File: BandedSmithWaterman.cpp

package info (click to toggle)
libsmithwaterman 0.0+git20160702.2610e25-11
  • links: PTS, VCS
  • area: main
  • in suites: bullseye, sid
  • size: 296 kB
  • sloc: cpp: 2,115; makefile: 59; sh: 16
file content (670 lines) | stat: -rw-r--r-- 23,803 bytes parent folder | download | duplicates (5)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
#include "BandedSmithWaterman.h"

// define our static constants
const float CBandedSmithWaterman::FLOAT_NEGATIVE_INFINITY = (float)-1e+30;

const DirectionType CBandedSmithWaterman::Directions_STOP     = 0;
const DirectionType CBandedSmithWaterman::Directions_LEFT     = 1;
const DirectionType CBandedSmithWaterman::Directions_DIAGONAL = 2;
const DirectionType CBandedSmithWaterman::Directions_UP       = 3;

const PositionType CBandedSmithWaterman::Position_REF_AND_QUERY_ZERO    = 0;
const PositionType CBandedSmithWaterman::Position_REF_ZERO              = 1;
const PositionType CBandedSmithWaterman::Position_QUERY_ZERO            = 2;
const PositionType CBandedSmithWaterman::Position_REF_AND_QUERO_NONZERO = 3;

// constructor
CBandedSmithWaterman::CBandedSmithWaterman(float matchScore, float mismatchScore, float gapOpenPenalty, float gapExtendPenalty, unsigned int bandWidth) 
: mCurrentMatrixSize(0)
, mCurrentAnchorSize(0)
, mCurrentAQSumSize(0)
, mBandwidth(bandWidth)
, mPointers(NULL)
, mMatchScore(matchScore)
, mMismatchScore(mismatchScore)
, mGapOpenPenalty(gapOpenPenalty)
, mGapExtendPenalty(gapExtendPenalty)
, mAnchorGapScores(NULL)
, mBestScores(NULL)
, mReversedAnchor(NULL)
, mReversedQuery(NULL)
, mUseHomoPolymerGapOpenPenalty(false)
{
	CreateScoringMatrix();

	//if((bandWidth % 2) != 1) {
		//printf("ERROR: The bandwidth must be an odd number.\n");
		//exit(1);
	//}

	try {
		mBestScores	 = new float[bandWidth + 2];
		mAnchorGapScores = new float[bandWidth + 2];
	} catch(bad_alloc) {
		printf("ERROR: Unable to allocate enough memory for the banded Smith-Waterman algorithm.\n");
		exit(1);
	}
}

// destructor
CBandedSmithWaterman::~CBandedSmithWaterman(void) {
	if(mPointers)              delete [] mPointers;
	if(mAnchorGapScores)       delete [] mAnchorGapScores;
	if(mBestScores)            delete [] mBestScores;
	if(mReversedAnchor)        delete [] mReversedAnchor;
	if(mReversedQuery)         delete [] mReversedQuery;
}

// aligns the query sequence to the anchor using the Smith Waterman Gotoh algorithm
void CBandedSmithWaterman::Align(unsigned int& referenceAl, string& cigarAl, const string& s1, const string& s2, pair< pair<unsigned int, unsigned int>, pair<unsigned int, unsigned int> >& hr) {


	
	unsigned int rowStart = min(hr.first.first, (unsigned int)hr.second.first);
	hr.first.first    -= rowStart;
	hr.second.first   -= rowStart;
	
	//bool isLegalBandWidth = (s2.length() - hr.QueryBegin) > (mBandwidth / 2);
	//     isLegalBandWidth = isLegalBandWidth && ((s1.length() - hr.Begin) > (mBandwidth / 2));



	// check the lengths of the input sequences
	//if( (s1.length() <= 0) || (s2.length() <= 0) || (s1.length() < s2.length()) ) {
	//	printf("ERROR: An unexpected sequence length was encountered during pairwise alignment.\n");
	//	printf("Sequence lengths are listed as following:\n");
	//	printf("1. Reference length: %u\n2. Query length: %u\n", s1.length(), s2.length());
		//printf("3. Hash region in reference:%4u-%4u\n", hr.Begin + rowStart, hr.End);
		//printf("4. Hash region in query:    %4u-%4u\n", hr.QueryBegin + rowStart, hr.QueryEnd);
	//	exit(1);
	//}

