+static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
+{
+ pu_data_t *d = (pu_data_t*)data;
+ bam_maqindel_ret_t *r = 0;
+ int i, rb, rms_mapq = -1, *proposed_indels = 0;
+ uint64_t rms_aux;
+ uint32_t cns = 0;
+
+ // if GLF is required, suppress -c completely
+ if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
+ // if d->hash is initialized, only output the sites in the hash table
+ if (d->hash) {
+ khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
+ if (k == kh_end(d->hash)) return 0;
+ proposed_indels = kh_val(d->hash, k);
+ }
+ // update d->ref if necessary
+ if (d->fai && (int)tid != d->tid) {
+ free(d->ref);
+ d->ref = faidx_fetch_seq(d->fai, d->h->target_name[tid], 0, 0x7fffffff, &d->len);
+ d->tid = tid;
+ }
+ rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
+ // when the indel-only mode is asked for, return if no reads mapped with indels
+ if (d->format & BAM_PLF_INDEL_ONLY) {
+ for (i = 0; i < n; ++i)
+ if (pu[i].indel != 0) break;
+ if (i == n) return 0;
+ }
+ // call the consensus and indel
+ if (d->format & BAM_PLF_CNS) { // call consensus
+ if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE)) { // use a random base or the 1st base as the consensus call
+ const bam_pileup1_t *p = (d->format & BAM_PLF_1STBASE)? pu : pu + (int)(drand48() * n);
+ int q = bam1_qual(p->b)[p->qpos];
+ int mapQ = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
+ uint32_t b = bam1_seqi(bam1_seq(p->b), p->qpos);
+ cns = b<<28 | 0xf<<24 | mapQ<<16 | q<<8;
+ } else if (d->format & BAM_PLF_ALLBASE) { // collapse all bases
+ uint64_t rmsQ = 0;
+ uint32_t b = 0;
+ for (i = 0; i < n; ++i) {
+ const bam_pileup1_t *p = pu + i;
+ int q = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
+ b |= bam1_seqi(bam1_seq(p->b), p->qpos);
+ rmsQ += q * q;
+ }
+ rmsQ = (uint64_t)(sqrt((double)rmsQ / n) + .499);
+ cns = b<<28 | 0xf<<24 | rmsQ<<16 | 60<<8;
+ } else {
+ glf1_t *g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
+ cns = g->depth == 0? (0xfu<<28 | 0xf<<24) : glf2cns(g, (int)(d->c->q_r + .499));
+ free(g);
+ }
+ }
+ if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref && pos < d->len) { // call indels
+ int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
+ if (proposed_indels) // the first element gives the size of the array
+ r = bam_maqindel(m, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
+ else r = bam_maqindel(m, pos, d->ido, pu, d->ref, 0, 0);
+ }
+ // when only variant sites are asked for, test if the site is a variant
+ if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) {
+ if (!(bam_nt16_table[rb] != 15 && cns>>28 != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP
+ if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel
+ if (r) bam_maqindel_ret_destroy(r);
+ return 0;
+ }
+ }
+ }
+ // print the first 3 columns
+ fputs(d->h->target_name[tid], stdout); putchar('\t');
+ printw(pos+1, stdout); putchar('\t'); putchar(rb); putchar('\t');
+ // print consensus information if required
+ if (d->format & BAM_PLF_CNS) {
+ putchar(bam_nt16_rev_table[cns>>28]); putchar('\t');
