BamWriter.cpp

   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: 30 March 2010 (DB)\r
   7 // ---------------------------------------------------------------------------\r
   8 // Uses BGZF routines were adapted from the bgzf.c code developed at the Broad\r
   9 // Institute.\r
  10 // ---------------------------------------------------------------------------\r
  11 // Provides the basic functionality for producing BAM files\r
  12 // ***************************************************************************\r
  13 \r
  14 // BGZF includes\r
  15 #include "BGZF.h"\r
  16 #include "BamWriter.h"\r
  17 using namespace BamTools;\r
  18 using namespace std;\r
  19 \r
  20 struct BamWriter::BamWriterPrivate {\r
  21 \r
  22     // data members\r
  23     BgzfData mBGZF;\r
  24     bool IsBigEndian;\r
  25     \r
  26     \r
  27     // constructor / destructor\r
  28     BamWriterPrivate(void) { \r
  29       IsBigEndian = SystemIsBigEndian();  \r
  30     }\r
  31     \r
  32     ~BamWriterPrivate(void) {\r
  33         mBGZF.Close();\r
  34     }\r
  35 \r
  36     // "public" interface\r
  37     void Close(void);\r
  38     void Open(const std::string& filename, const std::string& samHeader, const BamTools::RefVector& referenceSequences);\r
  39     void SaveAlignment(const BamTools::BamAlignment& al);\r
  40 \r
  41     // internal methods\r
  42     void CreatePackedCigar(const std::vector<CigarOp>& cigarOperations, std::string& packedCigar);\r
  43     void EncodeQuerySequence(const std::string& query, std::string& encodedQuery);\r
  44 };\r
  45 \r
  46 // -----------------------------------------------------\r
  47 // BamWriter implementation\r
  48 // -----------------------------------------------------\r
  49 \r
  50 // constructor\r
  51 BamWriter::BamWriter(void) {\r
  52     d = new BamWriterPrivate;\r
  53 }\r
  54 \r
  55 // destructor\r
  56 BamWriter::~BamWriter(void) {\r
  57     delete d;\r
  58     d = 0;\r
  59 }\r
  60 \r
  61 // closes the alignment archive\r
  62 void BamWriter::Close(void) {\r
  63     d->Close();\r
  64 }\r
  65 \r
  66 // opens the alignment archive\r
  67 void BamWriter::Open(const string& filename, const string& samHeader, const RefVector& referenceSequences) {\r
  68     d->Open(filename, samHeader, referenceSequences);\r
  69 }\r
  70 \r
  71 // saves the alignment to the alignment archive\r
  72 void BamWriter::SaveAlignment(const BamAlignment& al) {\r
  73     d->SaveAlignment(al);\r
  74 }\r
  75 \r
  76 // -----------------------------------------------------\r
  77 // BamWriterPrivate implementation\r
  78 // -----------------------------------------------------\r
  79 \r
  80 // closes the alignment archive\r
  81 void BamWriter::BamWriterPrivate::Close(void) {\r
  82     mBGZF.Close();\r
  83 }\r
  84 \r
  85 // creates a cigar string from the supplied alignment\r
  86 void BamWriter::BamWriterPrivate::CreatePackedCigar(const vector<CigarOp>& cigarOperations, string& packedCigar) {\r
  87 \r
  88     // initialize\r
  89     const unsigned int numCigarOperations = cigarOperations.size();\r
  90     packedCigar.resize(numCigarOperations * BT_SIZEOF_INT);\r
  91 \r
  92     // pack the cigar data into the string\r
  93     unsigned int* pPackedCigar = (unsigned int*)packedCigar.data();\r
  94 \r
  95     unsigned int cigarOp;\r
  96     vector<CigarOp>::const_iterator coIter;\r
  97     for(coIter = cigarOperations.begin(); coIter != cigarOperations.end(); coIter++) {\r
  98 \r
  99         switch(coIter->Type) {\r
 100             case 'M':\r
 101                   cigarOp = BAM_CMATCH;\r
 102                   break;\r
 103             case 'I':\r
