#include "sam.h"
#include "faidx.h"
#include "bam_maqcns.h"
-#include "bam_mcns.h"
#include "khash.h"
#include "glf.h"
#include "kstring.h"
#define BAM_PLF_1STBASE 0x80
#define BAM_PLF_ALLBASE 0x100
#define BAM_PLF_READPOS 0x200
+#define BAM_PLF_NOBAQ 0x400
typedef struct {
bam_header_t *h;
uint32_t format;
int tid, len, last_pos;
int mask;
+ int capQ_thres, min_baseQ;
int max_depth; // for indel calling, ignore reads with the depth too high. 0 for unlimited
char *ref;
glfFile fp_glf; // for glf output only
}
rmsQ = (uint64_t)(sqrt((double)rmsQ / n) + .499);
cns = b<<28 | 0xf<<24 | rmsQ<<16 | 60<<8;
- } else cns = bam_maqcns_call(n, pu, d->c);
+ } 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;
}
// 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 != bam_nt16_table[rb])) { // not a SNP
+ 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
printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
// print consensus information if required
- if (d->format & BAM_PLF_CNS) {
- int ref_q, rb4 = bam_nt16_table[rb];
- ref_q = 0;
- if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP
- ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff);
- if (ref_q > 255) ref_q = 255;
- }
- rms_mapq = cns>>16&0xff;
- printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq);
- }
+ if (d->format & BAM_PLF_CNS)
+ printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, cns&0xff, cns>>16&0xff);
// print pileup sequences
printf("%d\t", n);
rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line
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 = 0;
- d->tid = -1; d->mask = BAM_DEF_MASK;
+ d->max_depth = 1024; d->tid = -1; d->mask = BAM_DEF_MASK; d->min_baseQ = 13;
d->c = bam_maqcns_init();
- d->c->is_soap = 1; // change the default model
+ 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:PA")) >= 0) {
+ while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:PAQ:C:B")) >= 0) {
switch (c) {
- case 'a': d->c->is_soap = 1; break;
- case 'A': d->c->is_soap = 0; break;
+ 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;
default: fprintf(stderr, "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(stderr, "\n");
fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
fprintf(stderr, " -S the input is in SAM\n");
- fprintf(stderr, " -A use the MAQ model for SNP calling\n");
+ fprintf(stderr, " -B disable BAQ computation\n");
+ fprintf(stderr, " -A use the original MAQ model for SNP calling (DEPRECATED)\n");
fprintf(stderr, " -2 output the 2nd best call and quality\n");
fprintf(stderr, " -i only show lines/consensus with indels\n");
- fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
+ fprintf(stderr, " -Q INT min base quality (possibly capped by BAQ) [%d]\n", d->c->min_baseQ);
+ fprintf(stderr, " -C INT coefficient for adjusting mapQ of poor mappings [%d]\n", d->capQ_thres);
+ fprintf(stderr, " -m INT filtering reads with bits in INT [0x%x]\n", d->mask);
fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
- fprintf(stderr, " -d INT limit maximum depth for indels [unlimited]\n");
+ fprintf(stderr, " -d INT limit maximum depth for indels [%d]\n", d->max_depth);
fprintf(stderr, " -t FILE list of reference sequences (force -S)\n");
fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
- fprintf(stderr, " -c output the SOAPsnp consensus sequence\n");
+ fprintf(stderr, " -c compute the consensus sequence\n");
fprintf(stderr, " -v print variants only (for -c)\n");
- fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
- fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
- fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
- fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
- fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
- fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
+ fprintf(stderr, " -g output in the GLFv3 format (DEPRECATED)\n");
+ fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c) [%.4g]\n", d->c->theta);
+ fprintf(stderr, " -N INT number of haplotypes in the sample (for -c) [%d]\n", d->c->n_hap);
+ fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c) [%.4g]\n", d->c->het_rate);
+ fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c) [%.4g]\n", d->ido->r_indel);
+ fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c) [%d]\n", d->ido->q_indel);
fprintf(stderr, "\n");
free(fn_list); free(fn_fa); free(d);
return 1;
}
d->h = fp->header;
if (fn_pos) d->hash = load_pos(fn_pos, d->h);
- sampileup(fp, d->mask, pileup_func, d);
+ { // 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;
+ // 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
}
* mpileup *
***********/
+#include <assert.h>
+#include "bam2bcf.h"
+#include "sample.h"
+
+#define MPLP_GLF 0x10
+#define MPLP_NO_COMP 0x20
+#define MPLP_NO_ORPHAN 0x40
+#define MPLP_REALN 0x80
+
typedef struct {
- int vcf, max_mq, min_mq;
+ int max_mq, min_mq, flag, min_baseQ, capQ_thres;
char *reg, *fn_pos;
faidx_t *fai;
kh_64_t *hash;
typedef struct {
bamFile fp;
bam_iter_t iter;
- int min_mq;
+ int min_mq, flag, ref_id, capQ_thres;
+ char *ref;
} mplp_aux_t;
+typedef struct {
+ int n;
+ int *n_plp, *m_plp;
+ bam_pileup1_t **plp;
+} mplp_pileup_t;
+
static int mplp_func(void *data, bam1_t *b)
{
+ extern int bam_realn(bam1_t *b, const char *ref);
+ extern int bam_prob_realn(bam1_t *b, const char *ref);
+ extern int bam_cap_mapQ(bam1_t *b, char *ref, int thres);
mplp_aux_t *ma = (mplp_aux_t*)data;
- int ret;
+ int ret, skip = 0;
do {
+ int has_ref = (ma->ref && ma->ref_id == b->core.tid)? 1 : 0;
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);
+ if (ret < 0) break;
+ skip = 0;
+ if (has_ref && (ma->flag&MPLP_REALN)) bam_prob_realn(b, ma->ref);
+ if (has_ref && ma->capQ_thres > 10) {
+ int q = bam_cap_mapQ(b, ma->ref, ma->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 (b->core.qual < ma->min_mq) skip = 1;
+ else if ((ma->flag&MPLP_NO_ORPHAN) && (b->core.flag&1) && !(b->core.flag&2)) skip = 1;
+ } while (skip);
return ret;
}
+static void group_smpl(mplp_pileup_t *m, bam_sample_t *sm, kstring_t *buf,
+ int n, char *const*fn, int *n_plp, const bam_pileup1_t **plp)
+{
+ int i, j;
+ memset(m->n_plp, 0, m->n * sizeof(int));
+ for (i = 0; i < n; ++i) {
+ for (j = 0; j < n_plp[i]; ++j) {
+ const bam_pileup1_t *p = plp[i] + j;
+ uint8_t *q;
+ int id = -1;
+ q = bam_aux_get(p->b, "RG");
+ if (q) id = bam_smpl_rg2smid(sm, fn[i], (char*)q+1, buf);
+ if (id < 0) id = bam_smpl_rg2smid(sm, fn[i], 0, buf);
+ assert(id >= 0 && id < m->n);
+ if (m->n_plp[id] == m->m_plp[id]) {
+ m->m_plp[id] = m->m_plp[id]? m->m_plp[id]<<1 : 8;
+ m->plp[id] = realloc(m->plp[id], sizeof(bam_pileup1_t) * m->m_plp[id]);
+ }
+ m->plp[id][m->n_plp[id]++] = *p;
+ }
+ }
+}
+
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
+
+ bcf_callaux_t *bca = 0;
+ bcf_callret1_t *bcr = 0;
+ bcf_call_t bc;
+ bcf_t *bp = 0;
+ bcf_hdr_t *bh = 0;
+
+ bam_sample_t *sm = 0;
+ kstring_t buf;
+ mplp_pileup_t gplp;
+
+ memset(&gplp, 0, sizeof(mplp_pileup_t));
+ memset(&buf, 0, sizeof(kstring_t));
+ memset(&bc, 0, sizeof(bcf_call_t));
data = calloc(n, sizeof(void*));
plp = calloc(n, sizeof(void*));
n_plp = calloc(n, sizeof(int*));
+ sm = bam_smpl_init();
+
// 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");
+ data[i]->flag = conf->flag;
+ data[i]->capQ_thres = conf->capQ_thres;
+ data[i]->fp = strcmp(fn[i], "-") == 0? bam_dopen(fileno(stdin), "r") : bam_open(fn[i], "r");
h_tmp = bam_header_read(data[i]->fp);
+ bam_smpl_add(sm, fn[i], h_tmp->text);
if (conf->reg) {
int beg, end;
bam_index_t *idx;
bam_header_destroy(h_tmp);
}
}
+ gplp.n = sm->n;
+ gplp.n_plp = calloc(sm->n, sizeof(int));
+ gplp.m_plp = calloc(sm->n, sizeof(int));
+ gplp.plp = calloc(sm->n, sizeof(void*));