	
	// determine the hash region type
	unsigned int rowOffset;
	unsigned int columnOffset;
	PositionType positionType;

	if(hr.first.first == 0) {
		if(hr.second.first == 0) {
			rowOffset    = 1;
			columnOffset = (mBandwidth / 2) + 1;
			positionType = Position_REF_AND_QUERY_ZERO;
		} else {
			rowOffset    = 1 - hr.second.first;
			columnOffset = (mBandwidth / 2) + 1 + hr.second.first;
			positionType = Position_REF_ZERO;
		}
	} else {
		if(hr.second.first == 0) {
			rowOffset    = 1;
			columnOffset = (mBandwidth / 2) + 1 - hr.first.first;
			positionType = Position_QUERY_ZERO;
		} else {
			rowOffset    = 1 - hr.second.first;
			columnOffset = (mBandwidth / 2) + 1 + hr.second.first - hr.first.first;
			positionType = Position_REF_AND_QUERO_NONZERO;
		}
	}

	// =========================
	// Reinitialize the matrices
	// =========================
	
	ReinitializeMatrices(positionType, s1.length(), s2.length(), hr);

	// =======================================
	// Banded Smith-Waterman forward algorithm
	// =======================================

	unsigned int bestColumn	= 0;
	unsigned int bestRow	= 0;
	float bestScore         = FLOAT_NEGATIVE_INFINITY;
	float currentQueryGapScore;

	// rowNum and column indicate the row and column numbers in the Smith-Waterman matrix respectively
	unsigned int rowNum    = hr.second.first;
	unsigned int columnNum = hr.first.first;

	// indicates how many rows including blank elements in the Banded SmithWaterman
	int numBlankElements = (mBandwidth / 2) - columnNum;

	//cout << numBlankElements << endl;
	// upper triangle matrix in Banded Smith-Waterman
	for( ; numBlankElements > 0; numBlankElements--, rowNum++){
		// in the upper triangle matrix, we always start at the 0th column
		columnNum = 0;

		// columnEnd indicates how many columns which should be dealt with in the current row
		unsigned int columnEnd = min((mBandwidth - numBlankElements), ((unsigned int) s1.length() - columnNum + 1) );
		currentQueryGapScore = FLOAT_NEGATIVE_INFINITY;
		for( unsigned int j = 0; j < columnEnd; j++){
			float score = CalculateScore(s1, s2, rowNum, columnNum, currentQueryGapScore, rowOffset, columnOffset);
			//cout << s1[columnNum] << s2[rowNum] << score << endl;
			UpdateBestScore(bestRow, bestColumn, bestScore, rowNum, columnNum, score);
			columnNum++;
		}

		// replace the columnNum to the middle column in the Smith-Waterman matrix
		columnNum = columnNum - (mBandwidth / 2);
	}
	// complete matrix in Banded Smith-Waterman
	unsigned int completeNum = min((s1.length() - columnNum - (mBandwidth / 2)), (s2.length() - rowNum));
	//cout << completeNum << endl;
	for(unsigned int i = 0; i < completeNum; i++, rowNum++){
		columnNum = columnNum - (mBandwidth / 2);

		// there are mBandwidth columns which should be dealt with in each row
		currentQueryGapScore = FLOAT_NEGATIVE_INFINITY;

		for(unsigned int j = 0; j < mBandwidth; j++){
			float score = CalculateScore(s1, s2, rowNum, columnNum, currentQueryGapScore, rowOffset, columnOffset);
			UpdateBestScore(bestRow, bestColumn, bestScore, rowNum, columnNum, score);
			//cout << s1[columnNum] << s2[rowNum] << score << endl;
			columnNum++;
		}

		// replace the columnNum to the middle column in the Smith-Waterman matrix
		// because mBandwidth is an odd number, everytime the following equation shifts a column (pluses 1).
		columnNum = columnNum - (mBandwidth / 2);
	}
	
