bam2bcf.c

   1 #include <math.h>
   2 #include <stdint.h>
   3 #include "bam.h"
   4 #include "kstring.h"
   5 #include "bam2bcf.h"
   6 #include "errmod.h"
   7 #include "bcftools/bcf.h"
   8
   9 extern  void ks_introsort_uint32_t(size_t n, uint32_t a[]);
  10
  11 #define CALL_ETA 0.03f
  12 #define CALL_MAX 256
  13 #define CALL_DEFTHETA 0.83f
  14
  15 #define CAP_DIST 25
  16
  17 bcf_callaux_t *bcf_call_init(double theta, int min_baseQ)
  18 {
  19         bcf_callaux_t *bca;
  20         if (theta <= 0.) theta = CALL_DEFTHETA;
  21         bca = calloc(1, sizeof(bcf_callaux_t));
  22         bca->capQ = 60;
  23         bca->openQ = 40; bca->extQ = 20; bca->tandemQ = 100;
  24         bca->min_baseQ = min_baseQ;
  25         bca->e = errmod_init(1. - theta);
  26         return bca;
  27 }
  28
  29 void bcf_call_destroy(bcf_callaux_t *bca)
  30 {
  31         if (bca == 0) return;
  32         errmod_destroy(bca->e);
  33         free(bca->bases); free(bca->inscns); free(bca);
  34 }
  35 /* ref_base is the 4-bit representation of the reference base. It is
  36  * negative if we are looking at an indel. */
  37 int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r)
  38 {
  39         int i, n, ref4, is_indel, ori_depth = 0;
  40         memset(r, 0, sizeof(bcf_callret1_t));
  41         if (ref_base >= 0) {
  42                 ref4 = bam_nt16_nt4_table[ref_base];
  43                 is_indel = 0;
  44         } else ref4 = 4, is_indel = 1;
  45         if (_n == 0) return -1;
  46         // enlarge the bases array if necessary
  47         if (bca->max_bases < _n) {
  48                 bca->max_bases = _n;
  49                 kroundup32(bca->max_bases);
  50                 bca->bases = (uint16_t*)realloc(bca->bases, 2 * bca->max_bases);
  51         }
  52         // fill the bases array
  53         memset(r, 0, sizeof(bcf_callret1_t));
  54         for (i = n = 0; i < _n; ++i) {
  55                 const bam_pileup1_t *p = pl + i;
  56                 int q, b, mapQ, baseQ, is_diff, min_dist, seqQ;
  57                 // set base
  58                 if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
  59                 ++ori_depth;
  60                 baseQ = q = is_indel? p->aux&0xff : (int)bam1_qual(p->b)[p->qpos]; // base/indel quality
  61                 seqQ = is_indel? (p->aux>>8&0xff) : 99;
  62                 if (q < bca->min_baseQ) continue;
  63                 if (q > seqQ) q = seqQ;
  64                 mapQ = p->b->core.qual < bca->capQ? p->b->core.qual : bca->capQ;
  65                 if (q > mapQ) q = mapQ;
  66                 if (q > 63) q = 63;
  67                 if (q < 4) q = 4;
  68                 if (!is_indel) {
  69                         b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
  70                         b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
  71                         is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
  72                 } else {
  73                         b = p->aux>>16&0x3f;
  74                         is_diff = (b != 0);
  75                 }
  76                 bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
  77                 // collect annotations
  78                 r->qsum[b] += q;
  79                 ++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
  80                 min_dist = p->b->core.l_qseq - 1 - p->qpos;
  81                 if (min_dist > p->qpos) min_dist = p->qpos;
  82                 if (min_dist > CAP_DIST) min_dist = CAP_DIST;
  83                 r->anno[1<<2|is_diff<<1|0] += baseQ;
  84                 r->anno[1<<2|is_diff<<1|1] += baseQ * baseQ;
  85                 r->anno[2<<2|is_diff<<1|0] += mapQ;
  86                 r->anno[2<<2|is_diff<<1|1] += mapQ * mapQ;
  87                 r->anno[3<<2|is_diff<<1|0] += min_dist;
  88                 r->anno[3<<2|is_diff<<1|1] += min_dist * min_dist;
  89         }
  90         r->depth = n; r->ori_depth = ori_depth;
  91         // glfgen
