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->last_mnp_pos = -100;
  26         bca->e = errmod_init(1. - theta);
  27         return bca;
  28 }
  29
  30 void bcf_call_destroy(bcf_callaux_t *bca)
  31 {
  32         if (bca == 0) return;
  33         errmod_destroy(bca->e);
  34         free(bca->bases); free(bca->inscns); free(bca);
  35 }
  36 /* Compute genotype likelihood by combining likelihood of each
  37  * read. ref_base is the 4-bit representation of the reference base. It
  38  * is negative if we are looking at an indel or an MNP. Indel and MNP
  39  * are indistinguishable in this function. */
  40 int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r)
  41 {
  42         int i, n, ref4, is_indel, ori_depth = 0;
  43         memset(r, 0, sizeof(bcf_callret1_t));
  44         if (ref_base >= 0) {
  45                 ref4 = bam_nt16_nt4_table[ref_base];
  46                 is_indel = 0;
  47         } else ref4 = 4, is_indel = 1;
  48         if (_n == 0) return -1;
  49         // enlarge the bases array if necessary
  50         if (bca->max_bases < _n) {
  51                 bca->max_bases = _n;
  52                 kroundup32(bca->max_bases);
  53                 bca->bases = (uint16_t*)realloc(bca->bases, 2 * bca->max_bases);
  54         }
  55         // fill the bases array
  56         memset(r, 0, sizeof(bcf_callret1_t));
  57         for (i = n = 0; i < _n; ++i) {
  58                 const bam_pileup1_t *p = pl + i;
  59                 int q, b, mapQ, baseQ, is_diff, min_dist, seqQ;
  60                 // set base
  61                 if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
  62                 ++ori_depth;
  63                 baseQ = q = is_indel? p->aux&0xff : (int)bam1_qual(p->b)[p->qpos]; // base/indel quality
  64                 seqQ = is_indel? (p->aux>>8&0xff) : 99;
  65                 if (q < bca->min_baseQ) continue;
  66                 if (q > seqQ) q = seqQ;
  67                 mapQ = p->b->core.qual < bca->capQ? p->b->core.qual : bca->capQ;
  68                 if (q > mapQ) q = mapQ;
  69                 if (q > 63) q = 63;
  70                 if (q < 4) q = 4;
  71                 if (!is_indel) {
  72                         b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
  73                         b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
  74                         is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
  75                 } else {
  76                         b = p->aux>>16&0x3f;
  77                         is_diff = (b != 0);
  78                 }
  79                 bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
  80                 // collect annotations
  81                 r->qsum[b] += q;
  82                 ++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
  83                 min_dist = p->b->core.l_qseq - 1 - p->qpos;
  84                 if (min_dist > p->qpos) min_dist = p->qpos;
  85                 if (min_dist > CAP_DIST) min_dist = CAP_DIST;
  86                 r->anno[1<<2|is_diff<<1|0] += baseQ;
  87                 r->anno[1<<2|is_diff<<1|1] += baseQ * baseQ;
  88                 r->anno[2<<2|is_diff<<1|0] += mapQ;
  89                 r->anno[2<<2|is_diff<<1|1] += mapQ * mapQ;
  90                 r->anno[3<<2|is_diff<<1|0] += min_dist;
  91                 r->anno[3<<2|is_diff<<1|1] += min_dist * min_dist;
  92         }
  93         r->depth = n; r->ori_depth = ori_depth;
  94         // glfgen
  95         errmod_cal(bca->e, n, 5, bca->bases, r->p);
  96         return r->depth;
  97 }
  98 /* Combine individual calls. ref_base is a 4-bit A/C/G/T for a SNP, or
  99  * B2B_REF_MNP, or B2B_REF_INDEL. Indel and MNP are indistinguishable in
 100  * this function. */
 101 int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/, bcf_call_t *call)
 102 {
 103         int ref4, i, j, qsum[4];
 104         int64_t tmp;
 105         if (ref_base >= 0) {
 106                 call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
 107                 if (ref4 > 4) ref4 = 4;
 108         } else call->ori_ref = ref_base, ref4 = 0; // ref_base < 0
