3 Copyright (c) 2008-2010 Genome Research Ltd (GRL).
5 Permission is hereby granted, free of charge, to any person obtaining
6 a copy of this software and associated documentation files (the
7 "Software"), to deal in the Software without restriction, including
8 without limitation the rights to use, copy, modify, merge, publish,
9 distribute, sublicense, and/or sell copies of the Software, and to
10 permit persons to whom the Software is furnished to do so, subject to
11 the following conditions:
13 The above copyright notice and this permission notice shall be
14 included in all copies or substantial portions of the Software.
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
17 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
19 NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
20 BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
21 ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
26 /* Contact: Heng Li <lh3@sanger.ac.uk> */
34 BAM library provides I/O and various operations on manipulating files
35 in the BAM (Binary Alignment/Mapping) or SAM (Sequence Alignment/Map)
36 format. It now supports importing from or exporting to TAM, sorting,
37 merging, generating pileup, and quickly retrieval of reads overlapped
38 with a specified region.
40 @copyright Genome Research Ltd.
49 #define BAM_VIRTUAL_OFFSET16
51 /*! @abstract BAM file handler */
52 typedef BGZF *bamFile;
53 #define bam_open(fn, mode) bgzf_open(fn, mode)
54 #define bam_dopen(fd, mode) bgzf_fdopen(fd, mode)
55 #define bam_close(fp) bgzf_close(fp)
56 #define bam_read(fp, buf, size) bgzf_read(fp, buf, size)
57 #define bam_write(fp, buf, size) bgzf_write(fp, buf, size)
58 #define bam_tell(fp) bgzf_tell(fp)
59 #define bam_seek(fp, pos, dir) bgzf_seek(fp, pos, dir)
61 #define BAM_TRUE_OFFSET
63 typedef gzFile bamFile;
64 #define bam_open(fn, mode) gzopen(fn, mode)
65 #define bam_dopen(fd, mode) gzdopen(fd, mode)
66 #define bam_close(fp) gzclose(fp)
67 #define bam_read(fp, buf, size) gzread(fp, buf, size)
68 /* no bam_write/bam_tell/bam_seek() here */
72 @abstract Structure for the alignment header.
73 @field n_targets number of reference sequences
74 @field target_name names of the reference sequences
75 @field target_len lengths of the referene sequences
76 @field dict header dictionary
77 @field hash hash table for fast name lookup
78 @field rg2lib hash table for @RG-ID -> LB lookup
79 @field l_text length of the plain text in the header
80 @field text plain text
82 @discussion Field hash points to null by default. It is a private
89 void *dict, *hash, *rg2lib;
90 size_t l_text, n_text;
94 /*! @abstract the read is paired in sequencing, no matter whether it is mapped in a pair */
96 /*! @abstract the read is mapped in a proper pair */
97 #define BAM_FPROPER_PAIR 2
98 /*! @abstract the read itself is unmapped; conflictive with BAM_FPROPER_PAIR */
100 /*! @abstract the mate is unmapped */
101 #define BAM_FMUNMAP 8
102 /*! @abstract the read is mapped to the reverse strand */
103 #define BAM_FREVERSE 16
104 /*! @abstract the mate is mapped to the reverse strand */
105 #define BAM_FMREVERSE 32
106 /*! @abstract this is read1 */
107 #define BAM_FREAD1 64
108 /*! @abstract this is read2 */
109 #define BAM_FREAD2 128
110 /*! @abstract not primary alignment */
111 #define BAM_FSECONDARY 256
112 /*! @abstract QC failure */
113 #define BAM_FQCFAIL 512
114 /*! @abstract optical or PCR duplicate */
115 #define BAM_FDUP 1024
121 /*! @abstract defautl mask for pileup */
122 #define BAM_DEF_MASK (BAM_FUNMAP | BAM_FSECONDARY | BAM_FQCFAIL | BAM_FDUP)
124 #define BAM_CORE_SIZE sizeof(bam1_core_t)
127 * Describing how CIGAR operation/length is packed in a 32-bit integer.
