+
+/***********
+ * mpileup *
+ ***********/
+
+typedef struct {
+ int vcf, max_mq, min_mq;
+ char *reg, *fn_pos;
+ faidx_t *fai;
+ kh_64_t *hash;
+} mplp_conf_t;
+
+typedef struct {
+ bamFile fp;
+ bam_iter_t iter;
+ int min_mq;
+} mplp_aux_t;
+
+static int mplp_func(void *data, bam1_t *b)
+{
+ mplp_aux_t *ma = (mplp_aux_t*)data;
+ int ret;
+ do {
+ ret = ma->iter? bam_iter_read(ma->fp, ma->iter, b) : bam_read1(ma->fp, b);
+ } while (b->core.qual < ma->min_mq && ret >= 0);
+ return ret;
+}
+
+static int mpileup(mplp_conf_t *conf, int n, char **fn)
+{
+ mplp_aux_t **data;
+ mc_aux_t *ma = 0;
+ int i, tid, pos, *n_plp, beg0 = 0, end0 = 1u<<29, ref_len, ref_tid;
+ const bam_pileup1_t **plp;
+ bam_mplp_t iter;
+ bam_header_t *h = 0;
+ char *ref;
+ khash_t(64) *hash = 0;
+ // allocate
+ data = calloc(n, sizeof(void*));
+ plp = calloc(n, sizeof(void*));
+ n_plp = calloc(n, sizeof(int*));
+ // read the header and initialize data
+ for (i = 0; i < n; ++i) {
+ bam_header_t *h_tmp;
+ data[i] = calloc(1, sizeof(mplp_aux_t));
+ data[i]->min_mq = conf->min_mq;
+ data[i]->fp = bam_open(fn[i], "r");
+ h_tmp = bam_header_read(data[i]->fp);
+ if (conf->reg) {
+ int beg, end;
+ bam_index_t *idx;
+ idx = bam_index_load(fn[i]);
+ if (idx == 0) {
+ fprintf(stderr, "[%s] fail to load index for %d-th input.\n", __func__, i+1);
+ exit(1);
+ }
+ if (bam_parse_region(h_tmp, conf->reg, &tid, &beg, &end) < 0) {
+ fprintf(stderr, "[%s] malformatted region or wrong seqname for %d-th input.\n", __func__, i+1);
+ exit(1);
+ }
+ if (i == 0) beg0 = beg, end0 = end;
+ data[i]->iter = bam_iter_query(idx, tid, beg, end);
+ bam_index_destroy(idx);
+ }
+ if (i == 0) h = h_tmp;
+ else {
+ // FIXME: to check consistency
+ bam_header_destroy(h_tmp);
+ }
+ }
+ if (conf->fn_pos) hash = load_pos(conf->fn_pos, h);
+ // write the VCF header
+ if (conf->vcf) {
+ kstring_t s;
+ s.l = s.m = 0; s.s = 0;
+ kputs("#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT", &s);
+ for (i = 0; i < n; ++i) {
+ const char *p;
+ kputc('\t', &s);
+ if ((p = strstr(fn[i], ".bam")) != 0)
+ kputsn(fn[i], p - fn[i], &s);
+ else kputs(fn[i], &s);
+ }
+ puts(s.s);
+ free(s.s);
+ }
+ // mpileup
+ if (conf->vcf) ma = mc_init(n);
+ ref_tid = -1; ref = 0;
+ iter = bam_mplp_init(n, mplp_func, (void**)data);
+ while (bam_mplp_auto(iter, &tid, &pos, n_plp, plp) > 0) {
+ if (conf->reg && (pos < beg0 || pos >= end0)) continue; // out of the region requested
+ if (hash) {
+ khint_t k;
+ k = kh_get(64, hash, (uint64_t)tid<<32 | pos);
+ if (k == kh_end(hash)) continue;
+ }
+ if (tid != ref_tid) {
+ free(ref);
+ if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len);
+ ref_tid = tid;
+ }
+ if (conf->vcf) {
+ double f0, f; // the reference allele frequency
+ int j, _ref, _alt, _ref0, depth, rms_q, _ref0b;
+ uint64_t sqr_sum;
