5 * Created by Pat Schloss on 12/17/08.
6 * Copyright 2008 Patrick D. Schloss. All rights reserved.
8 * This is my implementation of the NAST (nearest alignment space termination) algorithm as described in:
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.
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.
18 #include "sequence.hpp"
19 #include "alignment.hpp"
22 /**************************************************************************************************/
24 Nast::Nast(Alignment* method, Sequence* cand, Sequence* temp) : alignment(method), candidateSeq(cand), templateSeq(temp) {
26 m = MothurOut::getInstance();
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.
32 m->errorOut(e, "Nast", "Nast");
37 /**************************************************************************************************/
39 void Nast::pairwiseAlignSeqs(){ // Here we call one of the pairwise alignment methods to align our unaligned candidate
40 // and template sequences
43 alignment->align(candidateSeq->getUnaligned(), templateSeq->getUnaligned());
45 string candAln = alignment->getSeqAAln();
46 string tempAln = alignment->getSeqBAln();
50 candidateSeq->setPairwise("");
51 templateSeq->setPairwise(templateSeq->getUnaligned());
55 if(tempAln[0] == '-'){
56 int pairwiseAlignmentLength = tempAln.length(); // we need to make sure that the candidate sequence alignment
57 for(int i=0;i<pairwiseAlignmentLength;i++){ // starts where the template sequence alignment starts, if it
58 if(isalpha(tempAln[i])){ // starts early, we nuke the beginning of the candidate
59 candAln = candAln.substr(i); // sequence
60 tempAln = tempAln.substr(i);
65 int pairwiseAlignmentLength = tempAln.length();
66 if(tempAln[pairwiseAlignmentLength-1] == '-'){ // we need to make sure that the candidate sequence alignment
67 for(int i=pairwiseAlignmentLength-1; i>=0; i--){// ends where the template sequence alignment ends, if it runs
68 if(isalpha(tempAln[i])){ // long, we nuke the end of the candidate sequence
69 candAln = candAln.substr(0,i+1);
70 tempAln = tempAln.substr(0,i+1);
78 candidateSeq->setPairwise(candAln); // set the pairwise sequences in the Sequence objects for
79 templateSeq->setPairwise(tempAln); // the candidate and template sequences
82 m->errorOut(e, "Nast", "pairwiseAlignSeqs");
87 /**************************************************************************************************/
89 void Nast::removeExtraGaps(string& candAln, string tempAln, string newTemplateAlign){
91 // here we do steps C-F of Fig. 2 from DeSantis et al.
94 int longAlignmentLength = newTemplateAlign.length();
96 for(int i=0; i<longAlignmentLength; i++){ // use the long alignment as the standard
97 int rightIndex, rightRoom, leftIndex, leftRoom;
99 // Part C of Fig. 2 from DeSantis et al.
100 if((isalpha(newTemplateAlign[i]) != isalpha(tempAln[i]))){ //if there is a discrepancy between the regapped
102 rightRoom = 0; leftRoom = 0;
104 // Part D of Fig. 2 from DeSantis et al. // template sequence and the official template sequence
105 for(leftIndex=i-1;leftIndex>0;leftIndex--){ // then we've got problems...
106 if(!isalpha(candAln[leftIndex])){
107 leftRoom = 1; //count how far it is to the nearest gap on the LEFT side of the anomaly
108 while(leftIndex-leftRoom>=0 && !isalpha(candAln[leftIndex-leftRoom])) { leftRoom++; }
113 for(rightIndex=i+1;rightIndex<longAlignmentLength-1;rightIndex++){
114 if(!isalpha(candAln[rightIndex])){
115 rightRoom = 1; //count how far it is to the nearest gap on the RIGHT side of the anomaly
116 while(rightIndex+rightRoom<longAlignmentLength && !isalpha(candAln[rightIndex+rightRoom])) { rightRoom++; }
121 int insertLength = 0; // figure out how long the anomaly is
122 while(!isalpha(newTemplateAlign[i + insertLength])) { insertLength++; }
123 if(insertLength > maxInsertLength){ maxInsertLength = insertLength; }
125 if((leftRoom + rightRoom) >= insertLength){
127 // Parts D & E from Fig. 2 of DeSantis et al.
