nast.cpp

   1 /*
   2  *  nast.cpp
   3  *
   4  *
   5  *  Created by Pat Schloss on 12/17/08.
   6  *  Copyright 2008 Patrick D. Schloss. All rights reserved.
   7  *
   8  *      This is my implementation of the NAST (nearest alignment space termination) algorithm as described in:
   9  *
  10  *      DeSantis TZ, Hugenholtz P, Keller K, Brodie EL, Larsen N, Piceno YM, Phan R, & Anderson GL.  2006.  NAST: a multiple
  11  *              sequence alignment server for comparative analysis of 16S rRNA genes.  Nucleic Acids Research.  34:W394-9.
  12  *
  13  *      To construct an object one needs to provide a method of getting a pairwise alignment (alignment) and the template
  14  *      and candidate sequence that are to be aligned to each other.
  15  *
  16  */
  17
  18 #include "sequence.hpp"
  19 #include "alignment.hpp"
  20 #include "nast.hpp"
  21
  22 /**************************************************************************************************/
  23
  24 Nast::Nast(Alignment* method, Sequence* cand, Sequence* temp) : alignment(method), candidateSeq(cand), templateSeq(temp) {
  25         try {
  26                 m = MothurOut::getInstance();
  27                 maxInsertLength = 0;
  28                 pairwiseAlignSeqs();    //      This is part A in Fig. 2 of DeSantis et al.
  29                 regapSequences();               //      This is parts B-F in Fig. 2 of DeSantis et al.
  30         }
  31         catch(exception& e) {
  32                 m->errorOut(e, "Nast", "Nast");
  33                 exit(1);
  34         }
  35 }
  36
  37 /**************************************************************************************************/
  38
  39 void Nast::pairwiseAlignSeqs(){ //      Here we call one of the pairwise alignment methods to align our unaligned candidate
  40                                                                 //      and template sequences
  41         try {
  42                 alignment->align(candidateSeq->getUnaligned(), templateSeq->getUnaligned());
  43
  44                 string candAln = alignment->getSeqAAln();
  45                 string tempAln = alignment->getSeqBAln();
  46
  47                 if(candAln == ""){
  48
  49                         candidateSeq->setPairwise("");
  50                         templateSeq->setPairwise(templateSeq->getUnaligned());
  51
  52                 }
  53                 else{
  54                         if(tempAln[0] == '-'){
  55                                 int pairwiseAlignmentLength = tempAln.length(); //      we need to make sure that the candidate sequence alignment
  56                                 for(int i=0;i<pairwiseAlignmentLength;i++){             //      starts where the template sequence alignment starts, if it
  57                                         if(isalpha(tempAln[i])){                                        //      starts early, we nuke the beginning of the candidate
  58                                                 candAln = candAln.substr(i);                    //      sequence
  59                                                 tempAln = tempAln.substr(i);
  60                                                 break;
  61                                         }
  62                                 }
  63                         }
  64                         int pairwiseAlignmentLength = tempAln.length();
  65                         if(tempAln[pairwiseAlignmentLength-1] == '-'){          //      we need to make sure that the candidate sequence alignment
  66                                 for(int i=pairwiseAlignmentLength-1; i>=0; i--){//      ends where the template sequence alignment ends, if it runs
  67                                         if(isalpha(tempAln[i])){                                        //      long, we nuke the end of the candidate sequence
  68                                                 candAln = candAln.substr(0,i+1);
  69                                                 tempAln = tempAln.substr(0,i+1);
  70                                                 break;
  71                                         }
  72                                 }
  73                         }
  74
  75                 }
  76
  77                 candidateSeq->setPairwise(candAln);                                     //      set the pairwise sequences in the Sequence objects for
  78                 templateSeq->setPairwise(tempAln);                                      //      the candidate and template sequences
  79         }
  80         catch(exception& e) {
  81                 m->errorOut(e, "Nast", "pairwiseAlignSeqs");
  82                 exit(1);
  83         }
  84 }
  85
  86 /**************************************************************************************************/
  87
  88 void Nast::removeExtraGaps(string& candAln, string tempAln, string newTemplateAlign){
  89
  90 //      here we do steps C-F of Fig. 2 from DeSantis et al.