129 #define BAM_CIGAR_SHIFT 4
130 #define BAM_CIGAR_MASK ((1 << BAM_CIGAR_SHIFT) - 1)
135 /*! @abstract CIGAR: match */
137 /*! @abstract CIGAR: insertion to the reference */
139 /*! @abstract CIGAR: deletion from the reference */
141 /*! @abstract CIGAR: skip on the reference (e.g. spliced alignment) */
142 #define BAM_CREF_SKIP 3
143 /*! @abstract CIGAR: clip on the read with clipped sequence present in qseq */
144 #define BAM_CSOFT_CLIP 4
145 /*! @abstract CIGAR: clip on the read with clipped sequence trimmed off */
146 #define BAM_CHARD_CLIP 5
147 /*! @abstract CIGAR: padding */
151 @abstract Structure for core alignment information.
152 @field tid chromosome ID, defined by bam_header_t
153 @field pos 0-based leftmost coordinate
154 @field strand strand; 0 for forward and 1 otherwise
155 @field bin bin calculated by bam_reg2bin()
156 @field qual mapping quality
157 @field l_qname length of the query name
158 @field flag bitwise flag
159 @field n_cigar number of CIGAR operations
160 @field l_qseq length of the query sequence (read)
165 uint32_t bin:16, qual:8, l_qname:8;
166 uint32_t flag:16, n_cigar:16;
174 @abstract Structure for one alignment.
175 @field core core information about the alignment
176 @field l_aux length of auxiliary data
177 @field data_len current length of bam1_t::data
178 @field m_data maximum length of bam1_t::data
179 @field data all variable-length data, concatenated; structure: cigar-qname-seq-qual-aux
183 1. qname is zero tailing and core.l_qname includes the tailing '\0'.
184 2. l_qseq is calculated from the total length of an alignment block
185 on reading or from CIGAR.
189 int l_aux, data_len, m_data;
193 typedef struct __bam_iter_t *bam_iter_t;
195 #define bam1_strand(b) (((b)->core.flag&BAM_FREVERSE) != 0)
196 #define bam1_mstrand(b) (((b)->core.flag&BAM_FMREVERSE) != 0)
199 @abstract Get the CIGAR array
200 @param b pointer to an alignment
201 @return pointer to the CIGAR array
203 @discussion In the CIGAR array, each element is a 32-bit integer. The
204 lower 4 bits gives a CIGAR operation and the higher 28 bits keep the
207 #define bam1_cigar(b) ((uint32_t*)((b)->data + (b)->core.l_qname))
210 @abstract Get the name of the query
211 @param b pointer to an alignment
212 @return pointer to the name string, null terminated
214 #define bam1_qname(b) ((char*)((b)->data))
217 @abstract Get query sequence
218 @param b pointer to an alignment
219 @return pointer to sequence
221 @discussion Each base is encoded in 4 bits: 1 for A, 2 for C, 4 for G,
222 8 for T and 15 for N. Two bases are packed in one byte with the base
223 at the higher 4 bits having smaller coordinate on the read. It is
224 recommended to use bam1_seqi() macro to get the base.
226 #define bam1_seq(b) ((b)->data + (b)->core.n_cigar*4 + (b)->core.l_qname)
229 @abstract Get query quality
230 @param b pointer to an alignment
231 @return pointer to quality string
233 #define bam1_qual(b) ((b)->data + (b)->core.n_cigar*4 + (b)->core.l_qname + ((b)->core.l_qseq + 1)/2)
236 @abstract Get a base on read
237 @param s Query sequence returned by bam1_seq()
238 @param i The i-th position, 0-based
239 @return 4-bit integer representing the base.
