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();
29 pairwiseAlignSeqs(); // This is part A in Fig. 2 of DeSantis et al.
30 regapSequences(); // This is parts B-F in Fig. 2 of DeSantis et al.
34 m->errorOut(e, "Nast", "Nast");
39 /**************************************************************************************************/
41 void Nast::pairwiseAlignSeqs(){ // Here we call one of the pairwise alignment methods to align our unaligned candidate
42 // and template sequences
44 alignment->align(candidateSeq->getUnaligned(), templateSeq->getUnaligned());
46 string candAln = alignment->getSeqAAln();
47 string tempAln = alignment->getSeqBAln();
51 candidateSeq->setPairwise("");
52 templateSeq->setPairwise(templateSeq->getUnaligned());
56 if(tempAln[0] == '-'){
57 int pairwiseAlignmentLength = tempAln.length(); // we need to make sure that the candidate sequence alignment
58 for(int i=0;i<pairwiseAlignmentLength;i++){ // starts where the template sequence alignment starts, if it
59 if(isalpha(tempAln[i])){ // starts early, we nuke the beginning of the candidate
60 candAln = candAln.substr(i); // sequence
61 tempAln = tempAln.substr(i);
66 int pairwiseAlignmentLength = tempAln.length();
67 if(tempAln[pairwiseAlignmentLength-1] == '-'){ // we need to make sure that the candidate sequence alignment
68 for(int i=pairwiseAlignmentLength-1; i>=0; i--){// ends where the template sequence alignment ends, if it runs
69 if(isalpha(tempAln[i])){ // long, we nuke the end of the candidate sequence
70 candAln = candAln.substr(0,i+1);
71 tempAln = tempAln.substr(0,i+1);
79 candidateSeq->setPairwise(candAln); // set the pairwise sequences in the Sequence objects for
80 templateSeq->setPairwise(tempAln); // the candidate and template sequences
84 m->errorOut(e, "Nast", "pairwiseAlignSeqs");
89 /**************************************************************************************************/
91 void Nast::removeExtraGaps(string& candAln, string tempAln, string newTemplateAlign){
93 // here we do steps C-F of Fig. 2 from DeSantis et al.
96 int longAlignmentLength = newTemplateAlign.length();
98 for(int i=0; i<longAlignmentLength; i++){ // use the long alignment as the standard
99 int rightIndex, rightRoom, leftIndex, leftRoom;
101 // Part C of Fig. 2 from DeSantis et al.
102 if((isalpha(newTemplateAlign[i]) != isalpha(tempAln[i]))){ //if there is a discrepancy between the regapped
104 rightRoom = 0; leftRoom = 0;
106 // Part D of Fig. 2 from DeSantis et al. // template sequence and the official template sequence
107 for(leftIndex=i-1;leftIndex>0;leftIndex--){ // then we've got problems...
108 if(!isalpha(candAln[leftIndex])){
109 leftRoom = 1; //count how far it is to the nearest gap on the LEFT side of the anomaly
110 while(leftIndex-leftRoom>=0 && !isalpha(candAln[leftIndex-leftRoom])) { leftRoom++; }
115 for(rightIndex=i+1;rightIndex<longAlignmentLength-1;rightIndex++){
116 if(!isalpha(candAln[rightIndex])){
117 rightRoom = 1; //count how far it is to the nearest gap on the RIGHT side of the anomaly
118 while(rightIndex+rightRoom<longAlignmentLength && !isalpha(candAln[rightIndex+rightRoom])) { rightRoom++; }
123 int insertLength = 0; // figure out how long the anomaly is
124 while(!isalpha(newTemplateAlign[i + insertLength])) { insertLength++; }
125 if(insertLength > maxInsertLength){ maxInsertLength = insertLength; }
127 if((leftRoom + rightRoom) >= insertLength){
129 // Parts D & E from Fig. 2 of DeSantis et al.