+ printw(cns>>8&0xff, stdout); putchar('\t');
+ printw(cns&0xff, stdout); putchar('\t');
+ printw(cns>>16&0xff, stdout); putchar('\t');
+ }
+ // print pileup sequences
+ printw(n, stdout); putchar('\t');
+ for (i = 0; i < n; ++i)
+ pileup_seq(pu + i, pos, d->len, d->ref);
+ // finalize rms_mapq
+ if (d->format & BAM_PLF_CNS) {
+ for (i = rms_aux = 0; i < n; ++i) {
+ const bam_pileup1_t *p = pu + i;
+ int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
+ rms_aux += tmp * tmp;
+ }
+ rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499);
+ if (rms_mapq < 0) rms_mapq = rms_aux;
+ }
+ putchar('\t');
+ // print quality
+ for (i = 0; i < n; ++i) {
+ const bam_pileup1_t *p = pu + i;
+ int c = bam1_qual(p->b)[p->qpos] + 33;
+ if (c > 126) c = 126;
+ putchar(c);
+ }
+ if (d->format & BAM_PLF_2ND) { // print 2nd calls and qualities
+ const unsigned char *q;
+ putchar('\t');
+ for (i = 0; i < n; ++i) {
+ const bam_pileup1_t *p = pu + i;
+ q = bam_aux_get(p->b, "E2");
+ putchar(q? q[p->qpos + 1] : 'N');
+ }
+ putchar('\t');
+ for (i = 0; i < n; ++i) {
+ const bam_pileup1_t *p = pu + i;
+ q = bam_aux_get(p->b, "U2");
+ putchar(q? q[p->qpos + 1] : '!');
+ }
+ }
+ // print mapping quality if -s is flagged on the command line
+ if (d->format & BAM_PLF_SIMPLE) {
+ putchar('\t');
+ for (i = 0; i < n; ++i) {
+ int c = pu[i].b->core.qual + 33;
+ if (c > 126) c = 126;
+ putchar(c);
+ }
+ }
+ // print read position
+ if (d->format & BAM_PLF_READPOS) {
+ putchar('\t');
+ for (i = 0; i < n; ++i) {
+ int x = pu[i].qpos;
+ int l = pu[i].b->core.l_qseq;
+ printw(x < l/2? x+1 : -((l-1)-x+1), stdout); putchar(',');
+ }
+ }
+ putchar('\n');
+ // print the indel line if r has been calculated. This only happens if:
+ // a) -c or -i are flagged, AND b) the reference sequence is available
+ if (r) {
+ printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
+ if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
+ else printf("%s/%s\t", r->s[0], r->s[1]);
+ printf("%d\t%d\t", r->q_cns, r->q_ref);
+ printf("%d\t%d\t", rms_mapq, n);
+ printf("%s\t%s\t", r->s[0], r->s[1]);
+ //printf("%d\t%d\t", r->gl[0], r->gl[1]);
+ printf("%d\t%d\t%d\t", r->cnt1, r->cnt2, r->cnt_anti);
+ printf("%d\t%d\n", r->cnt_ref, r->cnt_ambi);
+ bam_maqindel_ret_destroy(r);
+ }
+ return 0;
+}
+
+int bam_pileup(int argc, char *argv[])
+{
+ int c, is_SAM = 0;
+ char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
+ pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
+ d->max_depth = 1024; d->tid = -1; d->mask = BAM_DEF_MASK; d->min_baseQ = 13;
+ d->c = bam_maqcns_init();
+ d->c->errmod = BAM_ERRMOD_MAQ2; // change the default model
+ d->ido = bam_maqindel_opt_init();
+ while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:PAQ:C:B")) >= 0) {
+ switch (c) {
+ case 'Q': d->c->min_baseQ = atoi(optarg); break;
+ case 'C': d->capQ_thres = atoi(optarg); break;
+ case 'B': d->format |= BAM_PLF_NOBAQ; break;
+ case 'a': d->c->errmod = BAM_ERRMOD_SOAP; break;
+ case 'A': d->c->errmod = BAM_ERRMOD_MAQ; break;
+ case 's': d->format |= BAM_PLF_SIMPLE; break;
+ case 't': fn_list = strdup(optarg); break;