	// lower triangle matrix
	numBlankElements = min(mBandwidth, ((unsigned int) s2.length() - rowNum));
	columnNum = columnNum - (mBandwidth / 2);
	for(unsigned int i = 0; numBlankElements > 0; i++, rowNum++, numBlankElements--) {

		mBestScores[ mBandwidth - i ] = FLOAT_NEGATIVE_INFINITY;;
		// columnEnd indicates how many columns which should be dealt with
		currentQueryGapScore = FLOAT_NEGATIVE_INFINITY;

		for( unsigned int j = columnNum; j < s1.length(); j++){
			float score = CalculateScore(s1, s2, rowNum, columnNum, currentQueryGapScore, rowOffset, columnOffset);
			UpdateBestScore(bestRow, bestColumn, bestScore, rowNum, columnNum, score);
			//cout << s1[columnNum] << s2[rowNum] << score << endl;
			columnNum++;
		}

		// replace the columnNum to the middle column in the Smith-Waterman matrix
		columnNum = columnNum - mBandwidth + i + 2;
	}

	// =========================================
	// Banded Smith-Waterman backtrace algorithm
	// =========================================

	Traceback(referenceAl, cigarAl, s1, s2, bestRow, bestColumn, rowOffset, columnOffset);

}

// calculates the score during the forward algorithm
float CBandedSmithWaterman::CalculateScore(const string& s1, const string& s2, const unsigned int rowNum, const unsigned int columnNum, float& currentQueryGapScore, const unsigned int rowOffset, const unsigned int columnOffset) {

	// initialize
	const unsigned int row      = rowNum + rowOffset;
	const unsigned int column   = columnOffset - rowNum + columnNum;
	const unsigned int position = row * (mBandwidth + 2) + column;

	// retrieve the similarity scores
	const float similarityScore      = mScoringMatrix[s1[columnNum] - 'A'][s2[rowNum] - 'A'];
	const float totalSimilarityScore = mBestScores[column] + similarityScore;

	// ================================
	// open a gap in the query sequence
	// ================================

	float queryGapExtendScore = currentQueryGapScore - mGapExtendPenalty;
	float queryGapOpenScore   = mBestScores[column - 1] - mGapOpenPenalty;

	// compute the homo-polymer gap score if enabled
	if(mUseHomoPolymerGapOpenPenalty)
		if((rowNum > 1) && (s2[rowNum] == s2[rowNum - 1]))
			queryGapOpenScore = mBestScores[column - 1] - mHomoPolymerGapOpenPenalty;

	if(queryGapExtendScore > queryGapOpenScore) {
		currentQueryGapScore = queryGapExtendScore;
		mPointers[position].mSizeOfHorizontalGaps = mPointers[position - 1].mSizeOfHorizontalGaps + 1;
	} else currentQueryGapScore = queryGapOpenScore;


	// ====================================
	// open a gap in the reference sequence
	// ====================================

	
	float anchorGapExtendScore = mAnchorGapScores[column + 1] - mGapExtendPenalty;
	float anchorGapOpenScore   = mBestScores[column + 1] - mGapOpenPenalty;

	// compute the homo-polymer gap score if enabled	
	if(mUseHomoPolymerGapOpenPenalty)
		if((columnNum > 1) && (s1[columnNum] == s1[columnNum - 1]))
			anchorGapOpenScore = mBestScores[column + 1] - mHomoPolymerGapOpenPenalty;

	if(anchorGapExtendScore > anchorGapOpenScore) {
		mAnchorGapScores[column] = anchorGapExtendScore;
		mPointers[position].mSizeOfVerticalGaps = mPointers[position - mBandwidth - 1].mSizeOfVerticalGaps + 1;
	} else mAnchorGapScores[column] = anchorGapOpenScore;

	// ======================================
	// calculate the best score and direction
	// ======================================

	//mBestScores[column] = MaxFloats(totalSimilarityScore, mAnchorGapScores[column], currentQueryGapScore);
	mBestScores[column] = MaxFloats(totalSimilarityScore, currentQueryGapScore, mAnchorGapScores[column]);

	// determine the traceback direction
	// diagonal (445364713) > stop (238960195) > up (214378647) > left (166504495)
	if(mBestScores[column] == 0)                            mPointers[position].Direction = Directions_STOP;
	else if(mBestScores[column] == totalSimilarityScore)    mPointers[position].Direction = Directions_UP;
	else if(mBestScores[column] == currentQueryGapScore)   mPointers[position].Direction = Directions_LEFT;
	else                                                    mPointers[position].Direction = Directions_DIAGONAL;

	return mBestScores[column];
}

// corrects the homopolymer gap order for forward alignments
void CBandedSmithWaterman::CorrectHomopolymerGapOrder(const unsigned int numBases, const unsigned int numMismatches) {