  91         try {
  92
  93                 //cout << candAln << endl;
  94                 //cout << tempAln << endl;
  95                 //cout << newTemplateAlign << endl;
  96                 //cout << endl;
  97
  98                 int longAlignmentLength = newTemplateAlign.length();
  99
 100                 for(int i=0; i<longAlignmentLength; i++){                               //      use the long alignment as the standard
 101                         int rightIndex, rightRoom, leftIndex, leftRoom;
 102
 103                         //      Part C of Fig. 2 from DeSantis et al.
 104                         if((isalpha(newTemplateAlign[i]) != isalpha(tempAln[i]))){      //if there is a discrepancy between the regapped
 105
 106                                 rightRoom = 0; leftRoom = 0;
 107
 108                                 //      Part D of Fig. 2 from DeSantis et al.           //      template sequence and the official template sequence
 109                                 for(leftIndex=i-1;leftIndex>0;leftIndex--){             //      then we've got problems...
 110                                         if(!isalpha(candAln[leftIndex])){
 111                                                 leftRoom = 1;   //count how far it is to the nearest gap on the LEFT side of the anomaly
 112                                                 while(leftIndex-leftRoom>=0 && !isalpha(candAln[leftIndex-leftRoom]))   {       leftRoom++;             }
 113                                                 break;
 114                                         }
 115                                 }
 116
 117                                 for(rightIndex=i+1;rightIndex<longAlignmentLength-1;rightIndex++){
 118                                         if(!isalpha(candAln[rightIndex])){
 119                                                 rightRoom = 1;  //count how far it is to the nearest gap on the RIGHT side of the anomaly
 120                                                 while(rightIndex+rightRoom<longAlignmentLength && !isalpha(candAln[rightIndex+rightRoom]))      {       rightRoom++;    }
 121                                                 break;
 122                                         }
 123                                 }
 124
 125                                 int insertLength = 0;                                                   //      figure out how long the anomaly is
 126                                 while(!isalpha(newTemplateAlign[i + insertLength]))     {       insertLength++; }
 127                                 if(insertLength > maxInsertLength){     maxInsertLength = insertLength; }
 128
 129                                 if((leftRoom + rightRoom) >= insertLength){
 130
 131                                         //      Parts D & E from Fig. 2 of DeSantis et al.
 132                                         if((i-leftIndex) <= (rightIndex-i)){            //      the left gap is closer - > move stuff left there's
 133
 134                                                 if(leftRoom >= insertLength){                   //      enough room to the left to move
 135                         //cout << "lr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
 136                                                         string leftTemplateString = newTemplateAlign.substr(0,i);
 137                                                         string rightTemplateString = newTemplateAlign.substr((i+insertLength));
 138                                                         newTemplateAlign = leftTemplateString + rightTemplateString;
 139                                                         longAlignmentLength = newTemplateAlign.length();
 140                         //cout << "lr candAln = " << candAln.length() << '\t' << leftIndex-insertLength+1 << '\t' << leftIndex+1 << endl;
 141                                                         string leftCandidateString = candAln.substr(0,(leftIndex-insertLength+1));
 142                                                         string rightCandidateString = candAln.substr((leftIndex+1));
 143                                                         candAln = leftCandidateString + rightCandidateString;
 144
 145                                                 }
 146                                                 else{                                                                   //      not enough room to the left, have to steal some space to
 147
 148                         //cout << "in else lr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
 149                                                         string leftTemplateString = newTemplateAlign.substr(0,i);       //      the right
 150                                                         string rightTemplateString = newTemplateAlign.substr((i+insertLength));
 151                                                         newTemplateAlign = leftTemplateString + rightTemplateString;
 152                                                         longAlignmentLength = newTemplateAlign.length();
 153                         //cout << " in else lr candAln = " << candAln.length() << '\t' << leftIndex-leftRoom+1 << '\t' << rightIndex-leftIndex-1 << '\t' << rightIndex+(insertLength-leftRoom) << endl;
 154                                                         string leftCandidateString = candAln.substr(0,(leftIndex-leftRoom+1));
 155                                                         string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
 156                                                         string rightCandidateString = candAln.substr((rightIndex+(insertLength-leftRoom)));
