+++ /dev/null
-#include <math.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include "bam_mcns.h"
-
-#define MC_MIN_QUAL 13
-#define MC_AVG_ERR 0.007
-#define MC_MAX_SUMQ 3000
-#define MC_MAX_SUMQP 1e-300
-#define MC_MAX_EM_ITER 16
-#define MC_EM_EPS 1e-4
-
-struct __mc_aux_t {
- int n, M;
- int ref, alt, alt2;
- double *q2p, *pdg; // pdg -> P(D|g)
- double *phi, *CMk; // CMk=\binom{M}{k}
- double *z, *zswap; // aux for afs
- double *afs, *afs1; // afs: accumulative AFS; afs1: site posterior distribution
- int *qsum, *bcnt;
-};
-
-void mc_init_prior(mc_aux_t *ma, int type, double theta)
-{
- int i;
- if (type == MC_PTYPE_COND2) {
- for (i = 0; i <= 2 * ma->n; ++i)
- ma->phi[i] = 2. * (i + 1) / (2 * ma->n + 1) / (2 * ma->n + 2);
- } else if (type == MC_PTYPE_FLAT) {
- for (i = 0; i <= ma->M; ++i)
- ma->phi[i] = 1. / (ma->M + 1);
- } else {
- double sum;
- for (i = 0, sum = 0.; i < 2 * ma->n; ++i)
- sum += (ma->phi[i] = theta / (2 * ma->n - i));
- ma->phi[2 * ma->n] = 1. - sum;
- }
-}
-
-mc_aux_t *mc_init(int n) // FIXME: assuming diploid
-{
- mc_aux_t *ma;
- int i;
- ma = calloc(1, sizeof(mc_aux_t));
- ma->n = n; ma->M = 2 * n;
- ma->q2p = calloc(MC_MAX_SUMQ + 1, sizeof(double));
- ma->qsum = calloc(4 * ma->n, sizeof(int));
- ma->bcnt = calloc(4 * ma->n, sizeof(int));
- ma->pdg = calloc(3 * ma->n, sizeof(double));
- ma->phi = calloc(ma->M + 1, sizeof(double));
- ma->CMk = calloc(ma->M + 1, sizeof(double));
- ma->z = calloc(2 * ma->n + 1, sizeof(double));
- ma->zswap = calloc(2 * ma->n + 1, sizeof(double));
- ma->afs = calloc(2 * ma->n + 1, sizeof(double));
- ma->afs1 = calloc(2 * ma->n + 1, sizeof(double));
- for (i = 0; i <= MC_MAX_SUMQ; ++i)
- ma->q2p[i] = pow(10., -i / 10.);
- for (i = 0; i <= ma->M; ++i)
- ma->CMk[i] = exp(lgamma(ma->M + 1) - lgamma(i + 1) - lgamma(ma->M - i + 1));
- mc_init_prior(ma, MC_PTYPE_FULL, 1e-3); // the simplest prior
- return ma;
-}
-
-void mc_destroy(mc_aux_t *ma)
-{
- if (ma) {
- free(ma->qsum); free(ma->bcnt);
- free(ma->q2p); free(ma->pdg);
- free(ma->phi); free(ma->CMk);
- free(ma->z); free(ma->zswap);
- free(ma->afs); free(ma->afs1);
- free(ma);
- }
-}
-
-static int sum_err(int *n, const bam_pileup1_t **plp, mc_aux_t *ma)
-{
- int i, j, tot = 0;
- memset(ma->qsum, 0, sizeof(int) * 4 * ma->n);
- memset(ma->bcnt, 0, sizeof(int) * 4 * ma->n);
- for (j = 0; j < ma->n; ++j) {
- int *qsum = ma->qsum + j * 4;
- int *bcnt = ma->bcnt + j * 4;
- for (i = 0; i < n[j]; ++i) {
- const bam_pileup1_t *p = plp[j] + i;
- int q, b;
- if (p->is_del || (p->b->core.flag&BAM_FUNMAP)) continue;
- q = bam1_qual(p->b)[p->qpos];
- if (p->b->core.qual < q) q = p->b->core.qual;
- if (q < MC_MIN_QUAL) continue; // small qual
- b = bam_nt16_nt4_table[(int)bam1_seqi(bam1_seq(p->b), p->qpos)];
- if (b > 3) continue; // N
- qsum[b] += q;
- ++bcnt[b];
- ++tot;
- }
- }
- return tot;
-}
-
-static void set_allele(int ref, mc_aux_t *ma)
-{
- int i, j, sum[4], tmp;
- sum[0] = sum[1] = sum[2] = sum[3] = 0;
- for (i = 0; i < ma->n; ++i)
- for (j = 0; j < 4; ++j)
- sum[j] += ma->qsum[i * 4 + j];
