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","fasta-name",false,true,true); parameters.push_back(pfasta);
  32                 CommandParameter pname("name", "InputTypes", "", "", "namecount", "none", "none","name",false,false,true); parameters.push_back(pname);
  33         CommandParameter pcount("count", "InputTypes", "", "", "namecount", "none", "none","count",false,false,true); parameters.push_back(pcount);
  34                 CommandParameter pdiffs("diffs", "Number", "", "0", "", "", "","",false,false); parameters.push_back(pdiffs);
  35                 CommandParameter ppercent("percent", "Number", "", "0", "", "", "","",false,false); parameters.push_back(ppercent);
  36                 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "","",false,false); parameters.push_back(pinputdir);
  37                 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "","",false,false); parameters.push_back(poutputdir);
  38
  39                 vector<string> myArray;
  40                 for (int i = 0; i < parameters.size(); i++) {   myArray.push_back(parameters[i].name);          }
  41                 return myArray;
  42         }
  43         catch(exception& e) {
  44                 m->errorOut(e, "ClusterFragmentsCommand", "setParameters");
  45                 exit(1);
  46         }
  47 }
  48 //**********************************************************************************************************************
  49 string ClusterFragmentsCommand::getHelpString(){
  50         try {
  51                 string helpString = "";
  52                 helpString += "The cluster.fragments command groups sequences that are part of a larger sequence.\n";
  53                 helpString += "The cluster.fragments command outputs a new fasta and name or count file.\n";
  54                 helpString += "The cluster.fragments command parameters are fasta, name, count, diffs and percent. The fasta parameter is required, unless you have a valid current file. \n";
  55                 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";
  56                 helpString += "The diffs parameter allows you to set the number of differences allowed, default=0. \n";
  57                 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";
  58                 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";
  59                 helpString += "The cluster.fragments command should be in the following format: \n";
  60                 helpString += "cluster.fragments(fasta=yourFastaFile, names=yourNamesFile) \n";
  61                 helpString += "Example cluster.fragments(fasta=amazon.fasta).\n";
  62                 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
  63                 return helpString;
  64         }
  65         catch(exception& e) {
  66                 m->errorOut(e, "ClusterFragmentsCommand", "getHelpString");
  67                 exit(1);
  68         }
  69 }
  70 //**********************************************************************************************************************
  71 string ClusterFragmentsCommand::getOutputPattern(string type) {
  72     try {
  73         string pattern = "";
  74
  75         if (type == "fasta") {  pattern = "[filename],fragclust.fasta"; }
  76         else if (type == "name") {  pattern = "[filename],fragclust.names"; }
  77         else if (type == "count") {  pattern = "[filename],fragclust.count_table"; }
  78         else { m->mothurOut("[ERROR]: No definition for type " + type + " output pattern.\n"); m->control_pressed = true;  }
  79
  80         return pattern;
  81     }
  82     catch(exception& e) {
  83         m->errorOut(e, "ClusterFragmentsCommand", "getOutputPattern");
  84         exit(1);
  85     }
  86 }
  87 //**********************************************************************************************************************
  88 ClusterFragmentsCommand::ClusterFragmentsCommand(){
  89         try {
  90                 abort = true; calledHelp = true;
  91                 setParameters();
  92                 vector<string> tempOutNames;
  93                 outputTypes["fasta"] = tempOutNames;
  94                 outputTypes["name"] = tempOutNames;
  95         outputTypes["count"] = tempOutNames;
  96         }
  97         catch(exception& e) {
  98                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
  99                 exit(1);
 100         }
 101 }
 102 //**********************************************************************************************************************
 103 ClusterFragmentsCommand::ClusterFragmentsCommand(string option) {
 104         try {
 105                 abort = false; calledHelp = false;
 106
 107                 //allow user to run help
 108                 if(option == "help") { help(); abort = true; calledHelp = true; }
 109                 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
 110
 111                 else {
 112                         vector<string> myArray = setParameters();
 113
 114                         OptionParser parser(option);
 115                         map<string, string> parameters = parser.getParameters();
 116
 117                         ValidParameters validParameter;
 118                         map<string, string>::iterator it;
 119
 120                         //check to make sure all parameters are valid for command
 121                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
 122                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
 123                         }
 124
 125                         //initialize outputTypes
 126                         vector<string> tempOutNames;
 127                         outputTypes["fasta"] = tempOutNames;
 128                         outputTypes["name"] = tempOutNames;
 129             outputTypes["count"] = tempOutNames;
 130
 131                         //if the user changes the input directory command factory will send this info to us in the output parameter
 132                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 133                         if (inputDir == "not found"){   inputDir = "";          }
 134                         else {
 135                                 string path;
 136                                 it = parameters.find("fasta");
 137                                 //user has given a template file
 138                                 if(it != parameters.end()){
 139                                         path = m->hasPath(it->second);
 140                                         //if the user has not given a path then, add inputdir. else leave path alone.