241 #define bam1_seqi(s, i) ((s)[(i)/2] >> 4*(1-(i)%2) & 0xf)
244 @abstract Get query sequence and quality
245 @param b pointer to an alignment
246 @return pointer to the concatenated auxiliary data
248 #define bam1_aux(b) ((b)->data + (b)->core.n_cigar*4 + (b)->core.l_qname + (b)->core.l_qseq + ((b)->core.l_qseq + 1)/2)
252 @abstract Round an integer to the next closest power-2 integer.
253 @param x integer to be rounded (in place)
254 @discussion x will be modified.
256 #define kroundup32(x) (--(x), (x)|=(x)>>1, (x)|=(x)>>2, (x)|=(x)>>4, (x)|=(x)>>8, (x)|=(x)>>16, ++(x))
260 @abstract Whether the machine is big-endian; modified only in
263 extern int bam_is_be;
265 /*! @abstract Table for converting a nucleotide character to the 4-bit encoding. */
266 extern unsigned char bam_nt16_table[256];
268 /*! @abstract Table for converting a 4-bit encoded nucleotide to a letter. */
269 extern char *bam_nt16_rev_table;
271 extern char bam_nt16_nt4_table[];
277 /*********************
278 * Low-level SAM I/O *
279 *********************/
281 /*! @abstract TAM file handler */
282 typedef struct __tamFile_t *tamFile;
285 @abstract Open a SAM file for reading, either uncompressed or compressed by gzip/zlib.
286 @param fn SAM file name
287 @return SAM file handler
289 tamFile sam_open(const char *fn);
292 @abstract Close a SAM file handler
293 @param fp SAM file handler
295 void sam_close(tamFile fp);
298 @abstract Read one alignment from a SAM file handler
299 @param fp SAM file handler
300 @param header header information (ordered names of chromosomes)
301 @param b read alignment; all members in b will be updated
302 @return 0 if successful; otherwise negative
304 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b);
307 @abstract Read header information from a TAB-delimited list file.
308 @param fn_list file name for the list
309 @return a pointer to the header structure
311 @discussion Each line in this file consists of chromosome name and
312 the length of chromosome.
314 bam_header_t *sam_header_read2(const char *fn_list);
317 @abstract Read header from a SAM file (if present)
318 @param fp SAM file handler
319 @return pointer to header struct; 0 if no @SQ lines available
321 bam_header_t *sam_header_read(tamFile fp);
324 @abstract Parse @SQ lines a update a header struct
325 @param h pointer to the header struct to be updated
326 @return number of target sequences
328 @discussion bam_header_t::{n_targets,target_len,target_name} will
329 be destroyed in the first place.
331 int sam_header_parse(bam_header_t *h);
332 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
335 @abstract Parse @RG lines a update a header struct
336 @param h pointer to the header struct to be updated
337 @return number of @RG lines
339 @discussion bam_header_t::rg2lib will be destroyed in the first
342 int sam_header_parse_rg(bam_header_t *h);
344 #define sam_write1(header, b) bam_view1(header, b)
347 /********************************
348 * APIs for string dictionaries *
349 ********************************/
351 int bam_strmap_put(void *strmap, const char *rg, const char *lib);
352 const char *bam_strmap_get(const void *strmap, const char *rg);
353 void *bam_strmap_dup(const void*);
354 void *bam_strmap_init();
355 void bam_strmap_destroy(void *strmap);
358 /*********************
359 * Low-level BAM I/O *
360 *********************/
363 @abstract Initialize a header structure.
364 @return the pointer to the header structure
366 @discussion This function also modifies the global variable
369 bam_header_t *bam_header_init();
372 @abstract Destroy a header structure.
373 @param header pointer to the header
375 void bam_header_destroy(bam_header_t *header);
378 @abstract Read a header structure from BAM.
379 @param fp BAM file handler, opened by bam_open()
380 @return pointer to the header structure
382 @discussion The file position indicator must be placed at the
383 beginning of the file. Upon success, the position indicator will
384 be set at the start of the first alignment.