 109         // calculate qsum
 110         memset(qsum, 0, 4 * sizeof(int));
 111         for (i = 0; i < n; ++i)
 112                 for (j = 0; j < 4; ++j)
 113                         qsum[j] += calls[i].qsum[j];
 114         for (j = 0; j < 4; ++j) qsum[j] = qsum[j] << 2 | j;
 115         // find the top 2 alleles
 116         for (i = 1; i < 4; ++i) // insertion sort
 117                 for (j = i; j > 0 && qsum[j] < qsum[j-1]; --j)
 118                         tmp = qsum[j], qsum[j] = qsum[j-1], qsum[j-1] = tmp;
 119         // set the reference allele and alternative allele(s)
 120         for (i = 0; i < 5; ++i) call->a[i] = -1;
 121         call->unseen = -1;
 122         call->a[0] = ref4;
 123         for (i = 3, j = 1; i >= 0; --i) {
 124                 if ((qsum[i]&3) != ref4) {
 125                         if (qsum[i]>>2 != 0) call->a[j++] = qsum[i]&3;
 126                         else break;
 127                 }
 128         }
 129         if (ref_base >= 0) { // for SNPs, find the "unseen" base
 130                 if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
 131                         call->unseen = j, call->a[j++] = qsum[i]&3;
 132                 call->n_alleles = j;
 133         } else { // for indels and MNPs
 134                 call->n_alleles = j;
 135                 if (call->n_alleles == 1) return -1; // no reliable supporting read. stop doing anything
 136         }
 137         // set the PL array
 138         if (call->n < n) {
 139                 call->n = n;
 140                 call->PL = realloc(call->PL, 15 * n);
 141         }
 142         {
 143                 int x, g[15], z;
 144                 double sum_min = 0.;
 145                 x = call->n_alleles * (call->n_alleles + 1) / 2;
 146                 // get the possible genotypes
 147                 for (i = z = 0; i < call->n_alleles; ++i)
 148                         for (j = i; j < call->n_alleles; ++j)
 149                                 g[z++] = call->a[i] * 5 + call->a[j];
 150                 for (i = 0; i < n; ++i) {
 151                         uint8_t *PL = call->PL + x * i;
 152                         const bcf_callret1_t *r = calls + i;
 153                         float min = 1e37;
 154                         for (j = 0; j < x; ++j)
 155                                 if (min > r->p[g[j]]) min = r->p[g[j]];
 156                         sum_min += min;
 157                         for (j = 0; j < x; ++j) {
 158                                 int y;
 159                                 y = (int)(r->p[g[j]] - min + .499);
 160                                 if (y > 255) y = 255;
 161                                 PL[j] = y;
 162                         }
 163                 }
 164 //              if (ref_base < 0) fprintf(stderr, "%d,%d,%f,%d\n", call->n_alleles, x, sum_min, call->unseen);
 165                 call->shift = (int)(sum_min + .499);
 166         }
 167         // combine annotations
 168         memset(call->anno, 0, 16 * sizeof(int));
 169         for (i = call->depth = call->ori_depth = 0, tmp = 0; i < n; ++i) {
 170                 call->depth += calls[i].depth;
 171                 call->ori_depth += calls[i].ori_depth;
 172                 for (j = 0; j < 16; ++j) call->anno[j] += calls[i].anno[j];
 173         }
 174         return 0;
 175 }
 176 /* Fill a bcf1_t record. The type of the record (SNP, MNP or INDEL) is
 177  * determined by bca->ori_ref. */
 178 int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP,
 179                                  const bcf_callaux_t *bca, const char *ref)
 180 {
 181         extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
 182         kstring_t s;
 183         int i, j;
 184         b->n_smpl = bc->n;
 185         b->tid = tid; b->pos = pos; b->qual = 0;
 186         s.s = b->str; s.m = b->m_str; s.l = 0;
 187         kputc('\0', &s);
 188         if (bc->ori_ref == B2B_REF_INDEL) { // an indel
 189                 // write REF
 190                 kputc(ref[pos], &s);
 191                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 192                 kputc('\0', &s);