 157                                                         candAln = leftCandidateString + insertString + rightCandidateString;
 158
 159                                                 }
 160                                         }
 161                                         else{                                                                           //      the right gap is closer - > move stuff right there's
 162                                                 if(rightRoom >= insertLength){                  //      enough room to the right to move
 163                         //cout << "rr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
 164                                                         string leftTemplateString = newTemplateAlign.substr(0,i);
 165                                                         string rightTemplateString = newTemplateAlign.substr((i+insertLength));
 166                                                         newTemplateAlign = leftTemplateString + rightTemplateString;
 167                                                         longAlignmentLength = newTemplateAlign.length();
 168                         //cout << "rr candAln = " << candAln.length() << '\t' << i << '\t' << rightIndex << '\t' << rightIndex+insertLength << endl;
 169                                                         string leftCandidateString = candAln.substr(0,rightIndex);
 170                                                         string rightCandidateString = candAln.substr((rightIndex+insertLength));
 171                                                         candAln = leftCandidateString + rightCandidateString;
 172
 173                                                 }
 174                                                 else{                                                                   //      not enough room to the right, have to steal some
 175                                                         //      space to the left lets move left and then right...
 176                                         //cout << "in else rr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
 177                                                         string leftTemplateString = newTemplateAlign.substr(0,i);
 178                                                         string rightTemplateString = newTemplateAlign.substr((i+insertLength));
 179                                                         newTemplateAlign = leftTemplateString + rightTemplateString;
 180                                                         longAlignmentLength = newTemplateAlign.length();
 181                                         //cout << "in else rr candAln = " << candAln.length() << '\t' << '\t' << (leftIndex-(insertLength-rightRoom)+1) << '\t' <<  (leftIndex+1,rightIndex-leftIndex-1) << '\t' << (rightIndex+rightRoom) << endl;
 182                                                         string leftCandidateString = candAln.substr(0,(leftIndex-(insertLength-rightRoom)+1));
 183                                                         string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
 184                                                         string rightCandidateString = candAln.substr((rightIndex+rightRoom));
 185                                                         candAln = leftCandidateString + insertString + rightCandidateString;
 186
 187                                                 }
 188                                         }
 189                                         i -= insertLength;
 190
 191                                 }
 192                                 else{
 193                         //      there could be a case where there isn't enough room in either direction to move stuff
 194 //cout << "in else else newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << (i+leftRoom+rightRoom) << endl;
 195                                         string leftTemplateString = newTemplateAlign.substr(0,i);
 196                                         string rightTemplateString = newTemplateAlign.substr((i+leftRoom+rightRoom));
 197                                         newTemplateAlign = leftTemplateString + rightTemplateString;
 198                                         longAlignmentLength = newTemplateAlign.length();
 199
 200                 //cout << "in else else newTemplateAlign = " << candAln.length() << '\t' << (leftIndex-leftRoom+1) << '\t' << (leftIndex+1) << '\t' << (rightIndex-leftIndex-1) << '\t' << (rightIndex+rightRoom) << endl;
 201                                         string leftCandidateString = candAln.substr(0,(leftIndex-leftRoom+1));
 202                                         string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
 203                                         string rightCandidateString = candAln.substr((rightIndex+rightRoom));
 204                                         candAln = leftCandidateString + insertString + rightCandidateString;
 205
 206                                         i -= (leftRoom + rightRoom);
 207                                 }
 208
 209 //                              i -= insertLength;
 210                         }
 211                 }
 212         }
 213         catch(exception& e) {
 214                 m->errorOut(e, "Nast", "removeExtraGaps");
 215                 exit(1);
 216         }
 217 }
 218
 219 /**************************************************************************************************/
 220
 221 void Nast::regapSequences(){    //This is essentially part B in Fig 2. of DeSantis et al.