386 bam_header_t *bam_header_read(bamFile fp);
389 @abstract Write a header structure to BAM.
390 @param fp BAM file handler
391 @param header pointer to the header structure
392 @return always 0 currently
394 int bam_header_write(bamFile fp, const bam_header_t *header);
397 @abstract Read an alignment from BAM.
398 @param fp BAM file handler
399 @param b read alignment; all members are updated.
400 @return number of bytes read from the file
402 @discussion The file position indicator must be
403 placed right before an alignment. Upon success, this function
404 will set the position indicator to the start of the next
405 alignment. This function is not affected by the machine
408 int bam_read1(bamFile fp, bam1_t *b);
411 @abstract Write an alignment to BAM.
412 @param fp BAM file handler
413 @param c pointer to the bam1_core_t structure
414 @param data_len total length of variable size data related to
416 @param data pointer to the concatenated data
417 @return number of bytes written to the file
419 @discussion This function is not affected by the machine
422 int bam_write1_core(bamFile fp, const bam1_core_t *c, int data_len, uint8_t *data);
425 @abstract Write an alignment to BAM.
426 @param fp BAM file handler
427 @param b alignment to write
428 @return number of bytes written to the file
430 @abstract It is equivalent to:
431 bam_write1_core(fp, &b->core, b->data_len, b->data)
433 int bam_write1(bamFile fp, const bam1_t *b);
436 @abstract Initiate a pointer to bam1_t struct
438 #define bam_init1() ((bam1_t*)calloc(1, sizeof(bam1_t)))
441 @abstract Free the memory allocated for an alignment.
442 @param b pointer to an alignment
444 #define bam_destroy1(b) do { \
445 if (b) { free((b)->data); free(b); } \
449 @abstract Format a BAM record in the SAM format
450 @param header pointer to the header structure
451 @param b alignment to print
452 @return a pointer to the SAM string
454 char *bam_format1(const bam_header_t *header, const bam1_t *b);
456 char *bam_format1_core(const bam_header_t *header, const bam1_t *b, int of);
459 @abstract Check whether a BAM record is plausibly valid
460 @param header associated header structure, or NULL if unavailable
461 @param b alignment to validate
462 @return 0 if the alignment is invalid; non-zero otherwise
464 @discussion Simple consistency check of some of the fields of the
465 alignment record. If the header is provided, several additional checks
466 are made. Not all fields are checked, so a non-zero result is not a
467 guarantee that the record is valid. However it is usually good enough
468 to detect when bam_seek() has been called with a virtual file offset
469 that is not the offset of an alignment record.
471 int bam_validate1(const bam_header_t *header, const bam1_t *b);
473 const char *bam_get_library(bam_header_t *header, const bam1_t *b);
481 @abstract Structure for one alignment covering the pileup position.
482 @field b pointer to the alignment
483 @field qpos position of the read base at the pileup site, 0-based
484 @field indel indel length; 0 for no indel, positive for ins and negative for del
485 @field is_del 1 iff the base on the padded read is a deletion
486 @field level the level of the read in the "viewer" mode
488 @discussion See also bam_plbuf_push() and bam_lplbuf_push(). The
489 difference between the two functions is that the former does not
490 set bam_pileup1_t::level, while the later does. Level helps the
491 implementation of alignment viewers, but calculating this has some
498 uint32_t is_del:1, is_head:1, is_tail:1, is_refskip:1;
501 typedef int (*bam_plp_auto_f)(void *data, bam1_t *b);
504 typedef struct __bam_plp_t *bam_plp_t;
506 bam_plp_t bam_plp_init(bam_plp_auto_f func, void *data);
507 int bam_plp_push(bam_plp_t iter, const bam1_t *b);
508 const bam_pileup1_t *bam_plp_next(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp);
509 const bam_pileup1_t *bam_plp_auto(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp);
510 void bam_plp_set_mask(bam_plp_t iter, int mask);
511 void bam_plp_reset(bam_plp_t iter);
512 void bam_plp_destroy(bam_plp_t iter);
515 typedef struct __bam_mplp_t *bam_mplp_t;
517 bam_mplp_t bam_mplp_init(int n, bam_plp_auto_f func, void **data);
518 void bam_mplp_destroy(bam_mplp_t iter);
519 int bam_mplp_auto(bam_mplp_t iter, int *_tid, int *_pos, int *n_plp, const bam_pileup1_t **plp);
522 @abstract Type of function to be called by bam_plbuf_push().