130 if((i-leftIndex) <= (rightIndex-i)){ // the left gap is closer - > move stuff left there's
132 if(leftRoom >= insertLength){ // enough room to the left to move
133 //cout << "lr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << insertLength << endl;
134 string leftTemplateString = newTemplateAlign.substr(0,i);
135 string rightTemplateString = newTemplateAlign.substr((i+insertLength));
136 newTemplateAlign = leftTemplateString + rightTemplateString;
137 longAlignmentLength = newTemplateAlign.length();
138 //cout << "lr candAln = " << candAln.length() << '\t' << leftIndex << '\t' << endl;
139 string leftCandidateString = candAln.substr(0,(leftIndex-insertLength+1));
140 string rightCandidateString = candAln.substr((leftIndex+1));
141 candAln = leftCandidateString + rightCandidateString;
143 }else{ // not enough room to the left, have to steal some space to the right
145 //cout << "in else lr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << insertLength << endl;
146 string leftTemplateString = newTemplateAlign.substr(0,i);
147 string rightTemplateString = newTemplateAlign.substr((i+insertLength));
148 newTemplateAlign = leftTemplateString + rightTemplateString;
149 longAlignmentLength = newTemplateAlign.length();
150 //cout << " in else lr candAln = " << candAln.length() << '\t' << " leftIndex = " << leftIndex << " leftroom = " << leftRoom << " rightIndex = " << rightIndex << '\t' << " rightroom = " << rightRoom << '\t' << endl;
151 string leftCandidateString = candAln.substr(0,(leftIndex-leftRoom+1));
152 string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
153 string rightCandidateString = candAln.substr((rightIndex+(insertLength-leftRoom)));
154 candAln = leftCandidateString + insertString + rightCandidateString;
157 }else{ // the right gap is closer - > move stuff right there's
158 if(rightRoom >= insertLength){ // enough room to the right to move
159 //cout << "rr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
160 string leftTemplateString = newTemplateAlign.substr(0,i);
161 string rightTemplateString = newTemplateAlign.substr((i+insertLength));
162 newTemplateAlign = leftTemplateString + rightTemplateString;
163 longAlignmentLength = newTemplateAlign.length();
164 //cout << "rr candAln = " << candAln.length() << '\t' << i << '\t' << rightIndex << '\t' << rightIndex+insertLength << endl;
165 string leftCandidateString = candAln.substr(0,rightIndex);
166 string rightCandidateString = candAln.substr((rightIndex+insertLength));
167 candAln = leftCandidateString + rightCandidateString;
170 else{ // not enough room to the right, have to steal some
171 // space to the left lets move left and then right...
172 //cout << "in else rr newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << i+insertLength << endl;
173 string leftTemplateString = newTemplateAlign.substr(0,i);
174 string rightTemplateString = newTemplateAlign.substr((i+insertLength));
175 newTemplateAlign = leftTemplateString + rightTemplateString;
176 longAlignmentLength = newTemplateAlign.length();
177 //cout << "in else rr candAln = " << candAln.length() << '\t' << '\t' << (leftIndex-(insertLength-rightRoom)+1) << '\t' << (leftIndex+1,rightIndex-leftIndex-1) << '\t' << (rightIndex+rightRoom) << endl;
178 string leftCandidateString = candAln.substr(0,(leftIndex-(insertLength-rightRoom)+1));
179 string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
180 string rightCandidateString = candAln.substr((rightIndex+rightRoom));
181 candAln = leftCandidateString + insertString + rightCandidateString;
186 if ((i - insertLength) < 0) { i = 0; }
187 else { i -= insertLength; }
191 // there could be a case where there isn't enough room in either direction to move stuff
192 //cout << "in else else newTemplateAlign = " << newTemplateAlign.length() << '\t' << i << '\t' << (i+leftRoom+rightRoom) << endl;
193 string leftTemplateString = newTemplateAlign.substr(0,i);
194 string rightTemplateString = newTemplateAlign.substr((i+leftRoom+rightRoom));
195 newTemplateAlign = leftTemplateString + rightTemplateString;
196 longAlignmentLength = newTemplateAlign.length();
198 //cout << "in else else newTemplateAlign = " << candAln.length() << '\t' << (leftIndex-leftRoom+1) << '\t' << (leftIndex+1) << '\t' << (rightIndex-leftIndex-1) << '\t' << (rightIndex+rightRoom) << endl;
199 string leftCandidateString = candAln.substr(0,(leftIndex-leftRoom+1));
200 string insertString = candAln.substr((leftIndex+1),(rightIndex-leftIndex-1));
201 string rightCandidateString = candAln.substr((rightIndex+rightRoom));
202 candAln = leftCandidateString + insertString + rightCandidateString;
204 i -= (leftRoom + rightRoom);
207 // i -= insertLength;
209 //if i is negative, we want to remove the extra gaps to the right
210 if (i < 0) { m->mothurOut("i is negative"); m->mothurOutEndLine(); }
214 catch(exception& e) {
215 m->errorOut(e, "Nast", "removeExtraGaps");
220 /**************************************************************************************************/
222 void Nast::regapSequences(){ //This is essentially part B in Fig 2. of DeSantis et al.
224 //cout << candidateSeq->getName() << endl;
225 string candPair = candidateSeq->getPairwise();
228 string tempPair = templateSeq->getPairwise();
229 string tempAln = templateSeq->getAligned(); // we use the template aligned sequence as our guide
231 int pairwiseLength = candPair.length();
232 int fullAlignLength = tempAln.length();
235 for(int i=0;i<fullAlignLength;i++) { candAln += '.'; }
236 candidateSeq->setAligned(candAln);
240 int fullAlignIndex = 0;
241 int pairwiseAlignIndex = 0;
242 string newTemplateAlign = ""; // this is going to be messy so we want a temporary template
244 while(tempAln[fullAlignIndex] == '.' || tempAln[fullAlignIndex] == '-'){
245 candAln += '.'; // add the initial '-' and '.' to the candidate and template
246 newTemplateAlign += tempAln[fullAlignIndex];// pairwise sequences
250 string lastLoop = "";
252 while(pairwiseAlignIndex<pairwiseLength){
253 //cout << pairwiseAlignIndex << '\t' << fullAlignIndex << '\t' << pairwiseLength << endl;
254 if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
255 && isalpha(candPair[pairwiseAlignIndex])){
256 // the template and candidate pairwise and template aligned have characters
257 // need to add character onto the candidatSeq.aligned sequence
259 candAln += candPair[pairwiseAlignIndex];
260 newTemplateAlign += tempPair[pairwiseAlignIndex];//
262 pairwiseAlignIndex++;
265 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
266 && isalpha(candPair[pairwiseAlignIndex])){
267 // the template pairwise and candidate pairwise are characters and the template aligned is a gap
268 // need to insert gaps into the candidateSeq.aligned sequence
271 newTemplateAlign += '-';//
274 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
275 && isalpha(candPair[pairwiseAlignIndex])){
276 // the template pairwise is a gap and the template aligned and pairwise sequences have characters
277 // this is the alpha scenario. add character to the candidateSeq.aligned sequence without progressing
278 // further through the tempAln sequence.