 104                   cigarOp = BAM_CINS;\r
 105                   break;\r
 106             case 'D':\r
 107                   cigarOp = BAM_CDEL;\r
 108                   break;\r
 109             case 'N':\r
 110                   cigarOp = BAM_CREF_SKIP;\r
 111                   break;\r
 112             case 'S':\r
 113                   cigarOp = BAM_CSOFT_CLIP;\r
 114                   break;\r
 115             case 'H':\r
 116                   cigarOp = BAM_CHARD_CLIP;\r
 117                   break;\r
 118             case 'P':\r
 119                   cigarOp = BAM_CPAD;\r
 120                   break;\r
 121             default:\r
 122                   printf("ERROR: Unknown cigar operation found: %c\n", coIter->Type);\r
 123                   exit(1);\r
 124         }\r
 125 \r
 126         *pPackedCigar = coIter->Length << BAM_CIGAR_SHIFT | cigarOp;\r
 127         pPackedCigar++;\r
 128     }\r
 129 }\r
 130 \r
 131 // encodes the supplied query sequence into 4-bit notation\r
 132 void BamWriter::BamWriterPrivate::EncodeQuerySequence(const string& query, string& encodedQuery) {\r
 133 \r
 134     // prepare the encoded query string\r
 135     const unsigned int queryLen = query.size();\r
 136     const unsigned int encodedQueryLen = (unsigned int)((queryLen / 2.0) + 0.5);\r
 137     encodedQuery.resize(encodedQueryLen);\r
 138     char* pEncodedQuery = (char*)encodedQuery.data();\r
 139     const char* pQuery = (const char*)query.data();\r
 140 \r
 141     unsigned char nucleotideCode;\r
 142     bool useHighWord = true;\r
 143 \r
 144     while(*pQuery) {\r
 145 \r
 146         switch(*pQuery) {\r
 147             \r
 148             case '=':\r
 149                 nucleotideCode = 0;\r
 150                 break;\r
 151                 \r
 152             case 'A':\r
 153                 nucleotideCode = 1;\r
 154                 break;\r
 155             \r
 156             case 'C':\r
 157                 nucleotideCode = 2;\r
 158                 break;\r
 159             \r
 160             case 'G':\r
 161                 nucleotideCode = 4;\r
 162                 break;\r
 163             \r
 164             case 'T':\r
 165                 nucleotideCode = 8;\r
 166                 break;\r
 167             \r
 168             case 'N':\r
 169                 nucleotideCode = 15;\r
 170                 break;\r
 171             \r
 172             default:\r
 173                 printf("ERROR: Only the following bases are supported in the BAM format: {=, A, C, G, T, N}. Found [%c]\n", *pQuery);\r
 174                 exit(1);\r
 175         }\r
 176 \r
 177         // pack the nucleotide code\r
 178         if(useHighWord) {\r
 179             *pEncodedQuery = nucleotideCode << 4;\r
 180             useHighWord = false;\r
 181         } else {\r
 182             *pEncodedQuery |= nucleotideCode;\r
 183             pEncodedQuery++;\r
 184             useHighWord = true;\r
 185         }\r
 186 \r
 187         // increment the query position\r
 188         pQuery++;\r
 189     }\r
 190 }\r
 191 \r
 192 // opens the alignment archive\r
 193 void BamWriter::BamWriterPrivate::Open(const string& filename, const string& samHeader, const RefVector& referenceSequences) {\r
 194 \r
 195     // open the BGZF file for writing\r
 196     mBGZF.Open(filename, "wb");\r
 197 \r
 198     // ================\r
 199     // write the header\r
 200     // ================\r
 201 \r
 202     // write the BAM signature\r
 203     const unsigned char SIGNATURE_LENGTH = 4;\r
 204     const char* BAM_SIGNATURE = "BAM\1";\r
 205     mBGZF.Write(BAM_SIGNATURE, SIGNATURE_LENGTH);\r
 206 \r
 207     // write the SAM header text length\r
