+ for (s = n_alt = 0; s < n; ++s) {
+ for (i = 0; i < n_plp[s]; ++i) {
+ bam_pileup1_t *p = plp[s] + i;
+ int x = types[p->aux>>16&0x3f];
+ for (j = 0; j < 4; ++j)
+ if (x == bca->indel_types[j]) break;
+ p->aux = j<<16 | (j == 4? 0 : (p->aux&0xffff));
+ if ((p->aux>>16&0x3f) > 0) ++n_alt;
+// fprintf(stderr, "X pos=%d read=%d:%d name=%s call=%d type=%d q=%d seqQ=%d\n", pos, s, i, bam1_qname(p->b), p->aux>>16&63, bca->indel_types[p->aux>>16&63], p->aux&0xff, p->aux>>8&0xff);
+ }
+ }
+ }
+ free(score1); free(score2);
+ // free
+ free(types); free(inscns);
+ return n_alt > 0? 0 : -1;
+}
+
+#define END_SKIP 4 // must be larger than B2B_MAX_MNP
+#define MIN_MNP_FLANK_BAQ 20
+
+// the following routine is modified from bcf_call_gap_prep()
+int bcf_call_mnp_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_callaux_t *bca, const char *ref)
+{
+ extern void ks_introsort_uint32_t(int, uint32_t*);
+ int i, s, j, k, t, n_types, *types, max_rd_len, left, right, *score1;
+ int N, K, ref_seq, ref_type, n_alt;
+ char *ref2, *query;
+ uint8_t *_seqQ;
+ if (ref == 0 || bca == 0) return -1;
+ if (pos < bca->last_mnp_pos + B2B_MNP_WIN) return -2; // to avoid calling a TNP multiple times
+ if (pos < END_SKIP || ref[pos] == 0 || ref[pos+1] == 0) return -2; // end of the reference
+ { // determine if there is an MNP
+ int r[2];
+ uint8_t *tt;
+ r[0] = bam_nt16_table[(int)ref[pos]];
+ r[1] = bam_nt16_table[(int)ref[pos+1]];
+ for (s = 0; s < n; ++s) {
+ for (i = 0; i < n_plp[s]; ++i) {
+ bam_pileup1_t *p = plp[s] + i;
+ int left, rght;
+ if (p->qpos < END_SKIP || p->qpos >= p->b->core.l_qseq - END_SKIP) continue;
+ tt = bam1_seq(p->b);
+ if (bam1_seqi(tt, p->qpos) == r[0] || bam1_seqi(tt, p->qpos+1) == r[1]) continue; // no MNP
+ tt = bam1_qual(p->b);
+ for (left = 0, k = p->qpos - END_SKIP; k <= p->qpos; ++k)
+ left = left > tt[k]? left : tt[k];
+ for (rght = 0, k = p->qpos; k < p->b->core.l_qseq - END_SKIP; ++k)
+ rght = rght > tt[k]? left : tt[k];
+ if (left >= MIN_MNP_FLANK_BAQ && rght >= MIN_MNP_FLANK_BAQ) break; // bracketed by good bases
+ }
+ if (i != n_plp[s]) break;
+ }
+ if (s == n) return -1; // there is no MNP at this position.
+ }
+ for (s = N = 0; s < n; ++s) N += n_plp[s]; // N is the total number of reads
+ { // construct the MNP consensus
+ uint8_t *tt;
+ uint32_t *aux, x, *cnt;
+ int m;
+ m = max_rd_len = 0;
+ aux = calloc(N + 1, 4);
+ for (i = 0, x = 0; i < B2B_MAX_MNP; ++i)
+ x |= bam_nt16_nt4_table[bam_nt16_table[(int)ref[pos+i]]] << 2*i;
+ ref_seq = aux[m++] = x;
+ bca->indelreg = 0;
+ for (s = 0; s < n; ++s) {
+ for (i = 0; i < n_plp[s]; ++i) {
+ bam_pileup1_t *p = plp[s] + i;
+ int stop;
+ j = bam_cigar2qlen(&p->b->core, bam1_cigar(p->b));
+ if (j > max_rd_len) max_rd_len = j;
+ if (p->qpos < END_SKIP || p->qpos >= p->b->core.l_qseq - END_SKIP) continue;
+ tt = bam1_seq(p->b);
+ for (k = j = stop = 0, x = 0; k < B2B_MAX_MNP; ++k) {
+ int c = bam_nt16_nt4_table[bam1_seqi(tt, p->qpos + k)];
+ if (c > 3) break;
+ if (c != (ref_seq>>k*2&3) && !stop) ++j;
+ else stop = 1;
+ x |= c << 2*k;
+ }
+// fprintf(stderr, "* %d, %d\n", x, j);
+ if (k == B2B_MAX_MNP && j >= 2) aux[m++] = x;
+ }
+ }
+ ks_introsort(uint32_t, m, aux);
+ // squeeze out indentical types
+ for (i = 1, n_types = 1; i < m; ++i)
+ if (aux[i] != aux[i-1]) ++n_types;
+ if (n_types == 1) { // no indels
+ free(aux); return -1;
+ }
+ // count reads for each type
+ cnt = alloca(n_types * 4);
+ cnt[0] = 1<<8 | aux[0];
+ for (i = 1, t = 0; i < m; ++i) {
+ if (aux[i] != aux[i-1]) {
+ ++t;
+ cnt[t] = 1<<8 | aux[i];
+ } else cnt[t] += 1<<8;
+ }
+ free(aux);
+ // collect types
+ ks_introsort(uint32_t, n_types, cnt);
+ for (t = 0; t < n_types; ++t)
+ if ((cnt[t]&0xff) == ref_seq) break;
+ if (t == n_types - 1) --t; // we get at least two types
+ types = (int*)calloc(n_types, sizeof(int));
+ for (i = 0; t < n_types; ++i, ++t) types[i] = cnt[t]&0xff;
+ n_types = i;
+ // calculate MNP length
+ for (i = B2B_MAX_MNP - 1, k = 0; i >= 0; --i) {
+ int c = types[0] >> 2*i & 3;
+ for (t = 1; t < n_types; ++t)
+ if ((types[t] >> 2*i & 3) != c) break;
+ if (t == n_types) ++k;
+ else break;
+ }
+ bca->indelreg = B2B_MAX_MNP - k;
+ x = (1<<2*bca->indelreg) - 1;
+ for (t = 0; t < n_types; ++t) types[t] &= x;
+ ref_seq &= x;
+ for (t = 0; t < n_types; ++t)
+ if (types[t] == ref_seq) break;
+ ref_type = t;
+// fprintf(stderr, "%d, %d, [%x,%x] %d\n", pos, n_types, types[0], types[1], bca->indelreg);
+ }
+ { // calculate left and right boundary
+ left = pos > INDEL_WINDOW_SIZE? pos - INDEL_WINDOW_SIZE : 0;
+ right = pos + INDEL_WINDOW_SIZE;
+ // in case the alignments stand out the reference
+ for (i = pos; i < right; ++i)
+ if (ref[i] == 0) break;
+ right = i;
+ }
+ { // find the highest base quality
+ _seqQ = calloc(N, 1);