280 candAln += candPair[pairwiseAlignIndex];
281 newTemplateAlign += '-';//
282 pairwiseAlignIndex++;
284 else if(isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
285 && !isalpha(candPair[pairwiseAlignIndex])){
286 // the template pairwise and full alignment are characters and the candidate sequence has a gap
287 // should not be a big deal, just add the gap position to the candidateSeq.aligned sequence;
289 candAln += candPair[pairwiseAlignIndex];
290 newTemplateAlign += tempAln[fullAlignIndex];//
292 pairwiseAlignIndex++;
294 else if(!isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
295 && isalpha(candPair[pairwiseAlignIndex])){
296 // the template pairwise and aligned are gaps while the candidate pairwise has a character
297 // this would be an insertion, go ahead and add the character->seems to be the opposite of the alpha scenario
299 candAln += candPair[pairwiseAlignIndex];
300 newTemplateAlign += tempAln[fullAlignIndex];//
301 pairwiseAlignIndex++;
304 else if(isalpha(tempPair[pairwiseAlignIndex]) && !isalpha(tempAln[fullAlignIndex])
305 && !isalpha(candPair[pairwiseAlignIndex])){
306 // template pairwise has a character, but its full aligned sequence and candidate sequence have gaps
307 // this would happen like we need to add a gap. basically the opposite of the alpha situation
309 newTemplateAlign += tempAln[fullAlignIndex];//
313 else if(!isalpha(tempPair[pairwiseAlignIndex]) && isalpha(tempAln[fullAlignIndex])
314 && !isalpha(candPair[pairwiseAlignIndex])){
315 // template and candidate pairwise are gaps and the template aligned is not a gap this should not be possible
316 // would skip the gaps and not progress through full alignment sequence
319 m->mothurOut("We're into D " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); m->mothurOutEndLine();
320 pairwiseAlignIndex++;
323 // everything has a gap - not possible
326 m->mothurOut("We're into F " + toString(fullAlignIndex) + " " + toString(pairwiseAlignIndex)); m->mothurOutEndLine();
327 pairwiseAlignIndex++;
332 for(int i=fullAlignIndex;i<fullAlignLength;i++){
334 newTemplateAlign += tempAln[i];//
338 int end = candAln.length()-1;
340 for(int i=0;i<candAln.length();i++){
341 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; } // if we padded the alignemnt from
342 else{ start = i; break; } // blast with Z's, change them to
345 for(int i=candAln.length()-1;i>=0;i--){ // ditto.
346 if(candAln[i] == 'Z' || !isalnum(candAln[i])) { candAln[i] = '.'; }
347 else{ end = i; break; }
350 for(int i=start;i<=end;i++){ // go through the candidate alignment sequence and make sure that
351 candAln[i] = toupper(candAln[i]); // everything is upper case
355 if(candAln.length() != tempAln.length()){ // if the regapped candidate sequence is longer than the official
356 removeExtraGaps(candAln, tempAln, newTemplateAlign);// template alignment then we need to do steps C-F in Fig.
357 } // 2 of Desantis et al.
359 candidateSeq->setAligned(candAln);
360 //cout << "here" << endl;
362 catch(exception& e) {
363 m->errorOut(e, "Nast", "regapSequences");
368 /**************************************************************************************************/
370 float Nast::getSimilarityScore(){
373 string cand = candidateSeq->getAligned();
374 string temp = templateSeq->getAligned();
375 int alignmentLength = temp.length();
379 for(int i=0;i<alignmentLength;i++){
380 if(cand[i] == '-' && temp[i] == '-'){
383 else if(cand[i] != '.' && temp[i] != '.'){
386 if(cand[i] != temp[i]){
391 float similarity = 100 * (1. - mismatch / (float)denominator);
392 if(denominator == 0){ similarity = 0.0000; }
397 catch(exception& e) {
398 m->errorOut(e, "Nast", "getSimilarityScore");
403 /**************************************************************************************************/
405 int Nast::getMaxInsertLength(){
407 return maxInsertLength;
411 /**************************************************************************************************/