#define VC_VARONLY 16
#define VC_VCFIN 32
#define VC_UNCOMP 64
+#define VC_HWE 128
typedef struct {
- int flag, prior_type;
+ int flag, prior_type, n1;
char *fn_list, *prior_file;
double theta, pref;
} viewconf_t;
bcf_sync(n_smpl, b);
}
-static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p1rst_t *pr, double pref, int flag)
+static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p1rst_t *pr, double pref)
{
kstring_t s;
int is_var = (pr->p_ref < pref);
double p_hwe, r = is_var? pr->p_ref : 1. - pr->p_ref;
anno16_t a;
- p_hwe = test_hwe(pr->g);
+ p_hwe = pr->g[0] >= 0.? test_hwe(pr->g) : 1.0; // only do HWE g[] is calculated
test16(b, &a);
rm_info(n_smpl, b, "I16=");
ksprintf(&s, "AF1=%.3lf;AFE=%.3lf", 1.-pr->f_em, 1.-pr->f_exp);
ksprintf(&s, ";DP4=%d,%d,%d,%d;MQ=%d", a.d[0], a.d[1], a.d[2], a.d[3], a.mq);
if (a.is_tested) ksprintf(&s, ";PV4=%.2lg,%.2lg,%.2lg,%.2lg", a.p[0], a.p[1], a.p[2], a.p[3]);
- if (p_hwe <= .2) ksprintf(&s, ";GC=%.2lf,%.2lf,%.2lf;HWE=%.3lf", pr->g[2], pr->g[1], pr->g[0], p_hwe);
+ if (pr->g[0] >= 0. && p_hwe <= .2)
+ ksprintf(&s, ";GC=%.2lf,%.2lf,%.2lf;HWE=%.3lf", pr->g[2], pr->g[1], pr->g[0], p_hwe);
kputc('\0', &s);
kputs(b->fmt, &s); kputc('\0', &s);
free(b->str);
tid = begin = end = -1;
memset(&vc, 0, sizeof(viewconf_t));
- vc.prior_type = -1; vc.theta = 1e-3; vc.pref = 0.9;
- while ((c = getopt(argc, argv, "l:cGvLbSuP:t:p:")) >= 0) {
+ vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.9;
+ while ((c = getopt(argc, argv, "1:l:cHGvLbSuP:t:p:")) >= 0) {
switch (c) {
+ case '1': vc.n1 = atoi(optarg); break;
case 'l': vc.fn_list = strdup(optarg); break;
case 'G': vc.flag |= VC_NO_GENO; break;
case 'L': vc.flag |= VC_NO_PL; break;
case 'c': vc.flag |= VC_CALL; break;
case 'v': vc.flag |= VC_VARONLY; break;
case 'u': vc.flag |= VC_UNCOMP | VC_BCFOUT; break;
+ case 'H': vc.flag |= VC_HWE; break;
case 't': vc.theta = atof(optarg); break;
case 'p': vc.pref = atof(optarg); break;
case 'P':
fprintf(stderr, " -S input is VCF\n");
fprintf(stderr, " -G suppress all individual genotype information\n");
fprintf(stderr, " -L discard the PL genotype field\n");
+ fprintf(stderr, " -H perform Hardy-Weinberg test (slower)\n");
fprintf(stderr, " -v output potential variant sites only\n");
fprintf(stderr, " -l FILE list of sites to output [all sites]\n");
fprintf(stderr, " -t FLOAT scaled mutation rate [%.4lg]\n", vc.theta);
return 1;
}
} else bcf_p1_init_prior(p1, vc.prior_type, vc.theta);
+ if (vc.n1 > 0) bcf_p1_set_n1(p1, vc.n1);
}
if (vc.fn_list) hash = bcf_load_pos(vc.fn_list, h);
if (optind + 1 < argc && !(vc.flag&VC_VCFIN)) {
}
if (vc.flag & VC_CALL) {
bcf_p1rst_t pr;
- bcf_p1_cal(b, p1, &pr);
+ bcf_p1_cal(b, p1, &pr); // pr.g[3] is not calculated here
+ if (vc.flag&VC_HWE) bcf_p1_cal_g3(p1, pr.g);
if ((n_processed + 1) % 50000 == 0) bcf_p1_dump_afs(p1);
if (pr.p_ref >= vc.pref && (vc.flag & VC_VARONLY)) continue;
- update_bcf1(h->n_smpl, b, p1, &pr, vc.pref, vc.flag);
+ update_bcf1(h->n_smpl, b, p1, &pr, vc.pref);
}
if (vc.flag & VC_NO_GENO) {
b->n_gi = 0;
double *q2p, *pdg; // pdg -> P(D|g)
double *phi;
double *z, *zswap; // aux for afs
+ double *z1, *z2; // only calculated when n1 is set
+ double t, t1, t2;
double *afs, *afs1; // afs: accumulative AFS; afs1: site posterior distribution
const uint8_t *PL; // point to PL
int PL_len;
ma->phi = 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->z1 = calloc(ma->M + 1, sizeof(double)); // actually we do not need this large
