bam_md.c

   1 #include <unistd.h>
   2 #include <assert.h>
   3 #include <string.h>
   4 #include <ctype.h>
   5 #include <math.h>
   6 #include "faidx.h"
   7 #include "sam.h"
   8 #include "kstring.h"
   9 #include "kaln.h"
  10
  11 void bam_fillmd1_core(bam1_t *b, char *ref, int is_equal, int max_nm)
  12 {
  13         uint8_t *seq = bam1_seq(b);
  14         uint32_t *cigar = bam1_cigar(b);
  15         bam1_core_t *c = &b->core;
  16         int i, x, y, u = 0;
  17         kstring_t *str;
  18         uint8_t *old_md, *old_nm;
  19         int32_t old_nm_i = -1, nm = 0;
  20
  21         str = (kstring_t*)calloc(1, sizeof(kstring_t));
  22         for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
  23                 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
  24                 if (op == BAM_CMATCH) {
  25                         for (j = 0; j < l; ++j) {
  26                                 int z = y + j;
  27                                 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
  28                                 if (ref[x+j] == 0) break; // out of boundary
  29                                 if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) { // a match
  30                                         if (is_equal) seq[z/2] &= (z&1)? 0xf0 : 0x0f;
  31                                         ++u;
  32                                 } else {
  33                                         ksprintf(str, "%d", u);
  34                                         kputc(ref[x+j], str);
  35                                         u = 0; ++nm;
  36                                 }
  37                         }
  38                         if (j < l) break;
  39                         x += l; y += l;
  40                 } else if (op == BAM_CDEL) {
  41                         ksprintf(str, "%d", u);
  42                         kputc('^', str);
  43                         for (j = 0; j < l; ++j) {
  44                                 if (ref[x+j] == 0) break;
  45                                 kputc(ref[x+j], str);
  46                         }
  47                         u = 0;
  48                         if (j < l) break;
  49                         x += l; nm += l;
  50                 } else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) {
  51                         y += l;
  52                         if (op == BAM_CINS) nm += l;
  53                 } else if (op == BAM_CREF_SKIP) {
  54                         x += l;
  55                 }
  56         }
  57         ksprintf(str, "%d", u);
  58         // apply max_nm
  59         if (max_nm > 0 && nm >= max_nm) {
  60                 for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
  61                         int j, l = cigar[i]>>4, op = cigar[i]&0xf;
  62                         if (op == BAM_CMATCH) {
  63                                 for (j = 0; j < l; ++j) {
  64                                         int z = y + j;
  65                                         int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
  66                                         if (ref[x+j] == 0) break; // out of boundary
  67                                         if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) { // a match
  68                                                 seq[z/2] |= (z&1)? 0x0f : 0xf0;
  69                                                 bam1_qual(b)[z] = 0;
  70                                         }
  71                                 }
  72                                 if (j < l) break;
  73                                 x += l; y += l;
  74                         } else if (op == BAM_CDEL || op == BAM_CREF_SKIP) x += l;
  75                         else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) y += l;
  76                 }
  77         }
  78         // update NM
  79         old_nm = bam_aux_get(b, "NM");
  80         if (c->flag & BAM_FUNMAP) return;
  81         if (old_nm) old_nm_i = bam_aux2i(old_nm);
  82         if (!old_nm) bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
  83         else if (nm != old_nm_i) {
  84                 fprintf(stderr, "[bam_fillmd1] different NM for read '%s': %d -> %d\n", bam1_qname(b), old_nm_i, nm);
  85                 bam_aux_del(b, old_nm);
  86                 bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
  87         }
  88         // update MD
  89         old_md = bam_aux_get(b, "MD");
  90         if (!old_md) bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
  91         else {
  92                 int is_diff = 0;
  93                 if (strlen((char*)old_md+1) == str->l) {
  94                         for (i = 0; i < str->l; ++i)
  95                                 if (toupper(old_md[i+1]) != toupper(str->s[i]))
  96                                         break;
  97                         if (i < str->l) is_diff = 1;
