#include "sam.h"
#include "faidx.h"
#include "bam_maqcns.h"
-#include "bam_mcns.h"
-#include "bam2bcf.h"
#include "khash.h"
#include "glf.h"
#include "kstring.h"
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");
* mpileup *
***********/
-#define MPLP_VCF 0x1
-#define MPLP_VAR 0x2
-#define MPLP_AFALL 0x8
+#include <assert.h>
+#include "bam2bcf.h"
+#include "sample.h"
+
#define MPLP_GLF 0x10
#define MPLP_NO_COMP 0x20
-
-#define MPLP_AFS_BLOCK 0x10000
+#define MPLP_NO_ORPHAN 0x40
+#define MPLP_REALN 0x80
typedef struct {
- int max_mq, min_mq, prior_type, flag, min_baseQ;
+ int max_mq, min_mq, flag, min_baseQ, capQ_thres;
double theta;
char *reg, *fn_pos;
faidx_t *fai;
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_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_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;
const bam_pileup1_t **plp;
bam_mplp_t iter;
bam_header_t *h = 0;
- uint64_t N = 0;
char *ref;
khash_t(64) *hash = 0;
- mc_aux_t *ma = 0;
-
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]->flag = conf->flag;
+ data[i]->capQ_thres = conf->capQ_thres;
data[i]->fp = 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->flag & MPLP_GLF) {
bh->name = malloc(s.l);
memcpy(bh->name, s.s, s.l);
s.l = 0;
- for (i = 0; i < n; ++i) {
- const char *p;
- if ((p = strstr(fn[i], ".bam")) != 0)
- kputsn(fn[i], p - fn[i], &s);
- else kputs(fn[i], &s);
- kputc('\0', &s);
+ for (i = 0; i < sm->n; ++i) {
+ kputs(sm->smpl[i], &s); kputc('\0', &s);
}
bh->l_smpl = s.l;
bh->sname = malloc(s.l);
free(s.s);
bcf_hdr_sync(bh);
bcf_hdr_write(bp, bh);
- } else if (conf->flag & MPLP_VCF) {
- kstring_t s;
- s.l = s.m = 0; s.s = 0;
- puts("##fileformat=VCFv4.0");
- puts("##INFO=<ID=DP,Number=1,Type=Integer,Description=\"Total read depth\">");
- puts("##INFO=<ID=AF,Number=1,Type=Float,Description=\"Non-reference allele frequency \\argmax_f P(D|f)\">");
- puts("##INFO=<ID=AFE,Number=1,Type=Float,Description=\"Expected non-reference allele frequency\">");
- puts("##FILTER=<ID=Q13,Description=\"All min{baseQ,mapQ} below 13\">");
- puts("##FILTER=<ID=FPE,Description=\"Floating point error\">");
- 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->flag & MPLP_GLF) {
bca = bcf_call_init(-1., conf->min_baseQ);
- bcr = calloc(n, sizeof(bcf_callret1_t));
- } else if (conf->flag & MPLP_VCF) {
- ma = mc_init(n);
- mc_init_prior(ma, conf->prior_type, conf->theta);
+ bcr = calloc(sm->n, sizeof(bcf_callret1_t));
}
ref_tid = -1; ref = 0;
iter = bam_mplp_init(n, mplp_func, (void**)data);
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->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 < n; ++i)
- bcf_call_glfgen(n_plp[i], plp[i], ref16, bca, bcr + i);
- bcf_call_combine(n, bcr, ref16, &bc);
+ 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);
- //fprintf(stderr, "%d,%d,%d\n", b->tid, b->pos, b->l_str);
bcf_destroy(b);
- } else if (conf->flag & MPLP_VCF) {
- mc_rst_t r;
- int j, _ref0, depth, rms_q, _ref0b, is_var = 0, qref = 0, level = 2, tot;
- uint64_t sqr_sum;
- _ref0 = _ref0b = (ref && pos < ref_len)? ref[pos] : 'N';
- _ref0 = bam_nt16_nt4_table[bam_nt16_table[_ref0]];
- tot = mc_cal(_ref0, n_plp, plp, ma, &r, level);
- if (tot) { // has good bases
- double q;
- is_var = (r.p_ref < .5);
- q = is_var? r.p_ref : 1. - r.p_ref;
- if (q < 1e-308) q = 1e-308;