 222         try {
 223
 224                 string candPair = candidateSeq->getPairwise();
 225                 string candAln = "";
 226
 227                 string tempPair = templateSeq->getPairwise();
 228                 string tempAln = templateSeq->getAligned();             //      we use the template aligned sequence as our guide
 229
 230                 int pairwiseLength = candPair.length();
 231                 int fullAlignLength = tempAln.length();
 232
 233                 if(candPair == ""){
 234                         for(int i=0;i<fullAlignLength;i++)      {       candAln += '.';         }
 235                         candidateSeq->setAligned(candAln);
 236                         return;
 237                 }
 238
 239                 int fullAlignIndex = 0;
 240                 int pairwiseAlignIndex = 0;
 241                 string newTemplateAlign = "";                                   //      this is going to be messy so we want a temporary template
 242                 //      alignment string
 243                 while(tempAln[fullAlignIndex] == '.' || tempAln[fullAlignIndex]  == '-'){
 244                         candAln += '.';                                                         //      add the initial '-' and '.' to the candidate and template
 245                         newTemplateAlign += tempAln[fullAlignIndex];//  pairwise sequences
 246                         fullAlignIndex++;
 247                 }
 248
 249                 string lastLoop = "";
 250
 251                 while(pairwiseAlignIndex<pairwiseLength){
 252                         if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
 253                            && isalpha(candPair[pairwiseAlignIndex])){
 254                                 //  the template and candidate pairwise and template aligned have characters
 255                                 //      need to add character onto the candidatSeq.aligned sequence
 256
 257                                 candAln += candPair[pairwiseAlignIndex];
 258                                 newTemplateAlign += tempPair[pairwiseAlignIndex];//
 259
 260                                 pairwiseAlignIndex++;
 261                                 fullAlignIndex++;
 262                         }
 263                         else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
 264                                         && isalpha(candPair[pairwiseAlignIndex])){
 265                                 //      the template pairwise and candidate pairwise are characters and the template aligned is a gap
 266                                 //      need to insert gaps into the candidateSeq.aligned sequence
 267
 268                                 candAln += '-';
 269                                 newTemplateAlign += '-';//
 270                                 fullAlignIndex++;
 271                         }
 272                         else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
 273                                         && isalpha(candPair[pairwiseAlignIndex])){
 274                                 //  the template pairwise is a gap and the template aligned and pairwise sequences have characters
 275                                 //      this is the alpha scenario.  add character to the candidateSeq.aligned sequence without progressing
 276                                 //      further through the tempAln sequence.
 277
 278                                 candAln += candPair[pairwiseAlignIndex];
 279                                 newTemplateAlign += '-';//
 280                                 pairwiseAlignIndex++;
 281                         }
 282                         else if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
 283                                         && !isalpha(candPair[pairwiseAlignIndex])){
 284                                 //  the template pairwise and full alignment are characters and the candidate sequence has a gap
 285                                 //      should not be a big deal, just add the gap position to the candidateSeq.aligned sequence;
 286
 287                                 candAln += candPair[pairwiseAlignIndex];
 288                                 newTemplateAlign += tempAln[fullAlignIndex];//
 289                                 fullAlignIndex++;
 290                                 pairwiseAlignIndex++;
 291                         }
 292                         else if(!isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
 293                                         && isalpha(candPair[pairwiseAlignIndex])){
 294                                 //      the template pairwise and aligned are gaps while the candidate pairwise has a character
 295                                 //      this would be an insertion, go ahead and add the character->seems to be the opposite of the alpha scenario
 296
 297                                 candAln += candPair[pairwiseAlignIndex];
 298                                 newTemplateAlign += tempAln[fullAlignIndex];//
 299                                 pairwiseAlignIndex++;
 300                                 fullAlignIndex++;
 301                         }
 302                         else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
 303                                         && !isalpha(candPair[pairwiseAlignIndex])){
 304                                 //      template pairwise has a character, but its full aligned sequence and candidate sequence have gaps
