bam_md.c

   1 #include <unistd.h>
   2 #include <assert.h>
   3 #include <string.h>
   4 #include <ctype.h>
   5 #include "faidx.h"
   6 #include "sam.h"
   7 #include "kstring.h"
   8
   9 void bam_fillmd1(bam1_t *b, char *ref, int is_equal)
  10 {
  11         uint8_t *seq = bam1_seq(b);
  12         uint32_t *cigar = bam1_cigar(b);
  13         bam1_core_t *c = &b->core;
  14         int i, x, y, u = 0;
  15         kstring_t *str;
  16         uint8_t *old_md, *old_nm;
  17         int32_t old_nm_i = -1, nm = 0;
  18
  19         old_md = bam_aux_get(b, "MD");
  20         old_nm = bam_aux_get(b, "NM");
  21         if (c->flag & BAM_FUNMAP) return;
  22         if (old_nm) old_nm_i = bam_aux2i(old_nm);
  23         str = (kstring_t*)calloc(1, sizeof(kstring_t));
  24         for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
  25                 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
  26                 if (op == BAM_CMATCH) {
  27                         for (j = 0; j < l; ++j) {
  28                                 int z = y + j;
  29                                 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
  30                                 if (ref[x+j] == 0) break; // out of boundary
  31                                 if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) { // a match
  32                                         if (is_equal) seq[z/2] &= (z&1)? 0xf0 : 0x0f;
  33                                         ++u;
  34                                 } else {
  35                                         ksprintf(str, "%d", u);
  36                                         kputc(ref[x+j], str);
  37                                         u = 0; ++nm;
  38                                 }
  39                         }
  40                         if (j < l) break;
  41                         x += l; y += l;
  42                 } else if (op == BAM_CDEL) {
  43                         ksprintf(str, "%d", u);
  44                         kputc('^', str);
  45                         for (j = 0; j < l; ++j) {
  46                                 if (ref[x+j] == 0) break;
  47                                 kputc(ref[x+j], str);
  48                         }
  49                         u = 0;
  50                         if (j < l) break;
  51                         x += l; nm += l;
  52                 } else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) {
  53                         y += l;
  54                         if (op == BAM_CINS) nm += l;
  55                 } else if (op == BAM_CREF_SKIP) {
  56                         x += l;
  57                 }
  58         }
  59         ksprintf(str, "%d", u);
  60         if (!old_nm) bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
  61         else if (nm != old_nm_i) {
  62                 fprintf(stderr, "[bam_fillmd1] different NM for read '%s': %d -> %d\n", bam1_qname(b), old_nm_i, nm);
  63                 bam_aux_del(b, old_nm);
  64                 bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
  65         }
  66         if (!old_md) bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
  67         else {
  68                 int is_diff = 0;
  69                 if (strlen((char*)old_md+1) == str->l) {
  70                         for (i = 0; i < str->l; ++i)
  71                                 if (toupper(old_md[i+1]) != toupper(str->s[i]))
  72                                         break;
  73                         if (i < str->l) is_diff = 1;
  74                 } else is_diff = 1;
  75                 if (is_diff) {
  76                         fprintf(stderr, "[bam_fillmd1] different MD for read '%s': '%s' -> '%s'\n", bam1_qname(b), old_md+1, str->s);
  77                         bam_aux_del(b, old_md);
  78                         bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
  79                 }
  80         }
  81         free(str->s); free(str);
  82 }
  83
  84 int bam_fillmd(int argc, char *argv[])
  85 {
  86         int c, is_equal = 0, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed;
  87         samfile_t *fp, *fpout = 0;
  88         faidx_t *fai;
  89         char *ref = 0, mode_w[8], mode_r[8];
  90         bam1_t *b;
  91
  92         is_bam_out = is_sam_in = is_uncompressed = 0;
  93         mode_w[0] = mode_r[0] = 0;
  94         strcpy(mode_r, "r"); strcpy(mode_w, "w");
  95         while ((c = getopt(argc, argv, "eubS")) >= 0) {
  96                 switch (c) {
  97                 case 'e': is_equal = 1; break;
  98                 case 'b': is_bam_out = 1; break;
  99                 case 'u': is_uncompressed = is_bam_out = 1; break;
 100                 case 'S': is_sam_in = 1; break;
 101                 default: fprintf(stderr, "[bam_fillmd] unrecognized option '-%c'\n", c); return 1;
 102                 }
 103         }
 104         if (!is_sam_in) strcat(mode_r, "b");
 105         if (is_bam_out) strcat(mode_w, "b");
 106         else strcat(mode_w, "h");
 107         if (is_uncompressed) strcat(mode_w, "u");
 108         if (optind + 1 >= argc) {
 109                 fprintf(stderr, "\n");
 110                 fprintf(stderr, "Usage:   samtools fillmd [-eubS] <aln.bam> <ref.fasta>\n\n");
 111                 fprintf(stderr, "Options: -e       change identical bases to '='\n");
 112                 fprintf(stderr, "         -u       uncompressed BAM output (for piping)\n");
 113                 fprintf(stderr, "         -b       compressed BAM output\n");
 114                 fprintf(stderr, "         -S       the input is SAM with header\n\n");
 115                 return 1;
 116         }
 117         fp = samopen(argv[optind], mode_r, 0);
 118         if (fp == 0) return 1;
 119         if (is_sam_in && (fp->header == 0 || fp->header->n_targets == 0)) {
 120                 fprintf(stderr, "[bam_fillmd] input SAM does not have header. Abort!\n");
 121                 return 1;
 122         }
 123         fpout = samopen("-", mode_w, fp->header);
 124         fai = fai_load(argv[optind+1]);
 125
 126         b = bam_init1();
 127         while ((ret = samread(fp, b)) >= 0) {
 128                 if (b->core.tid >= 0) {
 129                         if (tid != b->core.tid) {
 130                                 free(ref);
 131                                 ref = fai_fetch(fai, fp->header->target_name[b->core.tid], &len);
 132                                 tid = b->core.tid;
 133                         }
 134                         bam_fillmd1(b, ref, is_equal);
 135                 }
 136                 samwrite(fpout, b);
 137         }
 138         bam_destroy1(b);
 139
 140         free(ref);
 141         fai_destroy(fai);
 142         samclose(fp); samclose(fpout);
 143         return 0;
 144 }