128 if((i-leftIndex) <= (rightIndex-i)){ // the left gap is closer - > move stuff left there's
130 if(leftRoom >= insertLength){ // enough room to the left to move
132 string leftTemplateString = newTemplateAlign.substr(0,i);
133 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
134 newTemplateAlign = leftTemplateString + rightTemplateString;
135 longAlignmentLength = newTemplateAlign.length();
137 string leftCandidateString = candAln.substr(0,leftIndex-insertLength+1);
138 string rightCandidateString = candAln.substr(leftIndex+1);
139 candAln = leftCandidateString + rightCandidateString;
142 else{ // not enough room to the left, have to steal some space to
143 string leftTemplateString = newTemplateAlign.substr(0,i); // the right
144 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
145 newTemplateAlign = leftTemplateString + rightTemplateString;
146 longAlignmentLength = newTemplateAlign.length();
148 string leftCandidateString = candAln.substr(0,leftIndex-leftRoom+1);
149 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
150 string rightCandidateString = candAln.substr(rightIndex+(insertLength-leftRoom));
151 candAln = leftCandidateString + insertString + rightCandidateString;
155 else{ // the right gap is closer - > move stuff right there's
156 if(rightRoom >= insertLength){ // enough room to the right to move
158 string leftTemplateString = newTemplateAlign.substr(0,i);
159 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
160 newTemplateAlign = leftTemplateString + rightTemplateString;
161 longAlignmentLength = newTemplateAlign.length();
163 string leftCandidateString = candAln.substr(0,rightIndex);
164 string rightCandidateString = candAln.substr(rightIndex+insertLength);
165 candAln = leftCandidateString + rightCandidateString;
168 else{ // not enough room to the right, have to steal some
170 // space to the left lets move left and then right...
171 string leftTemplateString = newTemplateAlign.substr(0,i);
172 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
173 newTemplateAlign = leftTemplateString + rightTemplateString;
174 longAlignmentLength = newTemplateAlign.length();
176 string leftCandidateString = candAln.substr(0,leftIndex-(insertLength-rightRoom)+1);
177 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
178 string rightCandidateString = candAln.substr(rightIndex+rightRoom);
179 candAln = leftCandidateString + insertString + rightCandidateString;
185 // there could be a case where there isn't enough room in
186 string leftTemplateString = newTemplateAlign.substr(0,i); // either direction to move stuff
187 string rightTemplateString = newTemplateAlign.substr(i+leftRoom+rightRoom);
188 newTemplateAlign = leftTemplateString + rightTemplateString;
189 longAlignmentLength = newTemplateAlign.length();
191 string leftCandidateString = candAln.substr(0,leftIndex-leftRoom+1);
192 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
193 string rightCandidateString = candAln.substr(rightIndex+rightRoom);
194 candAln = leftCandidateString + insertString + rightCandidateString;
202 catch(exception& e) {
203 m->errorOut(e, "Nast", "removeExtraGaps");
208 /**************************************************************************************************/
210 void Nast::regapSequences(){ //This is essentially part B in Fig 2. of DeSantis et al.
213 string candPair = candidateSeq->getPairwise();
216 string tempPair = templateSeq->getPairwise();
217 string tempAln = templateSeq->getAligned(); // we use the template aligned sequence as our guide
219 int pairwiseLength = candPair.length();
220 int fullAlignLength = tempAln.length();
223 for(int i=0;i<fullAlignLength;i++) { candAln += '.'; }
224 candidateSeq->setAligned(candAln);
228 int fullAlignIndex = 0;
229 int pairwiseAlignIndex = 0;
230 string newTemplateAlign = ""; // this is going to be messy so we want a temporary template
232 while(tempAln[fullAlignIndex] == '.' || tempAln[fullAlignIndex] == '-'){
233 candAln += '.'; // add the initial '-' and '.' to the candidate and template
234 newTemplateAlign += tempAln[fullAlignIndex];// pairwise sequences
238 string lastLoop = "";
240 while(pairwiseAlignIndex<pairwiseLength){
241 if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
242 && isalpha(candPair[pairwiseAlignIndex])){
243 // the template and candidate pairwise and template aligned have characters
244 // need to add character onto the candidatSeq.aligned sequence
246 candAln += candPair[pairwiseAlignIndex];
247 newTemplateAlign += tempPair[pairwiseAlignIndex];//
249 pairwiseAlignIndex++;
252 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
253 && isalpha(candPair[pairwiseAlignIndex])){
254 // the template pairwise and candidate pairwise are characters and the template aligned is a gap
255 // need to insert gaps into the candidateSeq.aligned sequence
258 newTemplateAlign += '-';//
261 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
262 && isalpha(candPair[pairwiseAlignIndex])){
263 // the template pairwise is a gap and the template aligned and pairwise sequences have characters
264 // this is the alpha scenario. add character to the candidateSeq.aligned sequence without progressing
265 // further through the tempAln sequence.