+ ma->z2 = calloc(ma->M + 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 < 256; ++i)
if (ma) {
free(ma->q2p); free(ma->pdg);
free(ma->phi);
- free(ma->z); free(ma->zswap);
+ free(ma->z); free(ma->zswap); free(ma->z1); free(ma->z2);
free(ma->afs); free(ma->afs1);
free(ma);
}
#define TINY 1e-20
-static void mc_cal_y(bcf_p1aux_t *ma)
+static void mc_cal_y_core(bcf_p1aux_t *ma, int beg)
{
double *z[2], *tmp, *pdg;
- int k, j, last_min, last_max;
+ int _j, last_min, last_max;
z[0] = ma->z;
z[1] = ma->zswap;
pdg = ma->pdg;
memset(z[1], 0, sizeof(double) * (ma->M + 1));
z[0][0] = 1.;
last_min = last_max = 0;
- for (j = 0; j < ma->n; ++j) {
- int _min = last_min, _max = last_max;
+ ma->t = 0.;
+ for (_j = beg; _j < ma->n; ++_j) {
+ int k, j = _j - beg, _min = last_min, _max = last_max;
double p[3], sum;
- pdg = ma->pdg + j * 3;
+ pdg = ma->pdg + _j * 3;
p[0] = pdg[0]; p[1] = 2. * pdg[1]; p[2] = pdg[2];
for (; _min < _max && z[0][_min] < TINY; ++_min) z[0][_min] = z[1][_min] = 0.;
for (; _max > _min && z[0][_max] < TINY; --_max) z[0][_max] = z[1][_max] = 0.;
+ k*(2*j+2-k) * p[1] * z[0][k-1]
+ k*(k-1)* p[2] * z[0][k-2];
for (k = _min, sum = 0.; k <= _max; ++k) sum += z[1][k];
+ ma->t += log(sum / ((2. * j + 2) * (2. * j + 1)));
for (k = _min; k <= _max; ++k) z[1][k] /= sum;
if (_min >= 1) z[1][_min-1] = 0.;
if (_min >= 2) z[1][_min-2] = 0.;
if (j < ma->n - 1) z[1][_max+1] = z[1][_max+2] = 0.;
+ if (_j == ma->n1 - 1) { // set pop1
+ ma->t1 = ma->t;
+ memcpy(ma->z1, z[1], sizeof(double) * (ma->n1 * 2 + 1));
+ }
tmp = z[0]; z[0] = z[1]; z[1] = tmp;
last_min = _min; last_max = _max;
}
if (z[0] != ma->z) memcpy(ma->z, z[0], sizeof(double) * (ma->M + 1));
}
+static void mc_cal_y(bcf_p1aux_t *ma)
+{
+ if (ma->n1 > 0 && ma->n1 < ma->n) {
+ int k;
+ double x;
+ memset(ma->z1, 0, sizeof(double) * (2 * ma->n1 + 1));
+ memset(ma->z2, 0, sizeof(double) * (2 * (ma->n - ma->n1) + 1));
+ ma->t1 = ma->t2 = 0.;
+ mc_cal_y_core(ma, ma->n1);
+ ma->t2 = ma->t;
+ memcpy(ma->z2, ma->z, sizeof(double) * (2 * (ma->n - ma->n1) + 1));
+ mc_cal_y_core(ma, 0);
+ // rescale z
+ x = exp(ma->t - (ma->t1 + ma->t2));
+ for (k = 0; k <= ma->M; ++k) ma->z[k] *= x;
+ } else mc_cal_y_core(ma, 0);
+/*
+ if (ma->n1 > 0 && ma->n1 < ma->n) {
+ int i;
+ double y[5];
+ printf("*****\n");
+ for (i = 0; i <= 2; ++i)
+ printf("(%lf,%lf) ", ma->z1[i], ma->z2[i]);
+ printf("\n");
+ y[0] = ma->z1[0] * ma->z2[0];
+ y[1] = 1./2. * (ma->z1[0] * ma->z2[1] + ma->z1[1] * ma->z2[0]);
+ y[2] = 1./6. * (ma->z1[0] * ma->z2[2] + ma->z1[2] * ma->z2[0]) + 4./6. * ma->z1[1] * ma->z2[1];
+ y[3] = 1./2. * (ma->z1[1] * ma->z2[2] + ma->z1[2] * ma->z2[1]);
+ y[4] = ma->z1[2] * ma->z2[2];
+ for (i = 0; i <= 4; ++i) printf("(%lf,%lf) ", ma->z[i], y[i]);
+ printf("\n");
+ }
+*/
+}
+
static double mc_cal_afs(bcf_p1aux_t *ma)
{
int k;
return sum / ma->M;
}
-static long double p1_cal_g3(bcf_p1aux_t *p1a, double g[3])
+long double bcf_p1_cal_g3(bcf_p1aux_t *p1a, double g[3])
{
long double pd = 0., g2[3];
int i, k;
flast = rst->f_em;
}
}
- p1_cal_g3(ma, rst->g);
+ rst->g[0] = rst->g[1] = rst->g[2] = -1.;
+// bcf_p1_cal_g3(ma, rst->g);
return 0;
}
int bcf_p1_call_gt(const bcf_p1aux_t *ma, double f0, int k);
void bcf_p1_dump_afs(bcf_p1aux_t *ma);
int bcf_p1_read_prior(bcf_p1aux_t *ma, const char *fn);
+ long double bcf_p1_cal_g3(bcf_p1aux_t *p1a, double g[3]);
+ int bcf_p1_set_n1(bcf_p1aux_t *b, int n1);
#ifdef __cplusplus
}