523 @param tid chromosome ID as is defined in the header
524 @param pos start coordinate of the alignment, 0-based
525 @param n number of elements in pl array
526 @param pl array of alignments
527 @param data user provided data
528 @discussion See also bam_plbuf_push(), bam_plbuf_init() and bam_pileup1_t.
530 typedef int (*bam_pileup_f)(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pl, void *data);
538 void bam_plbuf_set_mask(bam_plbuf_t *buf, int mask);
539 void bam_plbuf_reset(bam_plbuf_t *buf);
540 bam_plbuf_t *bam_plbuf_init(bam_pileup_f func, void *data);
541 void bam_plbuf_destroy(bam_plbuf_t *buf);
542 int bam_plbuf_push(const bam1_t *b, bam_plbuf_t *buf);
544 int bam_pileup_file(bamFile fp, int mask, bam_pileup_f func, void *func_data);
546 struct __bam_lplbuf_t;
547 typedef struct __bam_lplbuf_t bam_lplbuf_t;
549 void bam_lplbuf_reset(bam_lplbuf_t *buf);
551 /*! @abstract bam_plbuf_init() equivalent with level calculated. */
552 bam_lplbuf_t *bam_lplbuf_init(bam_pileup_f func, void *data);
554 /*! @abstract bam_plbuf_destroy() equivalent with level calculated. */
555 void bam_lplbuf_destroy(bam_lplbuf_t *tv);
557 /*! @abstract bam_plbuf_push() equivalent with level calculated. */
558 int bam_lplbuf_push(const bam1_t *b, bam_lplbuf_t *buf);
561 /*********************
562 * BAM indexing APIs *
563 *********************/
565 struct __bam_index_t;
566 typedef struct __bam_index_t bam_index_t;
569 @abstract Build index for a BAM file.
570 @discussion Index file "fn.bai" will be created.
571 @param fn name of the BAM file
572 @return always 0 currently
574 int bam_index_build(const char *fn);
577 @abstract Load index from file "fn.bai".
578 @param fn name of the BAM file (NOT the index file)
579 @return pointer to the index structure
581 bam_index_t *bam_index_load(const char *fn);
584 @abstract Destroy an index structure.
585 @param idx pointer to the index structure
587 void bam_index_destroy(bam_index_t *idx);
590 @abstract Type of function to be called by bam_fetch().
591 @param b the alignment
592 @param data user provided data
594 typedef int (*bam_fetch_f)(const bam1_t *b, void *data);
597 @abstract Retrieve the alignments that are overlapped with the
600 @discussion A user defined function will be called for each
601 retrieved alignment ordered by its start position.
603 @param fp BAM file handler
604 @param idx pointer to the alignment index
605 @param tid chromosome ID as is defined in the header
606 @param beg start coordinate, 0-based
607 @param end end coordinate, 0-based
608 @param data user provided data (will be transferred to func)
609 @param func user defined function
611 int bam_fetch(bamFile fp, const bam_index_t *idx, int tid, int beg, int end, void *data, bam_fetch_f func);
613 bam_iter_t bam_iter_query(const bam_index_t *idx, int tid, int beg, int end);
614 int bam_iter_read(bamFile fp, bam_iter_t iter, bam1_t *b);
615 void bam_iter_destroy(bam_iter_t iter);
618 @abstract Parse a region in the format: "chr2:100,000-200,000".