	// this is only required for alignments with mismatches
	//if(al.NumMismatches == 0) return;
	if ( numMismatches == 0 ) return;

	// localize the alignment data
	//char* pReference = al.Reference.Data();
	//char* pQuery     = al.Query.Data();
	//const unsigned int numBases = al.Reference.Length();
	char* pReference = mReversedAnchor;
	char* pQuery     = mReversedQuery;

	// initialize
	bool hasReferenceGap = false, hasQueryGap = false;
	char* pNonGapSeq = NULL;
	char* pGapSeq    = NULL;
	char nonGapBase  = 'J';

	// identify gapped regions
	for(unsigned int i = 0; i < numBases; i++) {

		// check if the current position is gapped
		hasReferenceGap = false;
		hasQueryGap     = false;

		if(pReference[i] == GAP) {
			hasReferenceGap = true;
			pNonGapSeq      = pQuery;
			pGapSeq         = pReference;
			nonGapBase      = pQuery[i];
		}

		if(pQuery[i] == GAP) {
			hasQueryGap = true;
			pNonGapSeq  = pReference;
			pGapSeq     = pQuery;
			nonGapBase  = pReference[i];
		}

		// continue if we don't have any gaps
		if(!hasReferenceGap && !hasQueryGap) continue;

		// sanity check
		if(hasReferenceGap && hasQueryGap) {
			printf("ERROR: Found a gap in both the reference sequence and query sequence.\n");
			exit(1);
		}

		// find the non-gapped length (forward)
		unsigned short numGappedBases = 0;
		unsigned short nonGapLength   = 0;
		unsigned short testPos = i;
		while(testPos < numBases) {

			const char gs  = pGapSeq[testPos];
			const char ngs = pNonGapSeq[testPos];

			bool isPartofHomopolymer = false;
			if(((gs == nonGapBase) || (gs == GAP)) && (ngs == nonGapBase)) isPartofHomopolymer = true;
			if(!isPartofHomopolymer) break;

			if(gs == GAP) numGappedBases++;
			else nonGapLength++;
			testPos++;
		}

		// fix the gap order
		if(numGappedBases != 0) {
			char* pCurrentSequence = pGapSeq + i;
			memset(pCurrentSequence, nonGapBase, nonGapLength);
			pCurrentSequence += nonGapLength;
			memset(pCurrentSequence, GAP, numGappedBases);
		}

		// increment
		i += numGappedBases + nonGapLength - 1;
	}
}

// creates a simple scoring matrix to align the nucleotides and the ambiguity code N
void CBandedSmithWaterman::CreateScoringMatrix(void) {

	unsigned int nIndex = 13;
	unsigned int xIndex = 23;

	// define the N score to be 1/4 of the span between mismatch and match
	//const short nScore = mMismatchScore + (short)(((mMatchScore - mMismatchScore) / 4.0) + 0.5);

	// calculate the scoring matrix
	for(unsigned char i = 0; i < MOSAIK_NUM_NUCLEOTIDES; i++) {
		for(unsigned char j = 0; j < MOSAIK_NUM_NUCLEOTIDES; j++) {

			// N.B. matching N to everything (while conceptually correct) leads to some
			// bad alignments, lets make N be a mismatch instead.

			// add the matches or mismatches to the hashtable (N is a mismatch)
			if((i == nIndex) || (j == nIndex)) mScoringMatrix[i][j] = mMismatchScore;
			else if((i == xIndex) || (j == xIndex)) mScoringMatrix[i][j] = mMismatchScore;
			else if(i == j) mScoringMatrix[i][j] = mMatchScore;
			else mScoringMatrix[i][j] = mMismatchScore;
		}
	}