+ case 'l': fn_pos = strdup(optarg); break;
+ case 'f': fn_fa = strdup(optarg); break;
+ case 'T': d->c->theta = atof(optarg); break;
+ case 'N': d->c->n_hap = atoi(optarg); break;
+ case 'r': d->c->het_rate = atof(optarg); d->ido->r_snp = d->c->het_rate; break;
+ case 'M': d->c->cap_mapQ = atoi(optarg); break;
+ case 'd': d->max_depth = atoi(optarg); break;
+ case 'c': d->format |= BAM_PLF_CNS; break;
+ case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
+ case 'v': d->format |= BAM_PLF_VAR_ONLY; break;
+ case 'm': d->mask = strtol(optarg, 0, 0); break;
+ case 'g': d->format |= BAM_PLF_GLF; break;
+ case '2': d->format |= BAM_PLF_2ND; break;
+ case 'P': d->format |= BAM_PLF_READPOS; break;
+ case 'I': d->ido->q_indel = atoi(optarg); break;
+ case 'G': d->ido->r_indel = atof(optarg); break;
+ case 'S': is_SAM = 1; break;
+ case 'R':
+ if (strcmp(optarg, "random") == 0) d->format |= BAM_PLF_RANBASE;
+ else if (strcmp(optarg, "first") == 0) d->format |= BAM_PLF_1STBASE;
+ else if (strcmp(optarg, "all") == 0) d->format |= BAM_PLF_ALLBASE;
+ else fprintf(pysamerr, "[bam_pileup] unrecognized -R\n");
+ break;
+ default: fprintf(pysamerr, "Unrecognizd option '-%c'.\n", c); return 1;
+ }
+ }
+ if (d->c->errmod != BAM_ERRMOD_MAQ2) d->c->theta += 0.02;
+ if (d->c->theta > 1.0) d->c->theta = 1.0;
+ if (fn_list) is_SAM = 1;
+ if (optind == argc) {
+ fprintf(pysamerr, "\n");
+ fprintf(pysamerr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
+ fprintf(pysamerr, "Option: -s simple (yet incomplete) pileup format\n");
+ fprintf(pysamerr, " -S the input is in SAM\n");
+ fprintf(pysamerr, " -B disable BAQ computation\n");
+ fprintf(pysamerr, " -A use the original MAQ model for SNP calling (DEPRECATED)\n");
+ fprintf(pysamerr, " -2 output the 2nd best call and quality\n");
+ fprintf(pysamerr, " -i only show lines/consensus with indels\n");
+ fprintf(pysamerr, " -Q INT min base quality (possibly capped by BAQ) [%d]\n", d->c->min_baseQ);
+ fprintf(pysamerr, " -C INT coefficient for adjusting mapQ of poor mappings [%d]\n", d->capQ_thres);
+ fprintf(pysamerr, " -m INT filtering reads with bits in INT [0x%x]\n", d->mask);
+ fprintf(pysamerr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
+ fprintf(pysamerr, " -d INT limit maximum depth for indels [%d]\n", d->max_depth);
+ fprintf(pysamerr, " -t FILE list of reference sequences (force -S)\n");
+ fprintf(pysamerr, " -l FILE list of sites at which pileup is output\n");
+ fprintf(pysamerr, " -f FILE reference sequence in the FASTA format\n\n");
+ fprintf(pysamerr, " -c compute the consensus sequence\n");
+ fprintf(pysamerr, " -v print variants only (for -c)\n");
+ fprintf(pysamerr, " -g output in the GLFv3 format (DEPRECATED)\n");
+ fprintf(pysamerr, " -T FLOAT theta in maq consensus calling model (for -c) [%.4g]\n", d->c->theta);
+ fprintf(pysamerr, " -N INT number of haplotypes in the sample (for -c) [%d]\n", d->c->n_hap);
+ fprintf(pysamerr, " -r FLOAT prior of a difference between two haplotypes (for -c) [%.4g]\n", d->c->het_rate);
+ fprintf(pysamerr, " -G FLOAT prior of an indel between two haplotypes (for -c) [%.4g]\n", d->ido->r_indel);
