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) {
27 pairwiseAlignSeqs(); // This is part A in Fig. 2 of DeSantis et al.
28 regapSequences(); // This is parts B-F in Fig. 2 of DeSantis et al.
31 errorOut(e, "Nast", "Nast");
36 /**************************************************************************************************/
38 void Nast::pairwiseAlignSeqs(){ // Here we call one of the pairwise alignment methods to align our unaligned candidate
39 // and template sequences
42 alignment->align(candidateSeq->getUnaligned(), templateSeq->getUnaligned());
44 string candAln = alignment->getSeqAAln();
45 string tempAln = alignment->getSeqBAln();
49 candidateSeq->setPairwise("");
50 templateSeq->setPairwise(templateSeq->getUnaligned());
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);
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);
77 candidateSeq->setPairwise(candAln); // set the pairwise sequences in the Sequence objects for
78 templateSeq->setPairwise(tempAln); // the candidate and template sequences
81 errorOut(e, "Nast", "pairwiseAlignSeqs");
86 /**************************************************************************************************/
88 void Nast::removeExtraGaps(string& candAln, string tempAln, string newTemplateAlign){
90 // here we do steps C-F of Fig. 2 from DeSantis et al.
93 int longAlignmentLength = newTemplateAlign.length();
95 for(int i=0; i<longAlignmentLength; i++){ // use the long alignment as the standard
96 int rightIndex, rightRoom, leftIndex, leftRoom;
98 // Part C of Fig. 2 from DeSantis et al.
99 if((isalpha(newTemplateAlign[i]) != isalpha(tempAln[i]))){ //if there is a discrepancy between the regapped
101 rightRoom = 0; leftRoom = 0;
103 // Part D of Fig. 2 from DeSantis et al. // template sequence and the official template sequence
104 for(leftIndex=i-1;leftIndex>0;leftIndex--){ // then we've got problems...
105 if(!isalpha(candAln[leftIndex])){
106 leftRoom = 1; //count how far it is to the nearest gap on the LEFT side of the anomaly
107 while(leftIndex-leftRoom>=0 && !isalpha(candAln[leftIndex-leftRoom])) { leftRoom++; }
112 for(rightIndex=i+1;rightIndex<longAlignmentLength-1;rightIndex++){
113 if(!isalpha(candAln[rightIndex])){
114 rightRoom = 1; //count how far it is to the nearest gap on the RIGHT side of the anomaly
115 while(rightIndex+rightRoom<longAlignmentLength && !isalpha(candAln[rightIndex+rightRoom])) { rightRoom++; }
120 int insertLength = 0; // figure out how long the anomaly is
121 while(!isalpha(newTemplateAlign[i + insertLength])) { insertLength++; }
122 if(insertLength > maxInsertLength){ maxInsertLength = insertLength; }
124 if((leftRoom + rightRoom) >= insertLength){
126 // Parts D & E from Fig. 2 of DeSantis et al.
127 if((i-leftIndex) <= (rightIndex-i)){ // the left gap is closer - > move stuff left there's
129 if(leftRoom >= insertLength){ // enough room to the left to move
131 string leftTemplateString = newTemplateAlign.substr(0,i);
132 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
133 newTemplateAlign = leftTemplateString + rightTemplateString;
134 longAlignmentLength = newTemplateAlign.length();
136 string leftCandidateString = candAln.substr(0,leftIndex-insertLength+1);
137 string rightCandidateString = candAln.substr(leftIndex+1);
138 candAln = leftCandidateString + rightCandidateString;
141 else{ // not enough room to the left, have to steal some space to
142 string leftTemplateString = newTemplateAlign.substr(0,i); // the right
143 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
144 newTemplateAlign = leftTemplateString + rightTemplateString;
145 longAlignmentLength = newTemplateAlign.length();
147 string leftCandidateString = candAln.substr(0,leftIndex-leftRoom+1);
148 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
149 string rightCandidateString = candAln.substr(rightIndex+(insertLength-leftRoom));
150 candAln = leftCandidateString + insertString + rightCandidateString;
154 else{ // the right gap is closer - > move stuff right there's
155 if(rightRoom >= insertLength){ // enough room to the right to move
157 string leftTemplateString = newTemplateAlign.substr(0,i);
158 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
159 newTemplateAlign = leftTemplateString + rightTemplateString;
160 longAlignmentLength = newTemplateAlign.length();
162 string leftCandidateString = candAln.substr(0,rightIndex);
163 string rightCandidateString = candAln.substr(rightIndex+insertLength);
164 candAln = leftCandidateString + rightCandidateString;
167 else{ // not enough room to the right, have to steal some
169 // space to the left lets move left and then right...
170 string leftTemplateString = newTemplateAlign.substr(0,i);
171 string rightTemplateString = newTemplateAlign.substr(i+insertLength);
172 newTemplateAlign = leftTemplateString + rightTemplateString;
173 longAlignmentLength = newTemplateAlign.length();
175 string leftCandidateString = candAln.substr(0,leftIndex-(insertLength-rightRoom)+1);
176 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
177 string rightCandidateString = candAln.substr(rightIndex+rightRoom);
178 candAln = leftCandidateString + insertString + rightCandidateString;
184 // there could be a case where there isn't enough room in
185 string leftTemplateString = newTemplateAlign.substr(0,i); // either direction to move stuff
186 string rightTemplateString = newTemplateAlign.substr(i+leftRoom+rightRoom);
187 newTemplateAlign = leftTemplateString + rightTemplateString;
188 longAlignmentLength = newTemplateAlign.length();
190 string leftCandidateString = candAln.substr(0,leftIndex-leftRoom+1);
191 string insertString = candAln.substr(leftIndex+1,rightIndex-leftIndex-1);
192 string rightCandidateString = candAln.substr(rightIndex+rightRoom);
193 candAln = leftCandidateString + insertString + rightCandidateString;
201 catch(exception& e) {
202 errorOut(e, "Nast", "removeExtraGaps");
207 /**************************************************************************************************/
209 void Nast::regapSequences(){ //This is essentially part B in Fig 2. of DeSantis et al.