- for (j = 0; j < 4; ++j) sum[j] = sum[j]<<2 | j;
- for (i = 1; i < 4; ++i) // insertion sort
- for (j = i; j > 0 && sum[j] < sum[j-1]; --j)
- tmp = sum[j], sum[j] = sum[j-1], sum[j-1] = tmp;
- ma->ref = sum[3]&3; ma->alt = sum[2]&3; ma->alt2 = -1;
- if (ma->ref != ref) { // the best base is not ref
- if (ref >= 0 && ref <= 3) { // ref is not N
- if (ma->alt == ref) tmp = ma->ref, ma->ref = ma->alt, ma->alt = tmp; // then switch alt and ref
- else ma->alt2 = ma->alt, ma->alt = ma->ref, ma->ref = ref; // then set ref as ref
- } else ma->alt2 = ma->alt, ma->alt = ma->ref, ma->ref = sum[0]&3; // then set the weakest as ref
- }
-}
-
-static void cal_pdg(mc_aux_t *ma)
-{
- int i, j;
- for (j = 0; j < ma->n; ++j) {
- int pi[3], *qsum, *bcnt;
- double *pdg = ma->pdg + j * 3;
- qsum = ma->qsum + j * 4;
- bcnt = ma->bcnt + j * 4;
- pi[1] = 3 * (bcnt[ma->ref] + bcnt[ma->alt]);
- pi[0] = qsum[ma->ref];
- pi[2] = qsum[ma->alt];
- for (i = 0; i < 3; ++i)
- pdg[i] = pi[i] > MC_MAX_SUMQ? MC_MAX_SUMQP : ma->q2p[pi[i]];
- }
-}
-// this calculates the naive allele frequency and Nielsen's frequency
-static double mc_freq0(const mc_aux_t *ma, double *_f)
-{
- int i, cnt;
- double f, f_nielsen, w_sum;
- *_f = -1.;
- for (i = cnt = 0, f = f_nielsen = w_sum = 0.; i < ma->n; ++i) {
- int *bcnt = ma->bcnt + i * 4;
- int x = bcnt[ma->ref] + bcnt[ma->alt];
- if (x) {
- double w, p;
- ++cnt;
- f += (double)bcnt[ma->ref] / x;
- p = (bcnt[ma->ref] - MC_AVG_ERR * x) / (1. - 2. * MC_AVG_ERR) / x;
- w = 2. * x / (1. + x);
- w_sum += w;
- f_nielsen += p * w;
- }
- }
- if (cnt) {
- f_nielsen /= w_sum;
- if (f_nielsen < 0.) f_nielsen = 0.;
- if (f_nielsen > 1.) f_nielsen = 1.;
- *_f = f_nielsen;
- return f / cnt;
- } else return -1.;
-}
-// f0 is the reference allele frequency
-static double mc_freq_iter(double f0, const mc_aux_t *ma)
-{
- double f, f3[3];
- int i;
- f3[0] = (1.-f0)*(1.-f0); f3[1] = 2.*f0*(1.-f0); f3[2] = f0*f0;
- for (i = 0, f = 0.; i < ma->n; ++i) {
- double *pdg;
- pdg = ma->pdg + i * 3;
- f += (pdg[1] * f3[1] + 2. * pdg[2] * f3[2])
- / (pdg[0] * f3[0] + pdg[1] * f3[1] + pdg[2] * f3[2]);
- }
- f /= ma->n * 2.;
- return f;
-}
-
-int mc_call_gt(const mc_aux_t *ma, double f0, int k)
-{
- double sum, g[3];
- double max, f3[3], *pdg = ma->pdg + k * 3;
- int q, i, max_i;
- f3[0] = (1.-f0)*(1.-f0); f3[1] = 2.*f0*(1.-f0); f3[2] = f0*f0;
- for (i = 0, sum = 0.; i < 3; ++i)
- sum += (g[i] = pdg[i] * f3[i]);
- for (i = 0, max = -1., max_i = 0; i < 3; ++i) {
- g[i] /= sum;
- if (g[i] > max) max = g[i], max_i = i;
- }
- max = 1. - max;
- if (max < 1e-308) max = 1e-308;
- q = (int)(-3.434 * log(max) + .499);
- if (q > 99) q = 99;
- return q<<2|max_i;
-}
-
-static void mc_cal_z(mc_aux_t *ma)
-{
- double *z[2], *tmp, *pdg;
- int i, j;
- z[0] = ma->z;
- z[1] = ma->zswap;
- pdg = ma->pdg;
- z[0][0] = 1.; z[0][1] = z[0][2] = 0.;
- for (j = 0; j < ma->n; ++j) {
- int max = (j + 1) * 2;
- double p[3];
- pdg = ma->pdg + j * 3;
- p[0] = pdg[0]; p[1] = 2. * pdg[1]; p[2] = pdg[2];
- z[1][0] = p[0] * z[0][0];
- z[1][1] = p[0] * z[0][1] + p[1] * z[0][0];