  98                 } else is_diff = 1;
  99                 if (is_diff) {
 100                         fprintf(stderr, "[bam_fillmd1] different MD for read '%s': '%s' -> '%s'\n", bam1_qname(b), old_md+1, str->s);
 101                         bam_aux_del(b, old_md);
 102                         bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
 103                 }
 104         }
 105         free(str->s); free(str);
 106 }
 107
 108 void bam_fillmd1(bam1_t *b, char *ref, int is_equal)
 109 {
 110         bam_fillmd1_core(b, ref, is_equal, 0);
 111 }
 112
 113 int bam_cap_mapQ(bam1_t *b, char *ref, int thres)
 114 {
 115         uint8_t *seq = bam1_seq(b), *qual = bam1_qual(b);
 116         uint32_t *cigar = bam1_cigar(b);
 117         bam1_core_t *c = &b->core;
 118         int i, x, y, mm, q, len, clip_l, clip_q;
 119         double t;
 120         if (thres < 0) thres = 40; // set the default
 121         mm = q = len = clip_l = clip_q = 0;
 122         for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
 123                 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
 124                 if (op == BAM_CMATCH) {
 125                         for (j = 0; j < l; ++j) {
 126                                 int z = y + j;
 127                                 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
 128                                 if (ref[x+j] == 0) break; // out of boundary
 129                                 if (c2 != 15 && c1 != 15 && qual[z] >= 13) { // not ambiguous
 130                                         ++len;
 131                                         if (c1 && c1 != c2 && qual[z] >= 13) { // mismatch
 132                                                 ++mm;
 133                                                 q += qual[z] > 33? 33 : qual[z];
 134                                         }
 135                                 }
 136                         }
 137                         if (j < l) break;
 138                         x += l; y += l; len += l;
 139                 } else if (op == BAM_CDEL) {
 140                         for (j = 0; j < l; ++j)
 141                                 if (ref[x+j] == 0) break;
 142                         if (j < l) break;
 143                         x += l;
 144                 } else if (op == BAM_CSOFT_CLIP) {
 145                         for (j = 0; j < l; ++j) clip_q += qual[y+j];
 146                         clip_l += l;
 147                         y += l;
 148                 } else if (op == BAM_CHARD_CLIP) {
 149                         clip_q += 13 * l;
 150                         clip_l += l;
 151                 } else if (op == BAM_CINS) y += l;
 152                 else if (op == BAM_CREF_SKIP) x += l;
 153         }
 154         for (i = 0, t = 1; i < mm; ++i)
 155                 t *= (double)len / (i+1);
 156         t = q - 4.343 * log(t) + clip_q / 5.;
 157         if (t > thres) return -1;
 158         if (t < 0) t = 0;
 159         t = sqrt((thres - t) / thres) * thres;
 160 //      fprintf(stderr, "%s %lf %d\n", bam1_qname(b), t, q);
 161         return (int)(t + .499);
 162 }
 163
 164 int bam_prob_realn(bam1_t *b, const char *ref)
 165 {
 166         int k, i, bw, x, y, yb, ye, xb, xe;
 167         uint32_t *cigar = bam1_cigar(b);
 168         bam1_core_t *c = &b->core;
 169         ka_probpar_t conf = ka_probpar_def;
 170         // find the start and end of the alignment
 171         if (c->flag & BAM_FUNMAP) return -1;
 172         x = c->pos, y = 0, yb = ye = xb = xe = -1;
 173         for (k = 0; k < c->n_cigar; ++k) {
 174                 int op, l;
 175                 op = cigar[k]&0xf; l = cigar[k]>>4;
 176                 if (op == BAM_CMATCH) {
 177                         if (yb < 0) yb = y;
 178                         if (xb < 0) xb = x;
 179                         ye = y + l; xe = x + l;
 180                         x += l; y += l;
 181                 } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
 182                 else if (op == BAM_CDEL) x += l;
 183                 else if (op == BAM_CREF_SKIP) return -1;
 184         }
 185         // set bandwidth and the start and the end
 186         bw = 7;
 187         if (abs((xe - xb) - (ye - yb)) > bw)
 188                 bw = abs((xe - xb) - (ye - yb)) + 3;
 189         conf.bw = bw;
 190         xb -= yb + bw/2; if (xb < 0) xb = 0;
 191         xe += c->l_qseq - ye + bw/2;
 192         if (xe - xb - c->l_qseq > bw)
 193                 xb += (xe - xb - c->l_qseq - bw) / 2, xe -= (xe - xb - c->l_qseq - bw) / 2;
 194         { // glocal
 195                 uint8_t *s, *r, *q, *seq = bam1_seq(b), *qual = bam1_qual(b);
 196                 int *state;
 197                 s = calloc(c->l_qseq, 1);
 198                 for (i = 0; i < c->l_qseq; ++i) s[i] = bam_nt16_nt4_table[bam1_seqi(seq, i)];
 199                 r = calloc(xe - xb, 1);
