1 // ***************************************************************************
\r
2 // BamWriter.cpp (c) 2009 Michael Str�mberg, Derek Barnett
\r
3 // Marth Lab, Department of Biology, Boston College
\r
4 // All rights reserved.
\r
5 // ---------------------------------------------------------------------------
\r
6 // Last modified: 17 August 2010 (DB)
\r
7 // ---------------------------------------------------------------------------
\r
8 // Uses BGZF routines were adapted from the bgzf.c code developed at the Broad
\r
10 // ---------------------------------------------------------------------------
\r
11 // Provides the basic functionality for producing BAM files
\r
12 // ***************************************************************************
\r
17 #include "BamWriter.h"
\r
18 using namespace BamTools;
\r
19 using namespace std;
\r
21 struct BamWriter::BamWriterPrivate {
\r
27 // constructor / destructor
\r
28 BamWriterPrivate(void) {
\r
29 IsBigEndian = SystemIsBigEndian();
\r
32 ~BamWriterPrivate(void) {
\r
36 // "public" interface
\r
38 bool Open(const string& filename, const string& samHeader, const RefVector& referenceSequences, bool isWriteUncompressed);
\r
39 void SaveAlignment(const BamAlignment& al);
\r
42 const unsigned int CalculateMinimumBin(const int begin, int end) const;
\r
43 void CreatePackedCigar(const vector<CigarOp>& cigarOperations, string& packedCigar);
\r
44 void EncodeQuerySequence(const string& query, string& encodedQuery);
\r
47 // -----------------------------------------------------
\r
48 // BamWriter implementation
\r
49 // -----------------------------------------------------
\r
52 BamWriter::BamWriter(void) {
\r
53 d = new BamWriterPrivate;
\r
57 BamWriter::~BamWriter(void) {
\r
62 // closes the alignment archive
\r
63 void BamWriter::Close(void) {
\r
67 // opens the alignment archive
\r
68 bool BamWriter::Open(const string& filename, const string& samHeader, const RefVector& referenceSequences, bool isWriteUncompressed) {
\r
69 return d->Open(filename, samHeader, referenceSequences, isWriteUncompressed);
\r
72 // saves the alignment to the alignment archive
\r
73 void BamWriter::SaveAlignment(const BamAlignment& al) {
\r
74 d->SaveAlignment(al);
\r
77 // -----------------------------------------------------
\r
78 // BamWriterPrivate implementation
\r
79 // -----------------------------------------------------
\r
81 // closes the alignment archive
\r
82 void BamWriter::BamWriterPrivate::Close(void) {
\r
86 // calculates minimum bin for a BAM alignment interval
\r
87 const unsigned int BamWriter::BamWriterPrivate::CalculateMinimumBin(const int begin, int end) const {
\r
89 if( (begin >> 14) == (end >> 14) ) return 4681 + (begin >> 14);
\r
90 if( (begin >> 17) == (end >> 17) ) return 585 + (begin >> 17);
\r
91 if( (begin >> 20) == (end >> 20) ) return 73 + (begin >> 20);
\r
92 if( (begin >> 23) == (end >> 23) ) return 9 + (begin >> 23);
\r
93 if( (begin >> 26) == (end >> 26) ) return 1 + (begin >> 26);
\r
97 // creates a cigar string from the supplied alignment
\r
98 void BamWriter::BamWriterPrivate::CreatePackedCigar(const vector<CigarOp>& cigarOperations, string& packedCigar) {
\r
101 const unsigned int numCigarOperations = cigarOperations.size();
\r
102 packedCigar.resize(numCigarOperations * BT_SIZEOF_INT);
\r
104 // pack the cigar data into the string
\r
105 unsigned int* pPackedCigar = (unsigned int*)packedCigar.data();
\r
107 unsigned int cigarOp;
\r
108 vector<CigarOp>::const_iterator coIter;
\r
109 for(coIter = cigarOperations.begin(); coIter != cigarOperations.end(); ++coIter) {
\r
111 switch(coIter->Type) {
\r
113 cigarOp = BAM_CMATCH;
\r
116 cigarOp = BAM_CINS;
\r
119 cigarOp = BAM_CDEL;
\r
122 cigarOp = BAM_CREF_SKIP;
\r
125 cigarOp = BAM_CSOFT_CLIP;
\r
128 cigarOp = BAM_CHARD_CLIP;
\r
131 cigarOp = BAM_CPAD;
\r
134 fprintf(stderr, "ERROR: Unknown cigar operation found: %c\n", coIter->Type);
\r
138 *pPackedCigar = coIter->Length << BAM_CIGAR_SHIFT | cigarOp;
\r
143 // encodes the supplied query sequence into 4-bit notation
\r