 141                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 142                                 }
 143
 144                                 it = parameters.find("name");
 145                                 //user has given a template file
 146                                 if(it != parameters.end()){
 147                                         path = m->hasPath(it->second);
 148                                         //if the user has not given a path then, add inputdir. else leave path alone.
 149                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 150                                 }
 151
 152                 it = parameters.find("count");
 153                                 //user has given a template file
 154                                 if(it != parameters.end()){
 155                                         path = m->hasPath(it->second);
 156                                         //if the user has not given a path then, add inputdir. else leave path alone.
 157                                         if (path == "") {       parameters["count"] = inputDir + it->second;            }
 158                                 }
 159                         }
 160
 161                         //check for required parameters
 162                         fastafile = validParameter.validFile(parameters, "fasta", true);
 163                         if (fastafile == "not found") {
 164                                 fastafile = m->getFastaFile();
 165                                 if (fastafile != "") { m->mothurOut("Using " + fastafile + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
 166                                 else {  m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
 167                         }
 168                         else if (fastafile == "not open") { fastafile = ""; abort = true; }
 169                         else { m->setFastaFile(fastafile); }
 170
 171                         //if the user changes the output directory command factory will send this info to us in the output parameter
 172                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = m->hasPath(fastafile);      }
 173
 174                         //check for optional parameter and set defaults
 175                         // ...at some point should added some additional type checking...
 176                         namefile = validParameter.validFile(parameters, "name", true);
 177                         if (namefile == "not found") { namefile =  "";  }
 178                         else if (namefile == "not open") { namefile = ""; abort = true; }
 179                         else {  readNameFile(); m->setNameFile(namefile); }
 180
 181             countfile = validParameter.validFile(parameters, "count", true);
 182                         if (countfile == "not open") { abort = true; countfile = ""; }
 183                         else if (countfile == "not found") { countfile = ""; }
 184                         else { ct.readTable(countfile, true, false); m->setCountTableFile(countfile); }
 185
 186             if ((countfile != "") && (namefile != "")) { m->mothurOut("When executing a cluster.fragments command you must enter ONLY ONE of the following: count or name."); m->mothurOutEndLine(); abort = true; }
 187
 188                         string temp;
 189                         temp = validParameter.validFile(parameters, "diffs", false);            if (temp == "not found"){       temp = "0";                             }
 190                         m->mothurConvert(temp, diffs);
 191
 192                         temp = validParameter.validFile(parameters, "percent", false);          if (temp == "not found"){       temp = "0";                             }
 193                         m->mothurConvert(temp, percent);
 194
 195                         if (countfile == "") {
 196                 if (namefile == "") {
 197                     vector<string> files; files.push_back(fastafile);
 198                     parser.getNameFile(files);
 199                 }
 200             }
 201
 202                 }
 203
 204         }
 205         catch(exception& e) {
 206                 m->errorOut(e, "ClusterFragmentsCommand", "ClusterFragmentsCommand");
 207                 exit(1);
 208         }
 209 }
 210 //**********************************************************************************************************************
 211 int ClusterFragmentsCommand::execute(){
 212         try {
 213
 214                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 215
 216                 int start = time(NULL);
 217
 218                 //reads fasta file and return number of seqs
 219                 int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 220
 221                 if (m->control_pressed) { return 0; }
 222
 223                 if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 224
 225                 //sort seqs by length of unaligned sequence
 226                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 227
 228                 int count = 0;
 229
 230                 //think about running through twice...
