+#include <math.h>
#include <stdio.h>
#include <unistd.h>
#include <ctype.h>
-#include "bam.h"
+#include "sam.h"
#include "faidx.h"
#include "bam_maqcns.h"
+#include "bam_mcns.h"
#include "khash.h"
#include "glf.h"
-KHASH_SET_INIT_INT64(64)
+#include "kstring.h"
+
+typedef int *indel_list_t;
+KHASH_MAP_INIT_INT64(64, indel_list_t)
#define BAM_PLF_SIMPLE 0x01
#define BAM_PLF_CNS 0x02
#define BAM_PLF_INDEL_ONLY 0x04
#define BAM_PLF_GLF 0x08
+#define BAM_PLF_VAR_ONLY 0x10
+#define BAM_PLF_2ND 0x20
+#define BAM_PLF_RANBASE 0x40
+#define BAM_PLF_1STBASE 0x80
+#define BAM_PLF_ALLBASE 0x100
+#define BAM_PLF_READPOS 0x200
typedef struct {
bam_header_t *h;
faidx_t *fai;
khash_t(64) *hash;
uint32_t format;
- int tid, len;
+ int tid, len, last_pos;
int mask;
+ int max_depth; // for indel calling, ignore reads with the depth too high. 0 for unlimited
char *ref;
- glfFile fp; // for glf output only
+ glfFile fp_glf; // for glf output only
} pu_data_t;
-char **bam_load_pos(const char *fn, int *_n);
+char **__bam_get_lines(const char *fn, int *_n);
void bam_init_header_hash(bam_header_t *header);
int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
{
- int n, tmp, i;
- char **list, *s;
- uint64_t x;
+ char **list;
+ int i, j, n, *fields, max_fields;
khash_t(64) *hash;
bam_init_header_hash(h);
- list = bam_load_pos(fn, &n);
+ list = __bam_get_lines(fn, &n);
hash = kh_init(64);
+ max_fields = 0; fields = 0;
for (i = 0; i < n; ++i) {
- x = (uint64_t)bam_get_tid(h, list[i]) << 32;
- s = list[i];
- while (*s++);
- x |= *((uint32_t*)s) - 1;
- kh_put(64, hash, x, &tmp);
- free(list[i]);
- }
- free(list);
+ char *str = list[i];
+ int chr, n_fields, ret;
+ khint_t k;
+ uint64_t x;
+ n_fields = ksplit_core(str, 0, &max_fields, &fields);
+ if (n_fields < 2) continue;
+ chr = bam_get_tid(h, str + fields[0]);
+ if (chr < 0) {
+ fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
+ continue;
+ }
+ x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
+ k = kh_put(64, hash, x, &ret);
+ if (ret == 0) {
+ fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
+ continue;
+ }
+ kh_val(hash, k) = 0;
+ if (n_fields > 2) {
+ // count
+ for (j = 2; j < n_fields; ++j) {
+ char *s = str + fields[j];
+ if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
+ }
+ if (j > 2) { // update kh_val()
+ int *q, y, z;
+ q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
+ q[0] = j - 2; z = j; y = 1;
+ for (j = 2; j < z; ++j)
+ q[y++] = atoi(str + fields[j]);
+ }
+ }
+ free(str);
+ }
+ free(list); free(fields);
return hash;
}
// an analogy to pileup_func() below
-static int glt_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
+static int glt3_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 rb;
+ int rb, *proposed_indels = 0;
glf1_t *g;
- glf2_t g2;
+ glf3_t *g3;
+
if (d->fai == 0) {
- fprintf(stderr, "[glt_func] reference sequence is required for generating GLT. Abort!\n");
+ fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
exit(1);
}
- if (d->hash && kh_get(64, d->hash, (uint64_t)tid<<32|pos) == kh_end(d->hash)) return 0;
+ if (d->hash) { // only output a list of sites
+ 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);
+ }
+ g3 = glf3_init1();
if (d->fai && (int)tid != d->tid) {
if (d->ref) { // then write the end mark
- memset(&g2, 0, sizeof(glf2_t));
- g2.type = GLF_TYPE_END;
- bgzf_write(d->fp, &g2, sizeof(glf2_t));
+ g3->rtype = GLF3_RTYPE_END;
+ glf3_write1(d->fp_glf, g3);
}
- glf_ref_write(d->fp, d->h->target_name[tid]); // write reference
+ glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
free(d->ref);
d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
d->tid = tid;
+ d->last_pos = 0;
}
rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
- memcpy(&g2, g, sizeof(glf1_t));
- g2.type = GLF_TYPE_NORMAL;
- g2.pos = pos;
- bgzf_write(d->fp, &g2, sizeof(glf2_t));
- r = bam_maqindel(n, pos, d->ido, pu, d->ref);
+ memcpy(g3, g, sizeof(glf1_t));
+ g3->rtype = GLF3_RTYPE_SUB;
+ g3->offset = pos - d->last_pos;
+ d->last_pos = pos;
+ glf3_write1(d->fp_glf, g3);
+ if (pos < d->len) {
+ int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
+ if (proposed_indels)
+ 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);
+ }
if (r) { // then write indel line
- int g3 = 3 * n, min;
- min = g3;
+ int het = 3 * n, min;
+ min = het;
if (min > r->gl[0]) min = r->gl[0];
if (min > r->gl[1]) min = r->gl[1];
- g3 -= min;
- g2.ref_base = 0;
- g2.type = GLF_TYPE_INDEL;
- memset(g2.lk, 0, 10);
- g2.lk[0] = r->gl[0] < 255? r->gl[0] : 255;