 193                 // write ALT
 194                 kputc(ref[pos], &s);
 195                 for (i = 1; i < 4; ++i) {
 196                         if (bc->a[i] < 0) break;
 197                         if (i > 1) {
 198                                 kputc(',', &s); kputc(ref[pos], &s);
 199                         }
 200                         if (bca->indel_types[bc->a[i]] < 0) { // deletion
 201                                 for (j = -bca->indel_types[bc->a[i]]; j < bca->indelreg; ++j)
 202                                         kputc(ref[pos+1+j], &s);
 203                         } else { // insertion; cannot be a reference unless a bug
 204                                 char *inscns = &bca->inscns[bc->a[i] * bca->maxins];
 205                                 for (j = 0; j < bca->indel_types[bc->a[i]]; ++j)
 206                                         kputc("ACGTN"[(int)inscns[j]], &s);
 207                                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
 208                         }
 209                 }
 210                 kputc('\0', &s);
 211         } else if (bc->ori_ref == B2B_REF_MNP) {
 212                 for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+j], &s);
 213                 kputc('\0', &s);
 214                 for (i = 1; i < 4; ++i) {
 215                         if (bc->a[i] < 0) break;
 216                         if (i > 1) kputc(',', &s);
 217                         for (j = 0; j < bca->indelreg; ++j)
 218                                 kputc("ACGT"[bca->indel_types[i]>>2*j&3], &s);
 219                 }
 220                 kputc('\0', &s);
 221         } else { // a SNP
 222                 kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
 223                 for (i = 1; i < 5; ++i) {
 224                         if (bc->a[i] < 0) break;
 225                         if (i > 1) kputc(',', &s);
 226                         kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
 227                 }
 228                 kputc('\0', &s);
 229         }
 230         kputc('\0', &s);
 231         // INFO
 232         if (bc->ori_ref == B2B_REF_INDEL) kputs("INDEL;", &s);
 233         else if (bc->ori_ref == B2B_REF_MNP) kputs("MNP;", &s);
 234         kputs("DP=", &s); kputw(bc->ori_depth, &s); kputs(";I16=", &s);
 235         for (i = 0; i < 16; ++i) {
 236                 if (i) kputc(',', &s);
 237                 kputw(bc->anno[i], &s);
 238         }
 239         kputc('\0', &s);
 240         // FMT
 241         kputs("PL", &s);
 242         if (bcr) {
 243                 kputs(":DP", &s);
 244                 if (is_SP) kputs(":SP", &s);
 245         }
 246         kputc('\0', &s);
 247         b->m_str = s.m; b->str = s.s; b->l_str = s.l;
 248         bcf_sync(b);
 249         memcpy(b->gi[0].data, bc->PL, b->gi[0].len * bc->n);
 250         if (bcr) {
 251                 uint16_t *dp = (uint16_t*)b->gi[1].data;
 252                 uint8_t *sp = is_SP? b->gi[2].data : 0;
 253                 for (i = 0; i < bc->n; ++i) {
 254                         bcf_callret1_t *p = bcr + i;
 255                         dp[i] = p->depth < 0xffff? p->depth : 0xffff;
 256                         if (is_SP) {
 257                                 if (p->anno[0] + p->anno[1] < 2 || p->anno[2] + p->anno[3] < 2
 258                                         || p->anno[0] + p->anno[2] < 2 || p->anno[1] + p->anno[3] < 2)
 259                                 {
 260                                         sp[i] = 0;
 261                                 } else {
 262                                         double left, right, two;
 263                                         int x;
 264                                         kt_fisher_exact(p->anno[0], p->anno[1], p->anno[2], p->anno[3], &left, &right, &two);
 265                                         x = (int)(-4.343 * log(two) + .499);
 266                                         if (x > 255) x = 255;
 267                                         sp[i] = x;
 268                                 }
 269                         }
 270                 }
 271         }
 272         return 0;
 273 }