619 @discussion bam_header_t::hash will be initialized if empty.
620 @param header pointer to the header structure
621 @param str string to be parsed
622 @param ref_id the returned chromosome ID
623 @param begin the returned start coordinate
624 @param end the returned end coordinate
625 @return 0 on success; -1 on failure
627 int bam_parse_region(bam_header_t *header, const char *str, int *ref_id, int *begin, int *end);
630 /**************************
631 * APIs for optional tags *
632 **************************/
635 @abstract Retrieve data of a tag
636 @param b pointer to an alignment struct
637 @param tag two-character tag to be retrieved
639 @return pointer to the type and data. The first character is the
640 type that can be 'iIsScCdfAZH'.
642 @discussion Use bam_aux2?() series to convert the returned data to
643 the corresponding type.
645 uint8_t *bam_aux_get(const bam1_t *b, const char tag[2]);
647 int32_t bam_aux2i(const uint8_t *s);
648 float bam_aux2f(const uint8_t *s);
649 double bam_aux2d(const uint8_t *s);
650 char bam_aux2A(const uint8_t *s);
651 char *bam_aux2Z(const uint8_t *s);
653 int bam_aux_del(bam1_t *b, uint8_t *s);
654 void bam_aux_append(bam1_t *b, const char tag[2], char type, int len, uint8_t *data);
655 uint8_t *bam_aux_get_core(bam1_t *b, const char tag[2]); // an alias of bam_aux_get()
663 @abstract Calculate the rightmost coordinate of an alignment on the
666 @param c pointer to the bam1_core_t structure
667 @param cigar the corresponding CIGAR array (from bam1_t::cigar)
668 @return the rightmost coordinate, 0-based
670 uint32_t bam_calend(const bam1_core_t *c, const uint32_t *cigar);
673 @abstract Calculate the length of the query sequence from CIGAR.
674 @param c pointer to the bam1_core_t structure
675 @param cigar the corresponding CIGAR array (from bam1_t::cigar)
676 @return length of the query sequence
678 int32_t bam_cigar2qlen(const bam1_core_t *c, const uint32_t *cigar);
685 @abstract Calculate the minimum bin that contains a region [beg,end).
686 @param beg start of the region, 0-based
687 @param end end of the region, 0-based
690 static inline int bam_reg2bin(uint32_t beg, uint32_t end)
693 if (beg>>14 == end>>14) return 4681 + (beg>>14);
694 if (beg>>17 == end>>17) return 585 + (beg>>17);
695 if (beg>>20 == end>>20) return 73 + (beg>>20);
696 if (beg>>23 == end>>23) return 9 + (beg>>23);
697 if (beg>>26 == end>>26) return 1 + (beg>>26);
702 @abstract Copy an alignment
703 @param bdst destination alignment struct
704 @param bsrc source alignment struct
705 @return pointer to the destination alignment struct
707 static inline bam1_t *bam_copy1(bam1_t *bdst, const bam1_t *bsrc)
709 uint8_t *data = bdst->data;
710 int m_data = bdst->m_data; // backup data and m_data
711 if (m_data < bsrc->m_data) { // double the capacity
712 m_data = bsrc->m_data; kroundup32(m_data);
713 data = (uint8_t*)realloc(data, m_data);
715 memcpy(data, bsrc->data, bsrc->data_len); // copy var-len data
716 *bdst = *bsrc; // copy the rest
717 // restore the backup
718 bdst->m_data = m_data;
724 @abstract Duplicate an alignment
725 @param src source alignment struct
726 @return pointer to the destination alignment struct
728 static inline bam1_t *bam_dup1(const bam1_t *src)
733 b->m_data = b->data_len;
734 b->data = (uint8_t*)calloc(b->data_len, 1);
735 memcpy(b->data, src->data, b->data_len);