212 string candPair = candidateSeq->getPairwise();
215 string tempPair = templateSeq->getPairwise();
216 string tempAln = templateSeq->getAligned(); // we use the template aligned sequence as our guide
218 int pairwiseLength = candPair.length();
219 int fullAlignLength = tempAln.length();
222 for(int i=0;i<fullAlignLength;i++) { candAln += '.'; }
223 candidateSeq->setAligned(candAln);
227 int fullAlignIndex = 0;
228 int pairwiseAlignIndex = 0;
229 string newTemplateAlign = ""; // this is going to be messy so we want a temporary template
231 while(tempAln[fullAlignIndex] == '.' || tempAln[fullAlignIndex] == '-'){
232 candAln += '.'; // add the initial '-' and '.' to the candidate and template
233 newTemplateAlign += tempAln[fullAlignIndex];// pairwise sequences
237 string lastLoop = "";
239 while(pairwiseAlignIndex<pairwiseLength){
240 if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
241 && isalpha(candPair[pairwiseAlignIndex])){
242 // the template and candidate pairwise and template aligned have characters
243 // need to add character onto the candidatSeq.aligned sequence
245 candAln += candPair[pairwiseAlignIndex];
246 newTemplateAlign += tempPair[pairwiseAlignIndex];//
248 pairwiseAlignIndex++;
251 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
252 && isalpha(candPair[pairwiseAlignIndex])){
253 // the template pairwise and candidate pairwise are characters and the template aligned is a gap
254 // need to insert gaps into the candidateSeq.aligned sequence
257 newTemplateAlign += '-';//
260 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
261 && isalpha(candPair[pairwiseAlignIndex])){
262 // the template pairwise is a gap and the template aligned and pairwise sequences have characters
263 // this is the alpha scenario. add character to the candidateSeq.aligned sequence without progressing
264 // further through the tempAln sequence.
266 candAln += candPair[pairwiseAlignIndex];
267 newTemplateAlign += '-';//
268 pairwiseAlignIndex++;
270 else if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
271 && !isalpha(candPair[pairwiseAlignIndex])){
272 // the template pairwise and full alignment are characters and the candidate sequence has a gap
273 // should not be a big deal, just add the gap position to the candidateSeq.aligned sequence;
275 candAln += candPair[pairwiseAlignIndex];
276 newTemplateAlign += tempAln[fullAlignIndex];//
278 pairwiseAlignIndex++;
280 else if(!isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
281 && isalpha(candPair[pairwiseAlignIndex])){
282 // the template pairwise and aligned are gaps while the candidate pairwise has a character
283 // this would be an insertion, go ahead and add the character->seems to be the opposite of the alpha scenario
285 candAln += candPair[pairwiseAlignIndex];
286 newTemplateAlign += tempAln[fullAlignIndex];//
287 pairwiseAlignIndex++;
290 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
291 && !isalpha(candPair[pairwiseAlignIndex])){
292 // template pairwise has a character, but its full aligned sequence and candidate sequence have gaps
293 // this would happen like we need to add a gap. basically the opposite of the alpha situation
295 newTemplateAlign += tempAln[fullAlignIndex];//
299 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
300 && !isalpha(candPair[pairwiseAlignIndex])){
301 // template and candidate pairwise are gaps and the template aligned is not a gap this should not be possible
302 // would skip the gaps and not progress through full alignment sequence
305 mothurOut("We're into D " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); mothurOutEndLine();
306 pairwiseAlignIndex++;
309 // everything has a gap - not possible
312 mothurOut("We're into F " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); mothurOutEndLine();
313 pairwiseAlignIndex++;
318 for(int i=fullAlignIndex;i<fullAlignLength;i++){
320 newTemplateAlign += tempAln[i];//
324 int end = candAln.length()-1;
326 for(int i=0;i<candAln.length();i++){
327 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; } // if we padded the alignemnt from
328 else{ start = i; break; } // blast with Z's, change them to
331 for(int i=candAln.length()-1;i>=0;i--){ // ditto.
332 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; }
333 else{ end = i; break; }
336 for(int i=start;i<=end;i++){ // go through the candidate alignment sequence and make sure that
337 candAln[i] = toupper(candAln[i]); // everything is upper case
341 if(candAln.length() != tempAln.length()){ // if the regapped candidate sequence is longer than the official
342 removeExtraGaps(candAln, tempAln, newTemplateAlign);// template alignment then we need to do steps C-F in Fig.
343 } // 2 of Desantis et al.
345 candidateSeq->setAligned(candAln);
347 catch(exception& e) {
348 errorOut(e, "Nast", "regapSequences");
353 /**************************************************************************************************/
355 float Nast::getSimilarityScore(){
358 string cand = candidateSeq->getAligned();
359 string temp = templateSeq->getAligned();
360 int alignmentLength = temp.length();
364 for(int i=0;i<alignmentLength;i++){
365 if(cand[i] == '-' && temp[i] == '-'){
368 else if(cand[i] != '.' && temp[i] != '.'){
371 if(cand[i] != temp[i]){
376 float similarity = 100 * (1. - mismatch / (float)denominator);
377 if(denominator == 0){ similarity = 0.0000; }
382 catch(exception& e) {
383 errorOut(e, "Nast", "getSimilarityScore");
388 /**************************************************************************************************/
390 int Nast::getMaxInsertLength(){
392 return maxInsertLength;
396 /**************************************************************************************************/