 231                 for (int i = 0; i < numSeqs; i++) {
 232
 233                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 234
 235                                 string iBases = alignSeqs[i].seq.getUnaligned();
 236
 237                                 //try to merge it with all smaller seqs
 238                                 for (int j = i+1; j < numSeqs; j++) {
 239
 240                                         if (m->control_pressed) { return 0; }
 241
 242                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 243
 244                                                 string jBases = alignSeqs[j].seq.getUnaligned();
 245
 246                                                 if (isFragment(iBases, jBases)) {
 247                             if (countfile != "") {
 248                                 ct.mergeCounts(alignSeqs[i].names, alignSeqs[j].names);
 249                             }else {
 250                                 //merge
 251                                 alignSeqs[i].names += ',' + alignSeqs[j].names;
 252                                 alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 253                             }
 254                                                         alignSeqs[j].active = 0;
 255                                                         alignSeqs[j].numIdentical = 0;
 256                                                         count++;
 257                                                 }
 258                                         }//end if j active
 259                                 }//end if i != j
 260
 261                                 //remove from active list
 262                                 alignSeqs[i].active = 0;
 263
 264                         }//end if active i
 265                         if(i % 100 == 0)        { m->mothurOutJustToScreen(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)+"\n");       }
 266                 }
 267
 268                 if(numSeqs % 100 != 0)  { m->mothurOutJustToScreen(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)+"\n"); }
 269
 270
 271                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 272         map<string, string> variables;
 273         variables["[filename]"] = fileroot;
 274                 string newFastaFile = getOutputFileName("fasta", variables);
 275                 string newNamesFile = getOutputFileName("name", variables);
 276         if (countfile != "") { newNamesFile = getOutputFileName("count", variables); }
 277
 278                 if (m->control_pressed) { return 0; }
 279
 280                 m->mothurOutEndLine();
 281                 m->mothurOut("Total number of sequences before cluster.fragments was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 282                 m->mothurOut("cluster.fragments removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 283
 284                 printData(newFastaFile, newNamesFile);
 285
 286                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 287
 288                 if (m->control_pressed) { m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
 289
 290                 m->mothurOutEndLine();
 291                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 292                 m->mothurOut(newFastaFile); m->mothurOutEndLine();
 293                 m->mothurOut(newNamesFile); m->mothurOutEndLine();
 294                 outputNames.push_back(newFastaFile);  outputNames.push_back(newNamesFile); outputTypes["fasta"].push_back(newFastaFile); outputTypes["name"].push_back(newNamesFile);
 295                 m->mothurOutEndLine();
 296
 297                 //set fasta file as new current fastafile
 298                 string current = "";
 299                 itTypes = outputTypes.find("fasta");
 300                 if (itTypes != outputTypes.end()) {
 301                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 302                 }
 303
 304                 itTypes = outputTypes.find("name");
 305                 if (itTypes != outputTypes.end()) {
 306                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 307                 }
 308
 309         itTypes = outputTypes.find("count");
 310                 if (itTypes != outputTypes.end()) {
 311                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setCountTableFile(current); }
 312                 }
 313
 314                 return 0;
 315
 316         }
 317         catch(exception& e) {
 318                 m->errorOut(e, "ClusterFragmentsCommand", "execute");
 319                 exit(1);
 320         }
 321 }
 322 //***************************************************************************************************************
 323 bool ClusterFragmentsCommand::isFragment(string seq1, string seq2){
 324         try {
 325                 bool fragment = false;
 326
 327                 //exact match
 328                 int pos = seq1.find(seq2);
 329                 if (pos != string::npos) { return true; }
 330                 //no match, no diffs wanted
 331                 else if ((diffs == 0) && (percent == 0)) { return false; }
 332                 else { //try aligning and see if you can find it
 333
 334                         //find number of acceptable differences for this sequence fragment
 335                         int totalDiffs;
 336                         if (diffs == 0) { //you didnt set diffs you want a percentage
 337                                 totalDiffs = floor((seq2.length() * (percent / 100.0)));
 338                         }else if (percent == 0) { //you didn't set percent you want diffs
 339                                 totalDiffs = diffs;
 340                         }else if ((percent != 0) && (diffs != 0)) { //you want both, set total diffs to smaller of 2
 341                                 totalDiffs = diffs;