267 candAln += candPair[pairwiseAlignIndex];
268 newTemplateAlign += '-';//
269 pairwiseAlignIndex++;
271 else if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
272 && !isalpha(candPair[pairwiseAlignIndex])){
273 // the template pairwise and full alignment are characters and the candidate sequence has a gap
274 // should not be a big deal, just add the gap position to the candidateSeq.aligned sequence;
276 candAln += candPair[pairwiseAlignIndex];
277 newTemplateAlign += tempAln[fullAlignIndex];//
279 pairwiseAlignIndex++;
281 else if(!isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
282 && isalpha(candPair[pairwiseAlignIndex])){
283 // the template pairwise and aligned are gaps while the candidate pairwise has a character
284 // this would be an insertion, go ahead and add the character->seems to be the opposite of the alpha scenario
286 candAln += candPair[pairwiseAlignIndex];
287 newTemplateAlign += tempAln[fullAlignIndex];//
288 pairwiseAlignIndex++;
291 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
292 && !isalpha(candPair[pairwiseAlignIndex])){
293 // template pairwise has a character, but its full aligned sequence and candidate sequence have gaps
294 // this would happen like we need to add a gap. basically the opposite of the alpha situation
296 newTemplateAlign += tempAln[fullAlignIndex];//
300 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
301 && !isalpha(candPair[pairwiseAlignIndex])){
302 // template and candidate pairwise are gaps and the template aligned is not a gap this should not be possible
303 // would skip the gaps and not progress through full alignment sequence
306 m->mothurOut("We're into D " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); m->mothurOutEndLine();
307 pairwiseAlignIndex++;
310 // everything has a gap - not possible
313 m->mothurOut("We're into F " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); m->mothurOutEndLine();
314 pairwiseAlignIndex++;
319 for(int i=fullAlignIndex;i<fullAlignLength;i++){
321 newTemplateAlign += tempAln[i];//
325 int end = candAln.length()-1;
327 for(int i=0;i<candAln.length();i++){
328 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; } // if we padded the alignemnt from
329 else{ start = i; break; } // blast with Z's, change them to
332 for(int i=candAln.length()-1;i>=0;i--){ // ditto.
333 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; }
334 else{ end = i; break; }
337 for(int i=start;i<=end;i++){ // go through the candidate alignment sequence and make sure that
338 candAln[i] = toupper(candAln[i]); // everything is upper case
342 if(candAln.length() != tempAln.length()){ // if the regapped candidate sequence is longer than the official
343 removeExtraGaps(candAln, tempAln, newTemplateAlign);// template alignment then we need to do steps C-F in Fig.
344 } // 2 of Desantis et al.
346 candidateSeq->setAligned(candAln);
348 catch(exception& e) {
349 m->errorOut(e, "Nast", "regapSequences");
354 /**************************************************************************************************/
356 float Nast::getSimilarityScore(){
359 string cand = candidateSeq->getAligned();
360 string temp = templateSeq->getAligned();
361 int alignmentLength = temp.length();
365 for(int i=0;i<alignmentLength;i++){
366 if(cand[i] == '-' && temp[i] == '-'){
369 else if(cand[i] != '.' && temp[i] != '.'){
372 if(cand[i] != temp[i]){
377 float similarity = 100 * (1. - mismatch / (float)denominator);
378 if(denominator == 0){ similarity = 0.0000; }
383 catch(exception& e) {
384 m->errorOut(e, "Nast", "getSimilarityScore");
389 /**************************************************************************************************/
391 int Nast::getMaxInsertLength(){
393 return maxInsertLength;
397 /**************************************************************************************************/