bam_md.c

   1 #include <unistd.h>
   2 #include <string.h>
   3 #include <ctype.h>
   4 #include "faidx.h"
   5 #include "bam.h"
   6 #include "kstring.h"
   7
   8 void bam_fillmd1(bam1_t *b, char *ref, int is_equal)
   9 {
  10         uint8_t *seq = bam1_seq(b);
  11         uint32_t *cigar = bam1_cigar(b);
  12         bam1_core_t *c = &b->core;
  13         int i, x, y, u = 0;
  14         kstring_t *str;
  15         uint8_t *old_md;
  16
  17         old_md = bam_aux_get(b, "MD");
  18         if (c->flag & BAM_FUNMAP) return;
  19         if (old_md && !is_equal) return; // no need to add MD
  20         str = (kstring_t*)calloc(1, sizeof(kstring_t));
  21         for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
  22                 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
  23                 if (op == BAM_CMATCH) {
  24                         for (j = 0; j < l; ++j) {
  25                                 int z = y + j;
  26                                 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
  27                                 if (ref[x+j] == 0) break; // out of boundary
  28                                 if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) {
  29                                         if (is_equal) seq[z/2] &= (z&1)? 0xf0 : 0x0f;
  30                                         ++u;
  31                                 } else {
  32                                         ksprintf(str, "%d", u);
  33                                         kputc(ref[x+j], str);
  34                                         u = 0;
  35                                 }
  36                         }
  37                         if (j < l) break;
  38                         x += l; y += l;
  39                 } else if (op == BAM_CDEL) {
  40                         ksprintf(str, "%d", u);
  41                         kputc('^', str);
  42                         for (j = 0; j < l; ++j) {
  43                                 if (ref[x+j] == 0) break;
  44                                 kputc(ref[x+j], str);
  45                         }
  46                         u = 0;
  47                         if (j < l) break;
  48                         x += l;
  49                 } else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) {
  50                         y += l;
  51                 } else if (op == BAM_CREF_SKIP) {
  52                         x += l;
  53                 }
  54         }
  55         ksprintf(str, "%d", u);
  56         if (!old_md) bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
  57         else {
  58                 int is_diff = 0;
  59                 if (strlen((char*)old_md+1) == str->l) {
  60                         for (i = 0; i < str->l; ++i)
  61                                 if (toupper(old_md[i+1]) != toupper(str->s[i]))
  62                                         break;
  63                         if (i < str->l) is_diff = 1;
  64                 } else is_diff = 1;
  65                 if (is_diff)
  66                         fprintf(stderr, "[bam_fillmd1] different MD for read '%s': '%s' != '%s'\n", bam1_qname(b), old_md+1, str->s);
  67         }
  68         free(str->s); free(str);
  69 }
  70
  71 int bam_fillmd(int argc, char *argv[])
  72 {
  73         int c, is_equal = 0, tid = -2, ret, len;
  74         bamFile fp, fpout = 0;
  75         bam_header_t *header;
  76         faidx_t *fai;
  77         char *ref = 0;
  78         bam1_t *b;
  79
  80         while ((c = getopt(argc, argv, "e")) >= 0) {
  81                 switch (c) {
  82                 case 'e': is_equal = 1; break;
  83                 default: fprintf(stderr, "[bam_fillmd] unrecognized option '-%c'\n", c); return 1;
  84                 }
  85         }
  86         if (optind + 1 >= argc) {
  87                 fprintf(stderr, "Usage: bam fillmd [-e] <aln.bam> <ref.fasta>\n");
  88                 return 1;
  89         }
  90         fp = strcmp(argv[optind], "-")? bam_open(argv[optind], "r") : bam_dopen(fileno(stdin), "r");
  91         assert(fp);
  92         header = bam_header_read(fp);
  93         fpout = bam_dopen(fileno(stdout), "w");
  94         bam_header_write(fpout, header);
  95         fai = fai_load(argv[optind+1]);
  96
  97         b = bam_init1();
  98         while ((ret = bam_read1(fp, b)) >= 0) {
  99                 if (b->core.tid >= 0) {
 100                         if (tid != b->core.tid) {
 101                                 free(ref);
 102                                 ref = fai_fetch(fai, header->target_name[b->core.tid], &len);
 103                                 tid = b->core.tid;
 104                         }
 105                         bam_fillmd1(b, ref, is_equal);
 106                 }
 107                 bam_write1(fpout, b);
 108         }
 109         bam_destroy1(b);
 110
 111         free(ref);
 112         fai_destroy(fai);
 113         bam_header_destroy(header);
 114         bam_close(fp); bam_close(fpout);
 115         return 0;
 116 }