 342                                 int percentDiff = floor((seq2.length() * (percent / 100.0)));
 343                                 if (percentDiff < totalDiffs) { totalDiffs = percentDiff; }
 344                         }
 345
 346                         Alignment* alignment = new NeedlemanOverlap(-1.0, 1.0, -1.0, (seq1.length()+totalDiffs+1));
 347
 348                         //use needleman to align
 349                         alignment->align(seq2, seq1);
 350                         string tempSeq2 = alignment->getSeqAAln();
 351                         string temp = alignment->getSeqBAln();
 352
 353                         delete alignment;
 354
 355                         //chop gap ends
 356                         int startPos = 0;
 357                         int endPos = tempSeq2.length()-1;
 358                         for (int i = 0; i < tempSeq2.length(); i++) {  if (isalpha(tempSeq2[i])) { startPos = i; break; } }
 359                         for (int i = tempSeq2.length()-1; i >= 0; i--) {  if (isalpha(tempSeq2[i])) { endPos = i; break; } }
 360
 361                         //count number of diffs
 362                         int numDiffs = 0;
 363                         for (int i = startPos; i <= endPos; i++) {
 364                                 if (tempSeq2[i] != temp[i]) { numDiffs++; }
 365                         }
 366
 367                         if (numDiffs <= totalDiffs) { fragment = true; }
 368
 369                 }
 370
 371                 return fragment;
 372
 373         }
 374         catch(exception& e) {
 375                 m->errorOut(e, "ClusterFragmentsCommand", "isFragment");
 376                 exit(1);
 377         }
 378 }
 379 /**************************************************************************************************/
 380 int ClusterFragmentsCommand::readFASTA(){
 381         try {
 382
 383                 ifstream inFasta;
 384                 m->openInputFile(fastafile, inFasta);
 385
 386                 while (!inFasta.eof()) {
 387
 388                         if (m->control_pressed) { inFasta.close(); return 0; }
 389
 390                         Sequence seq(inFasta);  m->gobble(inFasta);
 391
 392                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 393                                 if (namefile != "") {
 394                                         itSize = sizes.find(seq.getName());
 395
 396                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 397                                         else{
 398                                                 seqRNode tempNode(itSize->second, seq, names[seq.getName()], seq.getUnaligned().length());
 399                                                 alignSeqs.push_back(tempNode);
 400                                         }
 401                 }else if(countfile != "") {
 402                     seqRNode tempNode(ct.getNumSeqs(seq.getName()), seq, seq.getName(), seq.getUnaligned().length());
 403                     alignSeqs.push_back(tempNode);
 404                                 }else { //no names file, you are identical to yourself
 405                                         seqRNode tempNode(1, seq, seq.getName(), seq.getUnaligned().length());
 406                                         alignSeqs.push_back(tempNode);
 407                                 }
 408                         }
 409                 }
 410
 411                 inFasta.close();
 412                 return alignSeqs.size();
 413         }
 414
 415         catch(exception& e) {
 416                 m->errorOut(e, "ClusterFragmentsCommand", "readFASTA");
 417                 exit(1);
 418         }
 419 }
 420 /**************************************************************************************************/
 421 void ClusterFragmentsCommand::printData(string newfasta, string newname){
 422         try {
 423                 ofstream outFasta;
 424                 ofstream outNames;
 425
 426                 m->openOutputFile(newfasta, outFasta);
 427                 if (countfile == "") {  m->openOutputFile(newname, outNames); }
 428
 429                 for (int i = 0; i < alignSeqs.size(); i++) {
 430                         if (alignSeqs[i].numIdentical != 0) {
 431                                 alignSeqs[i].seq.printSequence(outFasta);
 432                                 if (countfile == "") {  outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;  }
 433                         }
 434                 }
 435
 436                 outFasta.close();
 437                 if (countfile == "") {  outNames.close(); }
 438         else { ct.printTable(newname); }
 439         }
 440         catch(exception& e) {
 441                 m->errorOut(e, "ClusterFragmentsCommand", "printData");
 442                 exit(1);
 443         }
 444 }
 445 /**************************************************************************************************/
 446
 447 void ClusterFragmentsCommand::readNameFile(){
 448         try {
 449                 ifstream in;
 450                 m->openInputFile(namefile, in);
 451                 string firstCol, secondCol;
 452
 453                 while (!in.eof()) {
 454                         in >> firstCol >> secondCol; m->gobble(in);
 455                         names[firstCol] = secondCol;
 456                         int size = 1;
 457
 458                         for(int i=0;i<secondCol.size();i++){
 459                                 if(secondCol[i] == ','){        size++; }
 460                         }
 461                         sizes[firstCol] = size;
 462                 }
 463                 in.close();
 464         }
 465         catch(exception& e) {
 466                 m->errorOut(e, "ClusterFragmentsCommand", "readNameFile");
 467                 exit(1);
 468         }
 469 }
 470 /**************************************************************************************************/
 471