- g2.lk[1] = r->gl[1] < 255? r->gl[1] : 255;
- g2.lk[2] = g3 < 255? g3 : 255;
- *(int16_t*)(g2.lk + 3) = (int16_t)r->indel1;
- *(int16_t*)(g2.lk + 5) = (int16_t)r->indel2;
- g2.min_lk = min < 255? min : 255;
- bgzf_write(d->fp, &g2, sizeof(glf2_t));
- if (r->indel1) bgzf_write(d->fp, r->s[0]+1, r->indel1>0? r->indel1 : -r->indel1);
- if (r->indel2) bgzf_write(d->fp, r->s[1]+1, r->indel2>0? r->indel2 : -r->indel2);
+ g3->ref_base = 0;
+ g3->rtype = GLF3_RTYPE_INDEL;
+ memset(g3->lk, 0, 10);
+ g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
+ g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
+ g3->lk[2] = het - min < 255? het - min : 255;
+ g3->offset = 0;
+ g3->indel_len[0] = r->indel1;
+ g3->indel_len[1] = r->indel2;
+ g3->min_lk = min < 255? min : 255;
+ g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
+ g3->indel_seq[0] = strdup(r->s[0]+1);
+ g3->indel_seq[1] = strdup(r->s[1]+1);
+ glf3_write1(d->fp_glf, g3);
bam_maqindel_ret_destroy(r);
}
free(g);
+ glf3_destroy1(g3);
return 0;
}
+static void pileup_seq(const bam_pileup1_t *p, int pos, int ref_len, const char *ref)
+{
+ if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
+ if (!p->is_del) {
+ int j, rb, c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
+ rb = (ref && pos < ref_len)? ref[pos] : 'N';
+ if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
+ else c = bam1_strand(p->b)? tolower(c) : toupper(c);
+ putchar(c);
+ if (p->indel > 0) {
+ printf("+%d", p->indel);
+ for (j = 1; j <= p->indel; ++j) {
+ c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
+ putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
+ }
+ } else if (p->indel < 0) {
+ printf("%d", p->indel);
+ for (j = 1; j <= -p->indel; ++j) {
+ c = (ref && (int)pos+j < ref_len)? ref[pos+j] : 'N';
+ putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
+ }
+ }
+ } else putchar('*');
+ if (p->is_tail) putchar('$');
+}
+
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, j, rb;
- uint32_t x;
- if (d->hash && kh_get(64, d->hash, (uint64_t)tid<<32|pos) == kh_end(d->hash)) return 0;
- if (d->format & BAM_PLF_GLF) return glt_func(tid, pos, n, pu, data);
+ 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 = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
+ 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 cns = bam_maqcns_call(n, pu, d->c);
+ }
+ 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 != 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);
- if (d->format & BAM_PLF_CNS) { // consensus
+ // print consensus information if required
+ if (d->format & BAM_PLF_CNS) {
int ref_q, rb4 = bam_nt16_table[rb];
- x = bam_maqcns_call(n, pu, d->c);
ref_q = 0;
- if (rb4 != 15 && x>>28 != 15 && x>>28 != rb4) { // a SNP
- ref_q = ((x>>24&0xf) == rb4)? x>>8&0xff : (x>>8&0xff) + (x&0xff);
+ 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;
}
- printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[x>>28], x>>8&0xff, ref_q, x>>16&0xff);
- }
- if (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) {
- if (d->ref) // indel calling
- r = bam_maqindel(n, pos, d->ido, pu, d->ref);
+ 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);
}
- // pileup strings
+ // 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
for (i = 0; i < n; ++i) {
const bam_pileup1_t *p = pu + i;
- if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
- if (!p->is_del) {
- int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
- if (toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
- else bam1_strand(p->b)? tolower(c) : toupper(c);
- putchar(c);
- if (p->indel > 0) {
- printf("+%d", p->indel);
- for (j = 1; j <= p->indel; ++j) {
- c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
- putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
- }
- } else if (p->indel < 0) {
- printf("%d", p->indel);
- for (j = 1; j <= -p->indel; ++j) {
- c = (d->ref && (int)pos+j < d->len)? d->ref[pos+j] : 'N';
- putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
- }
- }
- } else putchar('*');
- if (p->is_tail) putchar('$');
+ int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
+ rms_aux += tmp * tmp;
+ pileup_seq(p, pos, d->len, d->ref);
}
+ // finalize rms_mapq
+ 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) {
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;
+ printf("%d,", x < l/2? x+1 : -((l-1)-x+1));
+ }
+ }
putchar('\n');
- if (r) { // then print indel line
+ // 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%d\n", r->cnt1, r->cnt2, r->cnt_ambi, r->cnt_anti);