	// add ambiguity codes
	mScoringMatrix['M' - 'A']['A' - 'A'] = mMatchScore;	// M - A
	mScoringMatrix['A' - 'A']['M' - 'A'] = mMatchScore;
	// add ambiguity codes
	mScoringMatrix['M' - 'A']['A' - 'A'] = mMatchScore;	// M - A
	mScoringMatrix['A' - 'A']['M' - 'A'] = mMatchScore;
	mScoringMatrix['M' - 'A']['C' - 'A'] = mMatchScore; // M - C
	mScoringMatrix['C' - 'A']['M' - 'A'] = mMatchScore;

	mScoringMatrix['R' - 'A']['A' - 'A'] = mMatchScore;	// R - A
	mScoringMatrix['A' - 'A']['R' - 'A'] = mMatchScore;
	mScoringMatrix['R' - 'A']['G' - 'A'] = mMatchScore; // R - G
	mScoringMatrix['G' - 'A']['R' - 'A'] = mMatchScore;

	mScoringMatrix['W' - 'A']['A' - 'A'] = mMatchScore;	// W - A
	mScoringMatrix['A' - 'A']['W' - 'A'] = mMatchScore;
	mScoringMatrix['W' - 'A']['T' - 'A'] = mMatchScore; // W - T
	mScoringMatrix['T' - 'A']['W' - 'A'] = mMatchScore;

	mScoringMatrix['S' - 'A']['C' - 'A'] = mMatchScore;	// S - C
	mScoringMatrix['C' - 'A']['S' - 'A'] = mMatchScore;
	mScoringMatrix['S' - 'A']['G' - 'A'] = mMatchScore; // S - G
	mScoringMatrix['G' - 'A']['S' - 'A'] = mMatchScore;

	mScoringMatrix['Y' - 'A']['C' - 'A'] = mMatchScore;	// Y - C
	mScoringMatrix['C' - 'A']['Y' - 'A'] = mMatchScore;
	mScoringMatrix['Y' - 'A']['T' - 'A'] = mMatchScore; // Y - T
	mScoringMatrix['T' - 'A']['Y' - 'A'] = mMatchScore;

	mScoringMatrix['K' - 'A']['G' - 'A'] = mMatchScore;	// K - G
	mScoringMatrix['G' - 'A']['K' - 'A'] = mMatchScore;
	mScoringMatrix['K' - 'A']['T' - 'A'] = mMatchScore; // K - T
	mScoringMatrix['T' - 'A']['K' - 'A'] = mMatchScore;

	mScoringMatrix['V' - 'A']['A' - 'A'] = mMatchScore;	// V - A
	mScoringMatrix['A' - 'A']['V' - 'A'] = mMatchScore;
	mScoringMatrix['V' - 'A']['C' - 'A'] = mMatchScore; // V - C
	mScoringMatrix['C' - 'A']['V' - 'A'] = mMatchScore;
	mScoringMatrix['V' - 'A']['G' - 'A'] = mMatchScore; // V - G
	mScoringMatrix['G' - 'A']['V' - 'A'] = mMatchScore;

	mScoringMatrix['H' - 'A']['A' - 'A'] = mMatchScore;	// H - A
	mScoringMatrix['A' - 'A']['H' - 'A'] = mMatchScore;
	mScoringMatrix['H' - 'A']['C' - 'A'] = mMatchScore; // H - C
	mScoringMatrix['C' - 'A']['H' - 'A'] = mMatchScore;
	mScoringMatrix['H' - 'A']['T' - 'A'] = mMatchScore; // H - T
	mScoringMatrix['T' - 'A']['H' - 'A'] = mMatchScore;

	mScoringMatrix['D' - 'A']['A' - 'A'] = mMatchScore;	// D - A
	mScoringMatrix['A' - 'A']['D' - 'A'] = mMatchScore;
	mScoringMatrix['D' - 'A']['G' - 'A'] = mMatchScore; // D - G
	mScoringMatrix['G' - 'A']['D' - 'A'] = mMatchScore;
	mScoringMatrix['D' - 'A']['T' - 'A'] = mMatchScore; // D - T
	mScoringMatrix['T' - 'A']['D' - 'A'] = mMatchScore;

	mScoringMatrix['B' - 'A']['C' - 'A'] = mMatchScore;	// B - C
	mScoringMatrix['C' - 'A']['B' - 'A'] = mMatchScore;
	mScoringMatrix['B' - 'A']['G' - 'A'] = mMatchScore; // B - G
	mScoringMatrix['G' - 'A']['B' - 'A'] = mMatchScore;
	mScoringMatrix['B' - 'A']['T' - 'A'] = mMatchScore; // B - T
	mScoringMatrix['T' - 'A']['B' - 'A'] = mMatchScore;
}

