if (theta <= 0.) theta = CALL_DEFTHETA;
bca = calloc(1, sizeof(bcf_callaux_t));
bca->capQ = 60;
- bca->openQ = 40; bca->extQ = 20; bca->tandemQ = 70;
+ bca->openQ = 40; bca->extQ = 20; bca->tandemQ = 100;
bca->min_baseQ = min_baseQ;
bca->e = errmod_init(1. - theta);
return bca;
{
if (bca == 0) return;
errmod_destroy(bca->e);
- free(bca->bases); free(bca);
+ free(bca->bases); free(bca->inscns); free(bca);
}
-
-int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base /*4-bit*/, bcf_callaux_t *bca, bcf_callret1_t *r)
+/* ref_base is the 4-bit representation of the reference base. It is
+ * negative if we are looking at an indel. */
+int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r)
{
- int i, n, ref4;
+ int i, n, ref4, is_indel, ori_depth = 0;
memset(r, 0, sizeof(bcf_callret1_t));
- ref4 = bam_nt16_nt4_table[ref_base];
+ if (ref_base >= 0) {
+ ref4 = bam_nt16_nt4_table[ref_base];
+ is_indel = 0;
+ } else ref4 = 4, is_indel = 1;
if (_n == 0) return -1;
-
// enlarge the bases array if necessary
if (bca->max_bases < _n) {
bca->max_bases = _n;
memset(r, 0, sizeof(bcf_callret1_t));
for (i = n = 0; i < _n; ++i) {
const bam_pileup1_t *p = pl + i;
- int q, b, mapQ, baseQ, is_diff, min_dist;
+ int q, b, mapQ, baseQ, is_diff, min_dist, seqQ;
// set base
- if (p->is_del || (p->b->core.flag&BAM_FUNMAP)) continue; // skip unmapped reads and deleted bases
- baseQ = q = (int)bam1_qual(p->b)[p->qpos]; // base quality
+ if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
+ ++ori_depth;
+ baseQ = q = is_indel? p->aux&0xff : (int)bam1_qual(p->b)[p->qpos]; // base/indel quality
+ seqQ = is_indel? (p->aux>>8&0xff) : 99;
if (q < bca->min_baseQ) continue;
+ if (q > seqQ) q = seqQ;
mapQ = p->b->core.qual < bca->capQ? p->b->core.qual : bca->capQ;
if (q > mapQ) q = mapQ;
if (q > 63) q = 63;
if (q < 4) q = 4;
- b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
- b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
+ if (!is_indel) {
+ b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
+ b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
+ is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
+ } else {
+ b = p->aux>>16&0x3f;
+ is_diff = (b != 0);
+ }
bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
// collect annotations
r->qsum[b] += q;
- is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
min_dist = p->b->core.l_qseq - 1 - p->qpos;
if (min_dist > p->qpos) min_dist = p->qpos;
r->anno[3<<2|is_diff<<1|0] += min_dist;
r->anno[3<<2|is_diff<<1|1] += min_dist * min_dist;
}
- r->depth = n;
+ r->depth = n; r->ori_depth = ori_depth;
// glfgen
errmod_cal(bca->e, n, 5, bca->bases, r->p);
return r->depth;
{
int ref4, i, j, qsum[4];
int64_t tmp;
- call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
- if (ref4 > 4) ref4 = 4;
+ if (ref_base >= 0) {
+ call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
+ if (ref4 > 4) ref4 = 4;
+ } else call->ori_ref = -1, ref4 = 0;
// calculate qsum
memset(qsum, 0, 4 * sizeof(int));
for (i = 0; i < n; ++i)
else break;
}
}
- if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
- call->unseen = j, call->a[j++] = qsum[i]&3;
- call->n_alleles = j;
+ if (ref_base >= 0) { // for SNPs, find the "unseen" base
+ if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
+ call->unseen = j, call->a[j++] = qsum[i]&3;
+ call->n_alleles = j;
+ } else {
+ call->n_alleles = j;
+ if (call->n_alleles == 1) return -1; // no reliable supporting read. stop doing anything
+ }
// set the PL array
if (call->n < n) {
call->n = n;
PL[j] = y;
}
}
+// if (ref_base < 0) fprintf(stderr, "%d,%d,%f,%d\n", call->n_alleles, x, sum_min, call->unseen);
call->shift = (int)(sum_min + .499);
}
// combine annotations
memset(call->anno, 0, 16 * sizeof(int));
- for (i = call->depth = 0, tmp = 0; i < n; ++i) {
+ for (i = call->depth = call->ori_depth = 0, tmp = 0; i < n; ++i) {
call->depth += calls[i].depth;
+ call->ori_depth += calls[i].ori_depth;
for (j = 0; j < 16; ++j) call->anno[j] += calls[i].anno[j];
}
return 0;
}
-int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP)
+int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP,
+ const bcf_callaux_t *bca, const char *ref)
{
extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
kstring_t s;
- int i;
+ int i, j;
b->n_smpl = bc->n;
b->tid = tid; b->pos = pos; b->qual = 0;
s.s = b->str; s.m = b->m_str; s.l = 0;
kputc('\0', &s);
- kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
- for (i = 1; i < 5; ++i) {
- if (bc->a[i] < 0) break;
- if (i > 1) kputc(',', &s);
- kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
+ if (bc->ori_ref < 0) { // an indel
+ // write REF
+ kputc(ref[pos], &s);
+ for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
+ kputc('\0', &s);
+ // write ALT
+ kputc(ref[pos], &s);
+ for (i = 1; i < 4; ++i) {
+ if (bc->a[i] < 0) break;
+ if (i > 1) {
+ kputc(',', &s); kputc(ref[pos], &s);
+ }
+ if (bca->indel_types[bc->a[i]] < 0) { // deletion
+ for (j = -bca->indel_types[bc->a[i]]; j < bca->indelreg; ++j)
+ kputc(ref[pos+1+j], &s);
+ } else { // insertion; cannot be a reference unless a bug
+ char *inscns = &bca->inscns[bc->a[i] * bca->maxins];
+ for (j = 0; j < bca->indel_types[bc->a[i]]; ++j)
+ kputc("ACGTN"[(int)inscns[j]], &s);
+ for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
+ }
+ }
+ kputc('\0', &s);
+ } else { // a SNP
+ kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
+ for (i = 1; i < 5; ++i) {
+ if (bc->a[i] < 0) break;
+ if (i > 1) kputc(',', &s);
+ kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
+ }
+ kputc('\0', &s);
}
kputc('\0', &s);
- kputc('\0', &s);
// INFO
- kputs("I16=", &s);
+ if (bc->ori_ref < 0) kputs("INDEL;", &s);
+ kputs("DP=", &s); kputw(bc->ori_depth, &s); kputs(";I16=", &s);
for (i = 0; i < 16; ++i) {
if (i) kputc(',', &s);
kputw(bc->anno[i], &s);