 200                 for (i = xb; i < xe; ++i)
 201                         r[i-xb] = bam_nt16_nt4_table[bam_nt16_table[(int)ref[i]]];
 202                 state = calloc(c->l_qseq, sizeof(int));
 203                 q = calloc(c->l_qseq, 1);
 204                 ka_prob_glocal(r, xe-xb, s, c->l_qseq, qual, &conf, state, q);
 205                 for (k = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
 206                         int op = cigar[k]&0xf, l = cigar[k]>>4;
 207                         if (op == BAM_CMATCH) {
 208                                 for (i = y; i < y + l; ++i) {
 209                                         if ((state[i]&3) != 0 || state[i]>>2 != x - xb + (i - y)) qual[i] = 0;
 210                                         else qual[i] = qual[i] < q[i]? qual[i] : q[i];
 211                                 }
 212                                 x += l; y += l;
 213                         } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
 214                         else if (op == BAM_CDEL) x += l;
 215                 }
 216                 free(s); free(r); free(q); free(state);
 217         }
 218         return 0;
 219 }
 220
 221 int bam_fillmd(int argc, char *argv[])
 222 {
 223         int c, is_equal = 0, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed, max_nm = 0, is_realn, capQ = 0;
 224         samfile_t *fp, *fpout = 0;
 225         faidx_t *fai;
 226         char *ref = 0, mode_w[8], mode_r[8];
 227         bam1_t *b;
 228
 229         is_bam_out = is_sam_in = is_uncompressed = is_realn = 0;
 230         mode_w[0] = mode_r[0] = 0;
 231         strcpy(mode_r, "r"); strcpy(mode_w, "w");
 232         while ((c = getopt(argc, argv, "reubSC:n:")) >= 0) {
 233                 switch (c) {
 234                 case 'r': is_realn = 1; break;
 235                 case 'e': is_equal = 1; break;
 236                 case 'b': is_bam_out = 1; break;
 237                 case 'u': is_uncompressed = is_bam_out = 1; break;
 238                 case 'S': is_sam_in = 1; break;
 239                 case 'n': max_nm = atoi(optarg); break;
 240                 case 'C': capQ = atoi(optarg); break;
 241                 default: fprintf(stderr, "[bam_fillmd] unrecognized option '-%c'\n", c); return 1;
 242                 }
 243         }
 244         if (!is_sam_in) strcat(mode_r, "b");
 245         if (is_bam_out) strcat(mode_w, "b");
 246         else strcat(mode_w, "h");
 247         if (is_uncompressed) strcat(mode_w, "u");
 248         if (optind + 1 >= argc) {
 249                 fprintf(stderr, "\n");
 250                 fprintf(stderr, "Usage:   samtools fillmd [-eubrS] <aln.bam> <ref.fasta>\n\n");
 251                 fprintf(stderr, "Options: -e       change identical bases to '='\n");
 252                 fprintf(stderr, "         -u       uncompressed BAM output (for piping)\n");
 253                 fprintf(stderr, "         -b       compressed BAM output\n");
 254                 fprintf(stderr, "         -S       the input is SAM with header\n");
 255                 fprintf(stderr, "         -r       read-independent local realignment\n\n");
 256                 return 1;
 257         }
 258         fp = samopen(argv[optind], mode_r, 0);
 259         if (fp == 0) return 1;
 260         if (is_sam_in && (fp->header == 0 || fp->header->n_targets == 0)) {
 261                 fprintf(stderr, "[bam_fillmd] input SAM does not have header. Abort!\n");
 262                 return 1;
 263         }
 264         fpout = samopen("-", mode_w, fp->header);
 265         fai = fai_load(argv[optind+1]);
 266
 267         b = bam_init1();
 268         while ((ret = samread(fp, b)) >= 0) {
 269                 if (b->core.tid >= 0) {
 270                         if (tid != b->core.tid) {
 271                                 free(ref);
 272                                 ref = fai_fetch(fai, fp->header->target_name[b->core.tid], &len);
 273                                 tid = b->core.tid;
 274                                 if (ref == 0)
 275                                         fprintf(stderr, "[bam_fillmd] fail to find sequence '%s' in the reference.\n",
 276                                                         fp->header->target_name[tid]);
 277                         }
 278 //                      if (is_realn) bam_realn(b, ref);
 279                         if (is_realn) bam_prob_realn(b, ref);
 280                         if (capQ > 10) {
 281                                 int q = bam_cap_mapQ(b, ref, capQ);
 282                                 if (b->core.qual > q) b->core.qual = q;
 283                         }
 284                         if (ref) bam_fillmd1_core(b, ref, is_equal, max_nm);
 285                 }
 286                 samwrite(fpout, b);
 287         }
 288         bam_destroy1(b);
 289
 290         free(ref);
 291         fai_destroy(fai);
 292         samclose(fp); samclose(fpout);
 293         return 0;
 294 }