+ fprintf(pysamerr, " -I INT phred prob. of an indel in sequencing/prep. (for -c) [%d]\n", d->ido->q_indel);
+ fprintf(pysamerr, "\n");
+ fprintf(pysamerr, "Warning: Please use the `mpileup' command instead `pileup'. `Pileup' is deprecated!\n\n");
+ free(fn_list); free(fn_fa); free(d);
+ return 1;
+ }
+ if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE|BAM_PLF_ALLBASE)) d->format |= BAM_PLF_CNS;
+ if (fn_fa) d->fai = fai_load(fn_fa);
+ if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
+ if (d->format & BAM_PLF_GLF) { // for glf output
+ glf3_header_t *h;
+ h = glf3_header_init();
+ d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
+ glf3_header_write(d->fp_glf, h);
+ glf3_header_destroy(h);
+ }
+ if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
+ fprintf(pysamerr, "[bam_pileup] indels will not be called when -f is absent.\n");
+ if (fn_fa && is_SAM && fn_list == 0) fn_list = samfaipath(fn_fa);
+
+ {
+ samfile_t *fp;
+ fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
+ if (fp == 0 || fp->header == 0) {
+ fprintf(pysamerr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
+ return 1;
+ }
+ d->h = fp->header;
+ if (fn_pos) d->hash = load_pos(fn_pos, d->h);
+ { // run pileup
+ extern int bam_prob_realn(bam1_t *b, const char *ref);
+ extern int bam_cap_mapQ(bam1_t *b, char *ref, int thres);
+ bam1_t *b;
+ int ret, tid, pos, n_plp;
+ bam_plp_t iter;
+ const bam_pileup1_t *plp;
+ b = bam_init1();
+ iter = bam_plp_init(0, 0);
+ bam_plp_set_mask(iter, d->mask);
+ while ((ret = samread(fp, b)) >= 0) {
+ int skip = 0;
+ if ((int)b->core.tid < 0) break;
+ // update d->ref if necessary
+ if (d->fai && (int)b->core.tid != d->tid) {
+ free(d->ref);
+ d->ref = faidx_fetch_seq(d->fai, d->h->target_name[b->core.tid], 0, 0x7fffffff, &d->len);
+ d->tid = b->core.tid;
+ }
+ if (d->ref && (d->format&BAM_PLF_CNS) && !(d->format&BAM_PLF_NOBAQ)) bam_prob_realn(b, d->ref);
+ if (d->ref && (d->format&BAM_PLF_CNS) && d->capQ_thres > 10) {
+ int q = bam_cap_mapQ(b, d->ref, d->capQ_thres);
+ if (q < 0) skip = 1;
+ else if (b->core.qual > q) b->core.qual = q;
+ } else if (b->core.flag&BAM_FUNMAP) skip = 1;
+ else if ((d->format&BAM_PLF_CNS) && (b->core.flag&1) && !(b->core.flag&2)) skip = 1;
+ if (skip) continue;
+ bam_plp_push(iter, b);
+ while ((plp = bam_plp_next(iter, &tid, &pos, &n_plp)) != 0)
+ pileup_func(tid, pos, n_plp, plp, d);
+ }
+ bam_plp_push(iter, 0);
+ while ((plp = bam_plp_next(iter, &tid, &pos, &n_plp)) != 0)
+ pileup_func(tid, pos, n_plp, plp, d);
+ bam_plp_destroy(iter);
+ bam_destroy1(b);
+ }
+ samclose(fp); // d->h will be destroyed here
+ }
+
+ // free
+ if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
+ if (fn_pos) { // free the hash table
+ khint_t k;
+ for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
+ if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
+ kh_destroy(64, d->hash);
+ }
+ free(fn_pos); free(fn_list); free(fn_fa);
+ if (d->fai) fai_destroy(d->fai);
+ bam_maqcns_destroy(d->c);
+ free(d->ido); free(d->ref); free(d);
+ return 0;
+}
+
+/***********
+ * mpileup *
+ ***********/
+