 305                                 //      this would happen like we need to add a gap.  basically the opposite of the alpha situation
 306
 307                                 newTemplateAlign += tempAln[fullAlignIndex];//
 308                                 candAln += "-";
 309                                 fullAlignIndex++;
 310                         }
 311                         else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
 312                                         && !isalpha(candPair[pairwiseAlignIndex])){
 313                                 //      template and candidate pairwise are gaps and the template aligned is not a gap this should not be possible
 314                                 //      would skip the gaps and not progress through full alignment sequence
 315                                 //      not tested yet
 316
 317                                 m->mothurOut("We're into D " + toString(fullAlignIndex) + " " +  toString(pairwiseAlignIndex)); m->mothurOutEndLine();
 318                                 pairwiseAlignIndex++;
 319                         }
 320                         else{
 321                                 //      everything has a gap - not possible
 322                                 //      not tested yet
 323
 324                                 m->mothurOut("We're into F " +  toString(fullAlignIndex) + " " +  toString(pairwiseAlignIndex)); m->mothurOutEndLine();
 325                                 pairwiseAlignIndex++;
 326                                 fullAlignIndex++;
 327                         }
 328                 }
 329
 330                 for(int i=fullAlignIndex;i<fullAlignLength;i++){
 331                         candAln += '.';
 332                         newTemplateAlign += tempAln[i];//
 333                 }
 334
 335                 int start = 0;
 336                 int end = candAln.length()-1;
 337
 338                 for(int i=0;i<candAln.length();i++){
 339                         if(candAln[i] == 'Z' || !isalnum(candAln[i]))   {       candAln[i] = '.';       }       //      if we padded the alignemnt from
 340                         else{                   start = i;                      break;          }                                                       //      blast with Z's, change them to
 341                 }                                                                                                                                                               //      '.' characters
 342
 343                 for(int i=candAln.length()-1;i>=0;i--){                                                                                 //      ditto.
 344                         if(candAln[i] == 'Z' || !isalnum(candAln[i]))   {       candAln[i] = '.';       }
 345                         else{                   end = i;                        break;          }
 346                 }
 347
 348                 for(int i=start;i<=end;i++){                                    //      go through the candidate alignment sequence and make sure that
 349                         candAln[i] = toupper(candAln[i]);                       //      everything is upper case
 350                 }
 351
 352
 353                 if(candAln.length() != tempAln.length()){               //      if the regapped candidate sequence is longer than the official
 354                         removeExtraGaps(candAln, tempAln, newTemplateAlign);//  template alignment then we need to do steps C-F in Fig.
 355                 }                                                                                               //      2 of Desantis et al.
 356
 357                 candidateSeq->setAligned(candAln);
 358         }
 359         catch(exception& e) {
 360                 m->errorOut(e, "Nast", "regapSequences");
 361                 exit(1);
 362         }
 363 }
 364
 365 /**************************************************************************************************/
 366
 367 float Nast::getSimilarityScore(){
 368         try {
 369
 370                 string cand = candidateSeq->getAligned();
 371                 string temp = templateSeq->getAligned();
 372                 int alignmentLength = temp.length();
 373                 int mismatch = 0;
 374                 int denominator = 0;
 375
 376                 for(int i=0;i<alignmentLength;i++){
 377                         if(cand[i] == '-' && temp[i] == '-'){
 378
 379                         }
 380                         else if(cand[i] != '.' && temp[i] != '.'){
 381                                 denominator++;
 382
 383                                 if(cand[i] != temp[i]){
 384                                         mismatch++;
 385                                 }
 386                         }
 387                 }
 388                 float similarity = 100 * (1. - mismatch / (float)denominator);
 389                 if(denominator == 0){   similarity = 0.0000;    }
 390
 391                 return similarity;
 392
 393         }
 394         catch(exception& e) {
 395                 m->errorOut(e, "Nast", "getSimilarityScore");
 396                 exit(1);
 397         }
 398 }
 399
 400 /**************************************************************************************************/
 401
 402 int Nast::getMaxInsertLength(){
 403
 404         return maxInsertLength;
 405
 406 }
 407
 408 /**************************************************************************************************/