144 void BamWriter::BamWriterPrivate::EncodeQuerySequence(const string& query, string& encodedQuery) {
\r
146 // prepare the encoded query string
\r
147 const unsigned int queryLen = query.size();
\r
148 const unsigned int encodedQueryLen = (unsigned int)((queryLen / 2.0) + 0.5);
\r
149 encodedQuery.resize(encodedQueryLen);
\r
150 char* pEncodedQuery = (char*)encodedQuery.data();
\r
151 const char* pQuery = (const char*)query.data();
\r
153 unsigned char nucleotideCode;
\r
154 bool useHighWord = true;
\r
161 nucleotideCode = 0;
\r
165 nucleotideCode = 1;
\r
169 nucleotideCode = 2;
\r
173 nucleotideCode = 4;
\r
177 nucleotideCode = 8;
\r
181 nucleotideCode = 15;
\r
185 fprintf(stderr, "ERROR: Only the following bases are supported in the BAM format: {=, A, C, G, T, N}. Found [%c]\n", *pQuery);
\r
189 // pack the nucleotide code
\r
191 *pEncodedQuery = nucleotideCode << 4;
\r
192 useHighWord = false;
\r
194 *pEncodedQuery |= nucleotideCode;
\r
196 useHighWord = true;
\r
199 // increment the query position
\r
204 // opens the alignment archive
\r
205 bool BamWriter::BamWriterPrivate::Open(const string& filename, const string& samHeader, const RefVector& referenceSequences, bool isWriteUncompressed) {
\r
207 // open the BGZF file for writing, return failure if error
\r
208 if ( !mBGZF.Open(filename, "wb", isWriteUncompressed) )
\r
211 // ================
\r
212 // write the header
\r
213 // ================
\r
215 // write the BAM signature
\r
216 const unsigned char SIGNATURE_LENGTH = 4;
\r
217 const char* BAM_SIGNATURE = "BAM\1";
\r
218 mBGZF.Write(BAM_SIGNATURE, SIGNATURE_LENGTH);
\r
220 // write the SAM header text length
\r
221 uint32_t samHeaderLen = samHeader.size();
\r
222 if (IsBigEndian) SwapEndian_32(samHeaderLen);
\r
223 mBGZF.Write((char*)&samHeaderLen, BT_SIZEOF_INT);
\r
225 // write the SAM header text
\r
226 if(samHeaderLen > 0)
\r
227 mBGZF.Write(samHeader.data(), samHeaderLen);
\r
229 // write the number of reference sequences
\r
230 uint32_t numReferenceSequences = referenceSequences.size();
\r
231 if (IsBigEndian) SwapEndian_32(numReferenceSequences);
\r
232 mBGZF.Write((char*)&numReferenceSequences, BT_SIZEOF_INT);
\r
234 // =============================
\r
235 // write the sequence dictionary
\r
236 // =============================
\r
238 RefVector::const_iterator rsIter;
\r
239 for(rsIter = referenceSequences.begin(); rsIter != referenceSequences.end(); rsIter++) {
\r
241 // write the reference sequence name length
\r
242 uint32_t referenceSequenceNameLen = rsIter->RefName.size() + 1;
\r
243 if (IsBigEndian) SwapEndian_32(referenceSequenceNameLen);
\r
244 mBGZF.Write((char*)&referenceSequenceNameLen, BT_SIZEOF_INT);
\r
246 // write the reference sequence name
\r
247 mBGZF.Write(rsIter->RefName.c_str(), referenceSequenceNameLen);
\r
249 // write the reference sequence length
\r
250 int32_t referenceLength = rsIter->RefLength;
\r
251 if (IsBigEndian) SwapEndian_32(referenceLength);
\r
252 mBGZF.Write((char*)&referenceLength, BT_SIZEOF_INT);
\r
259 // saves the alignment to the alignment archive
\r
260 void BamWriter::BamWriterPrivate::SaveAlignment(const BamAlignment& al) {
\r
262 // if BamAlignment contains only the core data and a raw char data buffer
\r
263 // (as a result of BamReader::GetNextAlignmentCore())
\r
264 if ( al.SupportData.HasCoreOnly ) {
\r
266 // write the block size
\r
267 unsigned int blockSize = al.SupportData.BlockLength;
\r
268 if (IsBigEndian) SwapEndian_32(blockSize);
\r
269 mBGZF.Write((char*)&blockSize, BT_SIZEOF_INT);
\r
271 // assign the BAM core data
\r
272 uint32_t buffer[8];
\r
273 buffer[0] = al.RefID;
\r
274 buffer[1] = al.Position;
\r
275 buffer[2] = (al.Bin << 16) | (al.MapQuality << 8) | al.SupportData.QueryNameLength;
\r
276 buffer[3] = (al.AlignmentFlag << 16) | al.SupportData.NumCigarOperations;
\r
277 buffer[4] = al.SupportData.QuerySequenceLength;
\r
278 buffer[5] = al.MateRefID;
\r
279 buffer[6] = al.MatePosition;
\r
280 buffer[7] = al.InsertSize;
\r
282 // swap BAM core endian-ness, if necessary
\r
283 if ( IsBigEndian ) {
\r
284 for ( int i = 0; i < 8; ++i )
\r
285 SwapEndian_32(buffer[i]);