+
+ fprintf(stderr, "[%s] %d samples in %d input files\n", __func__, sm->n, n);
if (conf->fn_pos) hash = load_pos(conf->fn_pos, h);
// write the VCF header
- if (conf->vcf) {
+ if (conf->flag & MPLP_GLF) {
kstring_t s;
+ bh = calloc(1, sizeof(bcf_hdr_t));
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);
+ bp = bcf_open("-", (conf->flag&MPLP_NO_COMP)? "wu" : "w");
+ for (i = 0; i < h->n_targets; ++i) {
+ kputs(h->target_name[i], &s);
+ kputc('\0', &s);
+ }
+ bh->l_nm = s.l;
+ bh->name = malloc(s.l);
+ memcpy(bh->name, s.s, s.l);
+ s.l = 0;
+ for (i = 0; i < sm->n; ++i) {
+ kputs(sm->smpl[i], &s); kputc('\0', &s);
}
- puts(s.s);
+ bh->l_smpl = s.l;
+ bh->sname = malloc(s.l);
+ memcpy(bh->sname, s.s, s.l);
+ bh->l_txt = 0;
free(s.s);
+ bcf_hdr_sync(bh);
+ bcf_hdr_write(bp, bh);
+ bca = bcf_call_init(-1., conf->min_baseQ);
+ bcr = calloc(sm->n, sizeof(bcf_callret1_t));
}
- // 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 (k == kh_end(hash)) continue;
}
if (tid != ref_tid) {
- free(ref);
+ free(ref); ref = 0;
if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len);
+ for (i = 0; i < n; ++i) data[i]->ref = ref, data[i]->ref_id = tid;
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');
+ if (conf->flag & MPLP_GLF) {
+ int _ref0, ref16;
+ bcf1_t *b = calloc(1, sizeof(bcf1_t));
+ group_smpl(&gplp, sm, &buf, n, fn, n_plp, plp);
+ _ref0 = (ref && pos < ref_len)? ref[pos] : 'N';
+ ref16 = bam_nt16_table[_ref0];
+ for (i = 0; i < gplp.n; ++i)
+ bcf_call_glfgen(gplp.n_plp[i], gplp.plp[i], ref16, bca, bcr + i);
+ bcf_call_combine(gplp.n, bcr, ref16, &bc);
+ bcf_call2bcf(tid, pos, &bc, b);
+ bcf_write(bp, bh, b);
+ bcf_destroy(b);
} 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) {
putchar('\n');
}
}
+
+ bcf_close(bp);
+ bam_smpl_destroy(sm); free(buf.s);
+ for (i = 0; i < gplp.n; ++i) free(gplp.plp[i]);
+ free(gplp.plp); free(gplp.n_plp); free(gplp.m_plp);
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);
+ bcf_hdr_destroy(bh); bcf_call_destroy(bca); free(bc.PL); free(bcr);
bam_mplp_destroy(iter);
bam_header_destroy(h);
for (i = 0; i < n; ++i) {
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) {
+ mplp.min_baseQ = 13;
+ mplp.capQ_thres = 0;
+ while ((c = getopt(argc, argv, "gf:r:l:M:q:Q:uaORC:")) >= 0) {
switch (c) {
case 'f':
mplp.fai = fai_load(optarg);
break;
case 'r': mplp.reg = strdup(optarg); break;
case 'l': mplp.fn_pos = strdup(optarg); break;
- case 'V': mplp.vcf = 1; break;
+ case 'g': mplp.flag |= MPLP_GLF; break;
+ case 'u': mplp.flag |= MPLP_NO_COMP | MPLP_GLF; break;
+ case 'a': mplp.flag |= MPLP_NO_ORPHAN | MPLP_REALN; break;
+ case 'O': mplp.flag |= MPLP_NO_ORPHAN; break;
+ case 'R': mplp.flag |= MPLP_REALN; break;
+ case 'C': mplp.capQ_thres = atoi(optarg); break;
case 'M': mplp.max_mq = atoi(optarg); break;
case 'q': mplp.min_mq = atoi(optarg); break;
+ case 'Q': mplp.min_baseQ = atoi(optarg); break;
}
}
if (argc == 1) {
- fprintf(stderr, "Usage: samtools mpileup [-r reg] [-f in.fa] [-l pos] in1.bam [in2.bam [...]]\n");
+ fprintf(stderr, "\n");
+ fprintf(stderr, "Usage: samtools mpileup [options] in1.bam [in2.bam [...]]\n\n");
+ fprintf(stderr, "Options: -f FILE reference sequence file [null]\n");
+ fprintf(stderr, " -r STR region in which pileup is generated [null]\n");
+ fprintf(stderr, " -l FILE list of positions (format: chr pos) [null]\n");
+ fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", mplp.max_mq);
+ fprintf(stderr, " -Q INT min base quality [%d]\n", mplp.min_baseQ);
+ fprintf(stderr, " -q INT filter out alignment with MQ smaller than INT [%d]\n", mplp.min_mq);
+ fprintf(stderr, " -g generate BCF output\n");
+ fprintf(stderr, " -u do not compress BCF output\n");
+ fprintf(stderr, "\n");
+ fprintf(stderr, "Notes: Assuming diploid individuals.\n\n");
return 1;
}
mpileup(&mplp, argc - optind, argv + optind);