// enables homo-polymer scoring
void CBandedSmithWaterman::EnableHomoPolymerGapPenalty(float hpGapOpenPenalty) {
	mUseHomoPolymerGapOpenPenalty = true;
	mHomoPolymerGapOpenPenalty    = hpGapOpenPenalty;
}

// reinitializes the matrices
void CBandedSmithWaterman::ReinitializeMatrices(const PositionType& positionType, const unsigned int& s1Length, const unsigned int& s2Length, const pair< pair<unsigned int, unsigned int>, pair<unsigned int, unsigned int> > hr) {

/*
	try {
		mBestScores	 = new float[mBandwidth + 2];
		mAnchorGapScores = new float[mBandwidth + 2];
	} catch(bad_alloc) {
		printf("ERROR: Unable to allocate enough memory for the banded Smith-Waterman algorithm.\n");
		exit(1);
	}
*/

	const unsigned int numColumns = mBandwidth + 2;
	unsigned int numRows = 0;

	switch(positionType) {
		case Position_REF_AND_QUERY_ZERO:
			numRows = s2Length + 1;
			break;
		case Position_REF_ZERO:
			numRows = s2Length - hr.second.first + 2;
			break;
		case Position_QUERY_ZERO:
			numRows = min(s2Length + 1, s1Length - hr.first.first + 2);
			break;
		case Position_REF_AND_QUERO_NONZERO:
			numRows = min(s1Length - hr.first.first + 2, s2Length - hr.second.first + 2);
			break;
	}

	// update the size of the backtrace matrix
	if((numColumns * numRows) > mCurrentMatrixSize) {

		mCurrentMatrixSize = numColumns * numRows;
		if(mPointers) delete [] mPointers;

		try {
			mPointers = new ElementInfo[mCurrentMatrixSize];
		} catch(bad_alloc) {
			printf("ERROR: Unable to allocate enough memory for the banded Smith-Waterman algorithm.\n");
			exit(1);
		}
	}

	// initialize our backtrace matrix
	ElementInfo defaultElement;
	defaultElement.Direction = Directions_STOP;
	defaultElement.mSizeOfHorizontalGaps = 1;
	defaultElement.mSizeOfVerticalGaps   = 1;

	uninitialized_fill(mPointers, mPointers + mCurrentMatrixSize, defaultElement);

	// update the sequence character arrays
	if((s1Length + s2Length) > mCurrentAQSumSize) {

		mCurrentAQSumSize = s1Length + s2Length;
		if(mReversedAnchor)	delete [] mReversedAnchor;
		if(mReversedQuery)	delete [] mReversedQuery;

		try {
			mReversedAnchor	= new char[mCurrentAQSumSize + 1]; // reversed sequence #1
			mReversedQuery	= new char[mCurrentAQSumSize + 1]; // reversed sequence #2
		} catch(bad_alloc) {
			printf("ERROR: Unable to allocate enough memory for the banded Smith-Waterman algorithm.\n");
			exit(1);
		}
	}

	// initialize the gap score and score vectors
	uninitialized_fill(mAnchorGapScores, mAnchorGapScores + mBandwidth + 2, FLOAT_NEGATIVE_INFINITY);
	memset((char*)mBestScores, 0, SIZEOF_FLOAT * (mBandwidth + 2));
	mBestScores[0]              = FLOAT_NEGATIVE_INFINITY;
	mBestScores[mBandwidth + 1] = FLOAT_NEGATIVE_INFINITY;
}

// performs the backtrace algorithm
void CBandedSmithWaterman::Traceback(unsigned int& referenceAl, string& cigarAl, const string& s1, const string& s2, unsigned int bestRow, unsigned int bestColumn, const unsigned int rowOffset, const unsigned int columnOffset){


	unsigned int currentRow		 = bestRow;
	unsigned int currentColumn	 = bestColumn;
	unsigned int currentPosition = ((currentRow + rowOffset) * (mBandwidth + 2)) + (columnOffset - currentRow + currentColumn);