- for (i = 2; i <= max; ++i)
- z[1][i] = p[0] * z[0][i] + p[1] * z[0][i-1] + p[2] * z[0][i-2];
- if (j < ma->n - 1) z[1][max+1] = z[1][max+2] = 0.;
-// int k; for (k = 0; k <= max; ++k) printf("%d:%.3lg ", k, z[1][k]); putchar('\n');
- tmp = z[0]; z[0] = z[1]; z[1] = tmp;
- }
- if (z[0] != ma->z) memcpy(ma->z, z[0], sizeof(double) * (2 * ma->n + 1));
-}
-
-static double mc_add_afs(mc_aux_t *ma)
-{
- int k;
- long double sum = 0.;
- memset(ma->afs1, 0, sizeof(double) * (ma->M + 1));
- mc_cal_z(ma);
- for (k = 0, sum = 0.; k <= ma->M; ++k)
- sum += (long double)ma->phi[k] * ma->z[k] / ma->CMk[k];
- for (k = 0; k <= ma->M; ++k) {
- ma->afs1[k] = ma->phi[k] * ma->z[k] / ma->CMk[k] / sum;
- if (isnan(ma->afs1[k]) || isinf(ma->afs1[k])) return -1.;
- }
- for (k = 0, sum = 0.; k <= ma->M; ++k) {
- ma->afs[k] += ma->afs1[k];
- sum += k * ma->afs1[k];
- }
- return sum / ma->M;
-}
-
-int mc_cal(int ref, int *n, const bam_pileup1_t **plp, mc_aux_t *ma, mc_rst_t *rst, int level)
-{
- int i, tot;
- memset(rst, 0, sizeof(mc_rst_t));
- rst->f_em = rst->f_exp = -1.; rst->ref = rst->alt = -1;
- // precalculation
- tot = sum_err(n, plp, ma);
- if (tot == 0) return 0; // no good bases
- set_allele(ref, ma);
- cal_pdg(ma);
- // set ref/major allele
- rst->ref = ma->ref; rst->alt = ma->alt; rst->alt2 = ma->alt2;
- // calculate naive and Nielsen's freq
- rst->f_naive = mc_freq0(ma, &rst->f_nielsen);
- { // calculate f_em
- double flast = rst->f_naive;
- for (i = 0; i < MC_MAX_EM_ITER; ++i) {
- rst->f_em = mc_freq_iter(flast, ma);
- if (fabs(rst->f_em - flast) < MC_EM_EPS) break;
- flast = rst->f_em;
- }
- }
- if (level >= 2) {
- rst->f_exp = mc_add_afs(ma);
- rst->p_ref = ma->afs1[ma->M];
- }
- return tot;
-}
-
-void mc_dump_afs(mc_aux_t *ma)
-{
- int k;
- fprintf(stderr, "[afs]");
- for (k = 0; k <= ma->M; ++k)
- fprintf(stderr, " %d:%.3lf", k, ma->afs[ma->M - k]);
- fprintf(stderr, "\n");
- memset(ma->afs, 0, sizeof(double) * (ma->M + 1));
-}
#include "sam.h"
#include "faidx.h"
#include "bam_maqcns.h"
-#include "bam_mcns.h"
#include "bam2bcf.h"
#include "khash.h"
#include "glf.h"
* mpileup *
***********/
-#define MPLP_VCF 0x1
-#define MPLP_VAR 0x2
-#define MPLP_AFALL 0x8
#define MPLP_GLF 0x10
#define MPLP_NO_COMP 0x20
-#define MPLP_AFS_BLOCK 0x10000
-
typedef struct {
- int max_mq, min_mq, prior_type, flag, min_baseQ;
+ int max_mq, min_mq, flag, min_baseQ;
double theta;
char *reg, *fn_pos;
faidx_t *fai;
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;
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);
}
ref_tid = -1; ref = 0;
iter = bam_mplp_init(n, mplp_func, (void**)data);
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) {
}
}
bcf_close(bp);
- 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)
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) {
+ while ((c = getopt(argc, argv, "gf:r:l:M:q:t:Q:u")) >= 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 '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, "\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);
projects / samtools.git / commitdiff