  92         errmod_cal(bca->e, n, 5, bca->bases, r->p);
  93         return r->depth;
  94 }
  95
  96 int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/, bcf_call_t *call)
  97 {
  98         int ref4, i, j, qsum[4];
  99         int64_t tmp;
 100         if (ref_base >= 0) {
 101                 call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
 102                 if (ref4 > 4) ref4 = 4;
 103         } else call->ori_ref = -1, ref4 = 0;
 104         // calculate qsum
 105         memset(qsum, 0, 4 * sizeof(int));
 106         for (i = 0; i < n; ++i)
 107                 for (j = 0; j < 4; ++j)
 108                         qsum[j] += calls[i].qsum[j];
 109         for (j = 0; j < 4; ++j) qsum[j] = qsum[j] << 2 | j;
 110         // find the top 2 alleles
 111         for (i = 1; i < 4; ++i) // insertion sort
 112                 for (j = i; j > 0 && qsum[j] < qsum[j-1]; --j)
 113                         tmp = qsum[j], qsum[j] = qsum[j-1], qsum[j-1] = tmp;
 114         // set the reference allele and alternative allele(s)
 115         for (i = 0; i < 5; ++i) call->a[i] = -1;
 116         call->unseen = -1;
 117         call->a[0] = ref4;
 118         for (i = 3, j = 1; i >= 0; --i) {
 119                 if ((qsum[i]&3) != ref4) {
 120                         if (qsum[i]>>2 != 0) call->a[j++] = qsum[i]&3;
 121                         else break;
 122                 }
 123         }
 124         if (ref_base >= 0) { // for SNPs, find the "unseen" base
 125                 if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
 126                         call->unseen = j, call->a[j++] = qsum[i]&3;
 127                 call->n_alleles = j;
 128         } else {
 129                 call->n_alleles = j;
 130                 if (call->n_alleles == 1) return -1; // no reliable supporting read. stop doing anything
 131         }
 132         // set the PL array
 133         if (call->n < n) {
 134                 call->n = n;
 135                 call->PL = realloc(call->PL, 15 * n);
 136         }
 137         {
 138                 int x, g[15], z;
 139                 double sum_min = 0.;
 140                 x = call->n_alleles * (call->n_alleles + 1) / 2;
 141                 // get the possible genotypes
 142                 for (i = z = 0; i < call->n_alleles; ++i)
 143                         for (j = i; j < call->n_alleles; ++j)
 144                                 g[z++] = call->a[i] * 5 + call->a[j];
 145                 for (i = 0; i < n; ++i) {
 146                         uint8_t *PL = call->PL + x * i;
 147                         const bcf_callret1_t *r = calls + i;
 148                         float min = 1e37;
 149                         for (j = 0; j < x; ++j)
 150                                 if (min > r->p[g[j]]) min = r->p[g[j]];
 151                         sum_min += min;
 152                         for (j = 0; j < x; ++j) {
 153                                 int y;
 154                                 y = (int)(r->p[g[j]] - min + .499);
 155                                 if (y > 255) y = 255;
 156                                 PL[j] = y;
 157                         }
 158                 }
 159 //              if (ref_base < 0) fprintf(stderr, "%d,%d,%f,%d\n", call->n_alleles, x, sum_min, call->unseen);
 160                 call->shift = (int)(sum_min + .499);
 161         }
 162         // combine annotations
 163         memset(call->anno, 0, 16 * sizeof(int));
 164         for (i = call->depth = call->ori_depth = 0, tmp = 0; i < n; ++i) {
 165                 call->depth += calls[i].depth;
 166                 call->ori_depth += calls[i].ori_depth;
 167                 for (j = 0; j < 16; ++j) call->anno[j] += calls[i].anno[j];
 168         }
 169         return 0;
 170 }
 171
 172 int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP,
 173                                  const bcf_callaux_t *bca, const char *ref)
 174 {
 175         extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