+ 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;
+ 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->c = bam_maqcns_init();
+ d->c->is_soap = 1; // change the default model
d->ido = bam_maqindel_opt_init();
- while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:")) >= 0) {
+ while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:PA")) >= 0) {
switch (c) {
+ case 'a': d->c->is_soap = 1; break;
+ case 'A': d->c->is_soap = 0; 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 = 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 'm': d->mask = atoi(optarg); 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(stderr, "[bam_pileup] unrecognized -R\n");
+ break;
default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
}
}
+ if (fn_list) is_SAM = 1;
if (optind == argc) {
fprintf(stderr, "\n");
- fprintf(stderr, "Usage: bamtk pileup [options] <in.bam>|<in.sam>\n\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, " -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, " -t FILE list of reference sequences (assume the input is in SAM)\n");
+ 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, " -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 maq consensus sequence\n");
- fprintf(stderr, " -g output in the extended GLT format (suppressing -c/-i/-s)\n");
+ fprintf(stderr, " -c output the SOAPsnp 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);
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);
- free(fn_fa);
- bam_maqcns_prepare(d->c);
- if (d->format & BAM_PLF_GLF) {
- glf_header_t *h;
- h = glf_header_init();
- d->fp = bgzf_fdopen(fileno(stdout), "w");
- glf_header_write(d->fp, h);
- glf_header_destroy(h);
- }
- if (fn_list) {
- tamFile fp;
- bam1_t *b;
- int ret;
- bam_plbuf_t *buf = bam_plbuf_init(pileup_func, d);
- bam_plbuf_set_mask(buf, d->mask);
- d->h = sam_header_read2(fn_list);
- if (fn_pos) d->hash = load_pos(fn_pos, d->h);
- fp = sam_open(argv[optind]);
- b = (bam1_t*)calloc(1, sizeof(bam1_t));
- while ((ret = sam_read1(fp, d->h, b)) >= 0)
- bam_plbuf_push(b, buf);
- bam_plbuf_push(0, buf);
- bam_plbuf_destroy(buf);
- bam_destroy1(b);
- sam_close(fp);
- } else {
- bamFile fp;
- fp = (strcmp(argv[optind], "-") == 0)? bam_dopen(fileno(stdin), "r") : bam_open(argv[optind], "r");
- d->h = bam_header_read(fp);
+ 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(stderr, "[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(stderr, "[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);
- bam_pileup_file(fp, d->mask, pileup_func, d);
- bam_close(fp);
+ sampileup(fp, d->mask, pileup_func, d);
+ 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);
}
- if (d->format & BAM_PLF_GLF) bgzf_close(d->fp);
- free(fn_pos); free(fn_list);
- kh_destroy(64, d->hash);
- bam_header_destroy(d->h);
+ 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 *
+ ***********/
+
+#define MPLP_VCF 0x1
+#define MPLP_VAR 0x2
+#define MPLP_AFALL 0x8
+
+#define MPLP_AFS_BLOCK 0x10000
+
+typedef struct {
+ int max_mq, min_mq, prior_type, flag;
+ double theta;
+ 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;
+ uint64_t N = 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->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_VCF) {
+ ma = mc_init(n);
+ mc_init_prior(ma, conf->prior_type, conf->theta);
+ }
+ 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->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) {
+ 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 (conf->flag&MPLP_VCF) mc_dump_afs(ma);
+ 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;
+ mplp.prior_type = MC_PTYPE_FULL;
+ mplp.theta = 1e-3;
+ while ((c = getopt(argc, argv, "vVcFSP:f:r:l:VM:q:t:")) >= 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;
+ break;
+ case 'r': mplp.reg = strdup(optarg); break;
+ case 'l': mplp.fn_pos = strdup(optarg); 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 'M': mplp.max_mq = atoi(optarg); break;
+ case 'q': mplp.min_mq = atoi(optarg); break;
+ }
+ }
+ if (argc == 1) {
+ 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 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, " -c generate VCF output (consensus calling)\n");
+ fprintf(stderr, " -v show variant sites only\n");
+ fprintf(stderr, "\n");
+ fprintf(stderr, "Notes: Assuming error independency and diploid individuals.\n\n");
+ return 1;
+ }
+ mpileup(&mplp, argc - optind, argv + optind);
+ free(mplp.reg);
+ if (mplp.fai) fai_destroy(mplp.fai);
+ return 0;
+}
projects / samtools.git / blobdiff