\r
288 // write the BAM core
\r
289 mBGZF.Write((char*)&buffer, BAM_CORE_SIZE);
\r
291 // write the raw char data
\r
292 mBGZF.Write((char*)al.SupportData.AllCharData.data(), al.SupportData.BlockLength-BAM_CORE_SIZE);
\r
295 // otherwise, BamAlignment should contain character in the standard fields: Name, QueryBases, etc
\r
296 // ( resulting from BamReader::GetNextAlignment() *OR* being generated directly by client code )
\r
299 // calculate char lengths
\r
300 const unsigned int nameLength = al.Name.size() + 1;
\r
301 const unsigned int numCigarOperations = al.CigarData.size();
\r
302 const unsigned int queryLength = al.QueryBases.size();
\r
303 const unsigned int tagDataLength = al.TagData.size();
\r
305 // no way to tell if BamAlignment.Bin is already defined (no default, invalid value)
\r
306 // force calculation of Bin before storing
\r
307 const int endPosition = al.GetEndPosition();
\r
308 const unsigned int alignmentBin = CalculateMinimumBin(al.Position, endPosition);
\r
310 // create our packed cigar string
\r
311 string packedCigar;
\r
312 CreatePackedCigar(al.CigarData, packedCigar);
\r
313 const unsigned int packedCigarLength = packedCigar.size();
\r
315 // encode the query
\r
316 string encodedQuery;
\r
317 EncodeQuerySequence(al.QueryBases, encodedQuery);
\r
318 const unsigned int encodedQueryLength = encodedQuery.size();
\r
320 // write the block size
\r
321 const unsigned int dataBlockSize = nameLength + packedCigarLength + encodedQueryLength + queryLength + tagDataLength;
\r
322 unsigned int blockSize = BAM_CORE_SIZE + dataBlockSize;
\r
323 if (IsBigEndian) SwapEndian_32(blockSize);
\r
324 mBGZF.Write((char*)&blockSize, BT_SIZEOF_INT);
\r
326 // assign the BAM core data
\r
327 uint32_t buffer[8];
\r
328 buffer[0] = al.RefID;
\r
329 buffer[1] = al.Position;
\r
330 buffer[2] = (alignmentBin << 16) | (al.MapQuality << 8) | nameLength;
\r
331 buffer[3] = (al.AlignmentFlag << 16) | numCigarOperations;
\r
332 buffer[4] = queryLength;
\r
333 buffer[5] = al.MateRefID;
\r
334 buffer[6] = al.MatePosition;
\r
335 buffer[7] = al.InsertSize;
\r
337 // swap BAM core endian-ness, if necessary
\r
338 if ( IsBigEndian ) {
\r
339 for ( int i = 0; i < 8; ++i )
\r
340 SwapEndian_32(buffer[i]);
\r
343 // write the BAM core
\r
344 mBGZF.Write((char*)&buffer, BAM_CORE_SIZE);
\r
346 // write the query name
\r
347 mBGZF.Write(al.Name.c_str(), nameLength);
\r
349 // write the packed cigar
\r
350 if ( IsBigEndian ) {
\r
352 char* cigarData = (char*)calloc(sizeof(char), packedCigarLength);
\r
353 memcpy(cigarData, packedCigar.data(), packedCigarLength);
\r
355 for (unsigned int i = 0; i < packedCigarLength; ++i) {
\r
357 SwapEndian_32p(&cigarData[i]);
\r
360 mBGZF.Write(cigarData, packedCigarLength);
\r
364 mBGZF.Write(packedCigar.data(), packedCigarLength);
\r
366 // write the encoded query sequence
\r
367 mBGZF.Write(encodedQuery.data(), encodedQueryLength);
\r
369 // write the base qualities
\r
370 string baseQualities(al.Qualities);
\r
371 char* pBaseQualities = (char*)al.Qualities.data();
\r
372 for(unsigned int i = 0; i < queryLength; i++) {
\r
373 pBaseQualities[i] -= 33;
\r
375 mBGZF.Write(pBaseQualities, queryLength);
\r
377 // write the read group tag
\r
378 if ( IsBigEndian ) {
\r
380 char* tagData = (char*)calloc(sizeof(char), tagDataLength);
\r
381 memcpy(tagData, al.TagData.data(), tagDataLength);
\r
384 while ( (unsigned int)i < tagDataLength ) {
\r
386 i += 2; // skip tag type (e.g. "RG", "NM", etc)
\r
387 uint8_t type = toupper(tagData[i]); // lower & upper case letters have same meaning
\r
388 ++i; // skip value type
\r
398 SwapEndian_16p(&tagData[i]);
\r
399 i+=2; // sizeof(uint16_t)
\r
404 SwapEndian_32p(&tagData[i]);
\r
405 i+=4; // sizeof(uint32_t)
\r
409 SwapEndian_64p(&tagData[i]);
\r
410 i+=8; // sizeof(uint64_t)
\r
415 while (tagData[i]) { ++i; }
\r
416 ++i; // increment one more for null terminator
\r
420 fprintf(stderr, "ERROR: Invalid tag value type\n"); // shouldn't get here
\r
426 mBGZF.Write(tagData, tagDataLength);
\r
430 mBGZF.Write(al.TagData.data(), tagDataLength);
\r