seq2);
 282                 if (pos != string::npos) { return true; }
 283                 //no match, no diffs wanted
 284                 else if ((diffs == 0) && (percent == 0)) { return false; }
 285                 else { //try aligning and see if you can find it
 286
 287                         //find number of acceptable differences for this sequence fragment
 288                         int totalDiffs;
 289                         if (diffs == 0) { //you didnt set diffs you want a percentage
 290                                 totalDiffs = floor((seq2.length() * (percent / 100.0)));
 291                         }else if (percent == 0) { //you didn't set percent you want diffs
 292                                 totalDiffs = diffs;
 293                         }else if ((percent != 0) && (diffs != 0)) { //you want both, set total diffs to smaller of 2
 294                                 totalDiffs = diffs;
 295                                 int percentDiff = floor((seq2.length() * (percent / 100.0)));
 296                                 if (percentDiff < totalDiffs) { totalDiffs = percentDiff; }
 297                         }
 298
 299                         Alignment* alignment = new NeedlemanOverlap(-1.0, 1.0, -1.0, (seq1.length()+totalDiffs+1));
 300
 301                         //use needleman to align
 302                         alignment->align(seq2, seq1);
 303                         string tempSeq2 = alignment->getSeqAAln();
 304                         string temp = alignment->getSeqBAln();
 305
 306                         delete alignment;
 307
 308                         //chop gap ends
 309                         int startPos = 0;
 310                         int endPos = tempSeq2.length()-1;
 311                         for (int i = 0; i < tempSeq2.length(); i++) {  if (isalpha(tempSeq2[i])) { startPos = i; break; } }
 312                         for (int i = tempSeq2.length()-1; i >= 0; i--) {  if (isalpha(tempSeq2[i])) { endPos = i; break; } }
 313
 314                         //count number of diffs
 315                         int numDiffs = 0;
 316                         for (int i = startPos; i <= endPos; i++) {
 317                                 if (tempSeq2[i] != temp[i]) { numDiffs++; }
 318                         }
 319
 320                         if (numDiffs <= totalDiffs) { fragment = true; }
 321
 322                 }
 323
 324                 return fragment;
 325
 326         }
 327         catch(exception& e) {
 328                 m->errorOut(e, "ClusterFragmentsCommand", "isFragment");
 329                 exit(1);
 330         }
 331 }
 332 /**************************************************************************************************/
 333 int ClusterFragmentsCommand::readFASTA(){
 334         try {
 335
 336                 ifstream inFasta;
 337                 m->openInputFile(fastafile, inFasta);
 338
 339                 while (!inFasta.eof()) {
 340
 341                         if (m->control_pressed) { inFasta.close(); return 0; }
 342
 343                         Sequence seq(inFasta);  m->gobble(inFasta);
 344
 345                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 346                                 if (namefile != "") {
 347                                         itSize = sizes.find(seq.getName());
 348
 349                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 350                                         else{
 351                                                 seqRNode tempNode(itSize->second, seq, names[seq.getName()], seq.getUnaligned().length());
 352                                                 alignSeqs.push_back(tempNode);
 353                                         }
 354                                 }else { //no names file, you are identical to yourself
 355                                         seqRNode tempNode(1, seq, seq.getName(), seq.getUnaligned().length());
 356                                         alignSeqs.push_back(tempNode);
 357                                 }
 358                         }
 359                 }
 360
 361                 inFasta.close();
 362                 return alignSeqs.size();
 363         }
 364
 365         catch(exception& e) {
 366                 m->errorOut(e, "ClusterFragmentsCommand", "readFASTA");
 367                 exit(1);
 368         }
 369 }
 370 /**************************************************************************************************/
 371 void ClusterFragmentsCommand::printData(string newfasta, string newname){
 372         try {
 373                 ofstream outFasta;
 374                 ofstream outNames;
 375
 376                 m->openOutputFile(newfasta, outFasta);
 377                 m->openOutputFile(newname, outNames);
 378
 379                 for (int i = 0; i < alignSeqs.size(); i++) {
 380                         if (alignSeqs[i].numIdentical != 0) {
 381                                 alignSeqs[i].seq.printSequence(outFasta);
 382                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 383                         }
 384                 }
 385
 386                 outFasta.close();
 387                 outNames.close();
 388         }
 389         catch(exception& e) {
 390                 m->errorOut(e, "ClusterFragmentsCommand", "printData");
 391                 exit(1);
 392         }
 393 }
 394 /**************************************************************************************************/
 395
 396 void ClusterFragmentsCommand::readNameFile(){
 397         try {
 398                 ifstream in;
 399                 m->openInputFile(namefile, in);
 400                 string firstCol, secondCol;
 401
 402                 while (!in.eof()) {
 403                         in >> firstCol >> secondCol; m->gobble(in);
 404                         names[firstCol] = secondCol;
 405                         int size = 1;
 406
 407                         for(int i=0;i<secondCol.size();i++){
 408                                 if(secondCol[i] == ','){        size++; }
 409                         }
 410                         sizes[firstCol] = size;
 411                 }
 412                 in.close();
 413         }
 414         catch(exception& e) {
 415                 m->errorOut(e, "ClusterFragmentsCommand", "readNameFile");
 416                 exit(1);
 417         }
 418 }
 419
 420 /**************************************************************************************************/
 421