 176         kstring_t s;
 177         int i, j;
 178         b->n_smpl = bc->n;
 179         b->tid = tid; b->pos = pos; b->qual = 0;
 180         s.s = b->str; s.m = b->m_str; s.l = 0;
 181         kputc('\0', &s);
 182         if (bc->ori_ref < 0) { // an indel
 183                 // write REF
 184                 kputc(ref[pos], &s);
 185                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 186                 kputc('\0', &s);
 187                 // write ALT
 188                 kputc(ref[pos], &s);
 189                 for (i = 1; i < 4; ++i) {
 190                         if (bc->a[i] < 0) break;
 191                         if (i > 1) {
 192                                 kputc(',', &s); kputc(ref[pos], &s);
 193                         }
 194                         if (bca->indel_types[bc->a[i]] < 0) { // deletion
 195                                 for (j = -bca->indel_types[bc->a[i]]; j < bca->indelreg; ++j)
 196                                         kputc(ref[pos+1+j], &s);
 197                         } else { // insertion; cannot be a reference unless a bug
 198                                 char *inscns = &bca->inscns[bc->a[i] * bca->maxins];
 199                                 for (j = 0; j < bca->indel_types[bc->a[i]]; ++j)
 200                                         kputc("ACGTN"[(int)inscns[j]], &s);
 201                                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 202                         }
 203                 }
 204                 kputc('\0', &s);
 205         } else { // a SNP
 206                 kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
 207                 for (i = 1; i < 5; ++i) {
 208                         if (bc->a[i] < 0) break;
 209                         if (i > 1) kputc(',', &s);
 210                         kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
 211                 }
 212                 kputc('\0', &s);
 213         }
 214         kputc('\0', &s);
 215         // INFO
 216         if (bc->ori_ref < 0) kputs("INDEL;", &s);
 217         kputs("DP=", &s); kputw(bc->ori_depth, &s); kputs(";I16=", &s);
 218         for (i = 0; i < 16; ++i) {
 219                 if (i) kputc(',', &s);
 220                 kputw(bc->anno[i], &s);
 221         }
 222         kputc('\0', &s);
 223         // FMT
 224         kputs("PL", &s);
 225         if (bcr) {
 226                 kputs(":DP", &s);
 227                 if (is_SP) kputs(":SP", &s);
 228         }
 229         kputc('\0', &s);
 230         b->m_str = s.m; b->str = s.s; b->l_str = s.l;
 231         bcf_sync(b);
 232         memcpy(b->gi[0].data, bc->PL, b->gi[0].len * bc->n);
 233         if (bcr) {
 234                 uint16_t *dp = (uint16_t*)b->gi[1].data;
 235                 uint8_t *sp = is_SP? b->gi[2].data : 0;
 236                 for (i = 0; i < bc->n; ++i) {
 237                         bcf_callret1_t *p = bcr + i;
 238                         dp[i] = p->depth < 0xffff? p->depth : 0xffff;
 239                         if (is_SP) {
 240                                 if (p->anno[0] + p->anno[1] < 2 || p->anno[2] + p->anno[3] < 2
 241                                         || p->anno[0] + p->anno[2] < 2 || p->anno[1] + p->anno[3] < 2)
 242                                 {
 243                                         sp[i] = 0;
 244                                 } else {
 245                                         double left, right, two;
 246                                         int x;
 247                                         kt_fisher_exact(p->anno[0], p->anno[1], p->anno[2], p->anno[3], &left, &right, &two);
 248                                         x = (int)(-4.343 * log(two) + .499);
 249                                         if (x > 255) x = 255;
 250                                         sp[i] = x;
 251                                 }
 252                         }
 253                 }
 254         }
 255         return 0;
 256 }