 208     uint32_t samHeaderLen = samHeader.size();\r
 209     if ( IsBigEndian ) { SwapEndian_32(samHeaderLen); }\r
 210     mBGZF.Write((char*)&samHeaderLen, BT_SIZEOF_INT);\r
 211 \r
 212     // write the SAM header text\r
 213     if(samHeaderLen > 0) {\r
 214         mBGZF.Write(samHeader.data(), samHeaderLen);\r
 215     }\r
 216 \r
 217     // write the number of reference sequences\r
 218     uint32_t numReferenceSequences = referenceSequences.size();\r
 219     if ( IsBigEndian ) { SwapEndian_32(numReferenceSequences); }\r
 220     mBGZF.Write((char*)&numReferenceSequences, BT_SIZEOF_INT);\r
 221 \r
 222     // =============================\r
 223     // write the sequence dictionary\r
 224     // =============================\r
 225 \r
 226     RefVector::const_iterator rsIter;\r
 227     for(rsIter = referenceSequences.begin(); rsIter != referenceSequences.end(); rsIter++) {\r
 228 \r
 229         // write the reference sequence name length\r
 230         uint32_t referenceSequenceNameLen = rsIter->RefName.size() + 1;\r
 231         if ( IsBigEndian ) { SwapEndian_32(referenceSequenceNameLen); }\r
 232         mBGZF.Write((char*)&referenceSequenceNameLen, BT_SIZEOF_INT);\r
 233 \r
 234         // write the reference sequence name\r
 235         mBGZF.Write(rsIter->RefName.c_str(), referenceSequenceNameLen);\r
 236 \r
 237         // write the reference sequence length\r
 238         int32_t referenceLength = rsIter->RefLength;\r
 239         if ( IsBigEndian ) { SwapEndian_32(referenceLength); }\r
 240         mBGZF.Write((char*)&referenceLength, BT_SIZEOF_INT);\r
 241     }\r
 242 }\r
 243 \r
 244 // saves the alignment to the alignment archive\r
 245 void BamWriter::BamWriterPrivate::SaveAlignment(const BamAlignment& al) {\r
 246 \r
 247     // initialize\r
 248     const unsigned int nameLen            = al.Name.size() + 1;\r
 249     const unsigned int queryLen           = al.QueryBases.size();\r
 250     const unsigned int numCigarOperations = al.CigarData.size();\r
 251 \r
 252     // create our packed cigar string\r
 253     string packedCigar;\r
 254     CreatePackedCigar(al.CigarData, packedCigar);\r
 255     const unsigned int packedCigarLen = packedCigar.size();\r
 256 \r
 257     // encode the query\r
 258     string encodedQuery;\r
 259     EncodeQuerySequence(al.QueryBases, encodedQuery);\r
 260     const unsigned int encodedQueryLen = encodedQuery.size();\r
 261 \r
 262     // store the tag data length\r
 263     // -------------------------------------------\r
 264     // Modified: 3-25-2010 DWB\r
 265     // Contributed: ARQ\r
 266     // Fixed: "off by one" error when parsing tags in files produced by BamWriter\r
 267     const unsigned int tagDataLength = al.TagData.size();\r
 268     // original line: \r
 269     // const unsigned int tagDataLength = al.TagData.size() + 1;     \r
 270     // -------------------------------------------\r
 271     \r
 272     // assign the BAM core data\r
 273     uint32_t buffer[8];\r
 274     buffer[0] = al.RefID;\r
 275     buffer[1] = al.Position;\r
 276     buffer[2] = (al.Bin << 16) | (al.MapQuality << 8) | nameLen;\r
 277     buffer[3] = (al.AlignmentFlag << 16) | numCigarOperations;\r
 278     buffer[4] = queryLen;\r
 279     buffer[5] = al.MateRefID;\r
 280     buffer[6] = al.MatePosition;\r
 281     buffer[7] = al.InsertSize;\r
 282 \r
 283     // write the block size\r
 284     unsigned int dataBlockSize = nameLen + packedCigarLen + encodedQueryLen + queryLen + tagDataLength;\r
 285     unsigned int blockSize = BAM_CORE_SIZE + dataBlockSize;\r
 286     if ( IsBigEndian ) { SwapEndian_32(blockSize); }\r
 287     mBGZF.Write((char*)&blockSize, BT_SIZEOF_INT);\r
 288 \r