+ _ref0 = _ref0b = (ref && pos < ref_len)? ref[pos] : 'N';
+ _ref0 = bam_nt16_nt4_table[bam_nt16_table[_ref0]];
+ f = f0 = mc_freq0(_ref0, n_plp, plp, ma, &_ref, &_alt);
+ if (f >= 0.0) {
+ double flast = f;
+ for (j = 0; j < 10; ++j) {
+ f = mc_freq_iter(flast, ma);
+ if (fabs(f - flast) < 1e-3) break;
+ flast = f;
+ }
+ }
+ printf("%s\t%d\t.\t%c\t", h->target_name[tid], pos + 1, _ref0b);
+ if (_ref0 == _ref) putchar("ACGTN"[_alt]);
+ else printf("%c,%c", "ACGTN"[_ref], "ACGTN"[_alt]);
+ printf("\t0\t"); // FIXME: currently these not available
+ if (f0 < 0.) printf("Q13\t");
+ else printf(".\t");
+ for (i = depth = 0, sqr_sum = 0; i < n; ++i) {
+ depth += n_plp[i];
+ for (j = 0; j < n_plp[i]; ++j) {
+ int q = plp[i][j].b->core.qual;
+ if (q > conf->max_mq) q = conf->max_mq;
+ sqr_sum += q * q;
+ }
+ }
+ rms_q = (int)(sqrt((double)sqr_sum / depth) + .499);
+ printf("DP=%d;MQ=%d;AF=%.3lg", depth, rms_q, f0<0.?0.:1.-f);
+ printf("\tDP");
+ // output genotype information; FIXME: to be implmented...
+ for (i = 0; i < n; ++i)
+ printf("\t%d", n_plp[i]);
+ putchar('\n');
+ } else {
+ printf("%s\t%d\t%c", h->target_name[tid], pos + 1, (ref && pos < ref_len)? ref[pos] : 'N');
+ for (i = 0; i < n; ++i) {
+ int j;
+ printf("\t%d\t", n_plp[i]);
+ if (n_plp[i] == 0) printf("*\t*");
+ else {
+ for (j = 0; j < n_plp[i]; ++j)
+ pileup_seq(plp[i] + j, pos, ref_len, ref);
+ putchar('\t');
+ for (j = 0; j < n_plp[i]; ++j) {
+ const bam_pileup1_t *p = plp[i] + j;
+ int c = bam1_qual(p->b)[p->qpos] + 33;
+ if (c > 126) c = 126;
+ putchar(c);
+ }
+ }
+ }
+ putchar('\n');
+ }
+ }
+ if (hash) { // free the hash table
+ khint_t k;
+ for (k = kh_begin(hash); k < kh_end(hash); ++k)
+ if (kh_exist(hash, k)) free(kh_val(hash, k));
+ kh_destroy(64, hash);
+ }
+ mc_destroy(ma);
+ bam_mplp_destroy(iter);
+ bam_header_destroy(h);
+ for (i = 0; i < n; ++i) {
+ bam_close(data[i]->fp);
+ if (data[i]->iter) bam_iter_destroy(data[i]->iter);
+ free(data[i]);
+ }
+ free(data); free(plp); free(ref); free(n_plp);
+ return 0;
+}
+
+int bam_mpileup(int argc, char *argv[])
+{
+ int c;
+ mplp_conf_t mplp;
+ memset(&mplp, 0, sizeof(mplp_conf_t));
+ mplp.max_mq = 60;
+ while ((c = getopt(argc, argv, "f:r:l:VM:q:")) >= 0) {
+ switch (c) {
+ case 'f':
+ mplp.fai = fai_load(optarg);
+ if (mplp.fai == 0) return 1;
+ break;
+ case 'r': mplp.reg = strdup(optarg); break;
+ case 'l': mplp.fn_pos = strdup(optarg); break;
+ case 'V': mplp.vcf = 1; break;
+ case 'M': mplp.max_mq = atoi(optarg); break;
+ case 'q': mplp.min_mq = atoi(optarg); break;
+ }
+ }
+ if (argc == 1) {
+ fprintf(stderr, "Usage: samtools mpileup [-r reg] [-f in.fa] [-l pos] in1.bam [in2.bam [...]]\n");
+ return 1;
+ }
+ mpileup(&mplp, argc - optind, argv + optind);
+ free(mplp.reg);
+ if (mplp.fai) fai_destroy(mplp.fai);
+ return 0;
+}