	// record the numbers of row and column before the current row and column
	unsigned int previousRow	= bestRow;
	unsigned int previousColumn	= bestColumn;

	unsigned int gappedAnchorLen = 0;
	unsigned int gappedQueryLen  = 0;
	unsigned int numMismatches   = 0;

	bool keepProcessing = true;
	while(keepProcessing) {
		unsigned int nVerticalGap = 0;
		unsigned int nHorizontalGap = 0;
		switch(mPointers[currentPosition].Direction){
			case Directions_DIAGONAL:
				nVerticalGap = mPointers[currentPosition].mSizeOfVerticalGaps;
				for(unsigned int i = 0; i < nVerticalGap; i++){
					mReversedAnchor[gappedAnchorLen++] = GAP;
					mReversedQuery[gappedQueryLen++]   = s2[currentRow];

					numMismatches++;

					previousRow = currentRow;
					previousColumn = currentColumn;

					currentRow--;
				}
				break;

			case Directions_STOP:
				keepProcessing = false;
				//mReversedAnchor[gappedAnchorLen+1]='\0';
				//mReversedQuery [gappedQueryLen+1]='\0';
				break;

			case Directions_UP:

				mReversedAnchor[gappedAnchorLen++] = s1[currentColumn];
				mReversedQuery[gappedQueryLen++]   = s2[currentRow];

				if(s1[currentColumn] != s2[currentRow]) numMismatches++;
				previousRow = currentRow;
				previousColumn = currentColumn;

				currentRow--;
				currentColumn--;
				break;

			case Directions_LEFT:
				nHorizontalGap =  mPointers[currentPosition].mSizeOfHorizontalGaps;
				for(unsigned int i = 0; i < nHorizontalGap; i++){

					mReversedAnchor[gappedAnchorLen++] = s1[currentColumn];
					mReversedQuery[gappedQueryLen++]   = GAP;

					numMismatches++;

					previousRow = currentRow;
					previousColumn = currentColumn;


					currentColumn--;
				}
				break;
		}
		currentPosition = ((currentRow + rowOffset) * (mBandwidth + 2)) + (columnOffset - currentRow + currentColumn);
	}

	// correct the reference and query sequence order
	mReversedAnchor[gappedAnchorLen] = 0;
	mReversedQuery [gappedQueryLen] = 0;
	reverse(mReversedAnchor, mReversedAnchor + gappedAnchorLen);
	reverse(mReversedQuery,  mReversedQuery  + gappedQueryLen);

	//alignment.Reference = mReversedAnchor;
	//alignment.Query     = mReversedQuery;

	// assign the alignment endpoints
	//alignment.ReferenceBegin = previousColumn;
	//alignment.ReferenceEnd   = bestColumn;
	referenceAl  = previousColumn;
	/*  
	if(alignment.IsReverseComplement){
		alignment.QueryBegin = s2.length() - bestRow - 1; 
		alignment.QueryEnd   = s2.length() - previousRow - 1;
	} else {
		alignment.QueryBegin = previousRow; 
		alignment.QueryEnd   = bestRow;
	}
	*/
	
	//alignment.QueryLength	= alignment.QueryEnd - alignment.QueryBegin + 1;
	//alignment.NumMismatches = numMismatches;

	const unsigned int alLength = strlen(mReversedAnchor);
	unsigned int m = 0, d = 0, i = 0;
	bool dashRegion = false;
	ostringstream oCigar;

	if ( previousRow != 0 )
		oCigar << previousRow << 'S';

	for ( unsigned int j = 0; j < alLength; j++ ) {
		// m
		if ( ( mReversedAnchor[j] != GAP ) && ( mReversedQuery[j] != GAP ) ) {
			if ( dashRegion ) {
				if ( d != 0 ) oCigar << d << 'D';
				else          oCigar << i << 'I';
			}
			dashRegion = false;
			m++;
			d = 0;
			i = 0;
		}
		// I or D
		else {
			if ( !dashRegion )
				oCigar << m << 'M';
			dashRegion = true;
			m = 0;
			if ( mReversedAnchor[j] == GAP ) {
				if ( d != 0 ) oCigar << d << 'D';
				i++;
				d = 0;
			}
			else {
				if ( i != 0 ) oCigar << i << 'I';
				d++;
				i = 0;
			}
		}
	}
	
	if      ( m != 0 ) oCigar << m << 'M';
	else if ( d != 0 ) oCigar << d << 'D';
	else if ( i != 0 ) oCigar << i << 'I';

	if ( ( bestRow + 1 ) != s2.length() )
		oCigar << s2.length() - bestRow - 1 << 'S';

	cigarAl = oCigar.str();
	

	// correct the homopolymer gap order
	CorrectHomopolymerGapOrder(alLength, numMismatches);

}