 289     // write the BAM core\r
 290     if ( IsBigEndian ) { \r
 291         for ( int i = 0; i < 8; ++i ) { \r
 292             SwapEndian_32(buffer[i]); \r
 293         } \r
 294     }\r
 295     mBGZF.Write((char*)&buffer, BAM_CORE_SIZE);\r
 296 \r
 297     // write the query name\r
 298     mBGZF.Write(al.Name.c_str(), nameLen);\r
 299 \r
 300     // write the packed cigar\r
 301     if ( IsBigEndian ) {\r
 302       \r
 303         char* cigarData = (char*)calloc(sizeof(char), packedCigarLen);\r
 304         memcpy(cigarData, packedCigar.data(), packedCigarLen);\r
 305         \r
 306         for (unsigned int i = 0; i < packedCigarLen; ++i) {\r
 307             if ( IsBigEndian ) { \r
 308               SwapEndian_32p(&cigarData[i]); \r
 309             }\r
 310         }\r
 311         \r
 312         mBGZF.Write(cigarData, packedCigarLen);\r
 313         free(cigarData);\r
 314         \r
 315     } else { \r
 316         mBGZF.Write(packedCigar.data(), packedCigarLen);\r
 317     }\r
 318 \r
 319     // write the encoded query sequence\r
 320     mBGZF.Write(encodedQuery.data(), encodedQueryLen);\r
 321 \r
 322     // write the base qualities\r
 323     string baseQualities = al.Qualities;\r
 324     char* pBaseQualities = (char*)al.Qualities.data();\r
 325     for(unsigned int i = 0; i < queryLen; i++) { pBaseQualities[i] -= 33; }\r
 326     mBGZF.Write(pBaseQualities, queryLen);\r
 327 \r
 328     // write the read group tag\r
 329     if ( IsBigEndian ) {\r
 330       \r
 331         char* tagData = (char*)calloc(sizeof(char), tagDataLength);\r
 332         memcpy(tagData, al.TagData.data(), tagDataLength);\r
 333       \r
 334         int i = 0;\r
 335         while ( (unsigned int)i < tagDataLength ) {\r
 336             \r
 337             i += 2;                                 // skip tag type (e.g. "RG", "NM", etc)\r
 338             uint8_t type = toupper(tagData[i]);     // lower & upper case letters have same meaning \r
 339             ++i;                                    // skip value type\r
 340             \r
 341             switch (type) {\r
 342               \r
 343                 case('A') :\r
 344                 case('C') : \r
 345                     ++i;\r
 346                     break;\r
 347                             \r
 348                 case('S') : \r
 349                     SwapEndian_16p(&tagData[i]); \r
 350                     i+=2; // sizeof(uint16_t)\r
 351                     break;\r
 352                             \r
 353                 case('F') :\r
 354                 case('I') : \r
 355                     SwapEndian_32p(&tagData[i]);\r
 356                     i+=4; // sizeof(uint32_t)\r
 357                     break;\r
 358                             \r
 359                 case('D') : \r
 360                     SwapEndian_64p(&tagData[i]);\r
 361                     i+=8; // sizeof(uint64_t)\r
 362                     break;\r
 363                             \r
 364                 case('H') :\r
 365                 case('Z') : \r
 366                     while (tagData[i]) { ++i; }\r
 367                     ++i; // increment one more for null terminator\r
 368                     break;\r
 369                             \r
 370                 default : \r
 371                     printf("ERROR: Invalid tag value type\n"); // shouldn't get here\r
 372                     free(tagData);\r
 373                     exit(1); \r
 374             }\r
 375         }\r
 376         \r
 377         mBGZF.Write(tagData, tagDataLength);\r
 378         free(tagData);\r
 379     } else {\r
 380         mBGZF.Write(al.TagData.data(), tagDataLength);\r
 381     }\r
 382 }\r