- qref = (int)(-3.434 * log(q) + .499);
- if (qref > 99) qref = 99;
- }
- if ((conf->flag & MPLP_VAR) && !is_var) continue;
- ++N; // number of processed lines
- printf("%s\t%d\t.\t%c\t", h->target_name[tid], pos + 1, _ref0b);
- if (is_var) {
- putchar("ACGTN"[r.alt]);
- if (r.alt2 >= 0 && r.alt2 < 4) printf(",%c", "ACGT"[r.alt2]);
- } else putchar('.');
- printf("\t%d\t", qref);
- if (!tot) printf("Q13\t");
- else if (r.f_exp < 0.) printf("FPE\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", depth, rms_q);
- if (tot) {
- printf(";AF=%.3lf", 1. - r.f_em);
- if (level >= 2) printf(";AFE=%.3lf", 1-r.f_exp);
- if (conf->flag & MPLP_AFALL)
- printf(";AF0=%.3lf;AFN=%.3lf", 1-r.f_naive, 1-r.f_nielsen);
- }
- printf("\tGT:GQ:DP");
- if (tot) {
- for (i = 0; i < n; ++i) {
- int x = mc_call_gt(ma, r.f_exp, i);
- printf("\t%c/%c:%d:%d", "10"[((x&3)==2)], "10"[((x&3)>0)], x>>2, n_plp[i]);
- }
- } else for (i = 0; i < n; ++i) printf("\t./.:0:0");
- putchar('\n');
- if (N % MPLP_AFS_BLOCK == 0) mc_dump_afs(ma);
} 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);
- if (conf->flag&MPLP_VCF) mc_dump_afs(ma);
+ 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)
kh_destroy(64, hash);
}
bcf_hdr_destroy(bh); bcf_call_destroy(bca); free(bc.PL); free(bcr);
- mc_destroy(ma);
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;
- mplp.prior_type = MC_PTYPE_FULL;
mplp.theta = 1e-3;
mplp.min_baseQ = 13;
- while ((c = getopt(argc, argv, "gvVcFSP:f:r:l:VM:q:t:Q:u")) >= 0) {
+ mplp.capQ_thres = 0;
+ while ((c = getopt(argc, argv, "gf:r:l:M:q:t:Q:uaORC:")) >= 0) {
switch (c) {
case 't': mplp.theta = atof(optarg); break;
- case 'P':
- if (strcmp(optarg, "full") == 0) mplp.prior_type = MC_PTYPE_FULL;
- else if (strcmp(optarg, "cond2") == 0) mplp.prior_type = MC_PTYPE_COND2;
- else if (strcmp(optarg, "flat") == 0) mplp.prior_type = MC_PTYPE_FLAT;
- else {
- fprintf(stderr, "[%s] unrecognized prior type.\n", __func__);
- return 1;
- }
- break;
case 'f':
mplp.fai = fai_load(optarg);
if (mplp.fai == 0) return 1;
case 'r': mplp.reg = strdup(optarg); break;
case 'l': mplp.fn_pos = strdup(optarg); break;
case 'g': mplp.flag |= MPLP_GLF; break;
- case 'V':
- case 'c': mplp.flag |= MPLP_VCF; break;
- case 'F': mplp.flag |= MPLP_AFALL; break;
- case 'v': mplp.flag |= MPLP_VAR; break;
case 'u': mplp.flag |= MPLP_NO_COMP; 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 (mplp.flag&MPLP_GLF) mplp.flag &= ~MPLP_VCF;
if (argc == 1) {
fprintf(stderr, "\n");
fprintf(stderr, "Usage: samtools mpileup [options] in1.bam [in2.bam [...]]\n\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, " -t FLOAT scaled mutation rate [%lg]\n", mplp.theta);
- fprintf(stderr, " -P STR prior: full, flat, cond2 [full]\n");
- fprintf(stderr, " -Q INT min base quality [%d]\n", mplp.min_baseQ);
- fprintf(stderr, " -c generate VCF output (consensus calling)\n");
- fprintf(stderr, " -g generate GLF output\n");
- fprintf(stderr, " -v show variant sites only\n");
+ fprintf(stderr, " -g generate BCF output\n");
+ fprintf(stderr, " -u do not compress BCF output\n");
fprintf(stderr, "\n");
- fprintf(stderr, "Notes: Assuming error independency and diploid individuals.\n\n");
+ fprintf(stderr, "Notes: Assuming diploid individuals.\n\n");
return 1;
}
mpileup(&mplp, argc - optind, argv + optind);