#include "kstring.h"
#include "time.h"
+#ifdef _WIN32
+#define srand48(x) srand(x)
+#define lrand48() rand()
+#endif
+
#include "kseq.h"
KSTREAM_INIT(gzFile, gzread, 16384)
#define VC_NO_INDEL 8192
#define VC_ANNO_MAX 16384
#define VC_FIX_PL 32768
+#define VC_EM 0x10000
+#define VC_PAIRCALL 0x20000
+#define VC_QCNT 0x40000
typedef struct {
int flag, prior_type, n1, n_sub, *sublist, n_perm;
+ uint32_t *trio_aux;
char *prior_file, **subsam, *fn_dict;
uint8_t *ploidy;
- double theta, pref, indel_frac, min_perm_p, min_smpl_frac;
+ double theta, pref, indel_frac, min_perm_p, min_smpl_frac, min_lrt;
void *bed;
} viewconf_t;
void bed_destroy(void *_h);
int bed_overlap(const void *_h, const char *chr, int beg, int end);
-static double test_hwe(const double g[3])
-{
- extern double kf_gammaq(double p, double x);
- double fexp, chi2, f[3], n;
- int i;
- n = g[0] + g[1] + g[2];
- fexp = (2. * g[2] + g[1]) / (2. * n);
- if (fexp > 1. - 1e-10) fexp = 1. - 1e-10;
- if (fexp < 1e-10) fexp = 1e-10;
- f[0] = n * (1. - fexp) * (1. - fexp);
- f[1] = n * 2. * fexp * (1. - fexp);
- f[2] = n * fexp * fexp;
- for (i = 0, chi2 = 0.; i < 3; ++i)
- chi2 += (g[i] - f[i]) * (g[i] - f[i]) / f[i];
- return kf_gammaq(.5, chi2 / 2.);
-}
-
typedef struct {
double p[4];
int mq, depth, is_tested, d[4];
bcf_sync(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(bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p1rst_t *pr, double pref, int flag, double em[10], int cons_llr, int64_t cons_gt)
{
kstring_t s;
- int has_I16, is_var = (pr->p_ref < pref);
- double fq, r = is_var? pr->p_ref : pr->p_var;
+ int has_I16, is_var;
+ double fq, r;
anno16_t a;
has_I16 = test16(b, &a) >= 0? 1 : 0;
- rm_info(b, "I16=");
+ rm_info(b, "I16="); // FIXME: probably this function has a bug. If I move it below, I16 will not be removed!
memset(&s, 0, sizeof(kstring_t));
kputc('\0', &s); kputs(b->ref, &s); kputc('\0', &s);
kputs(b->alt, &s); kputc('\0', &s); kputc('\0', &s);
kputs(b->info, &s);
if (b->info[0]) kputc(';', &s);
-// ksprintf(&s, "AF1=%.4lg;AFE=%.4lg;CI95=%.4lg,%.4lg", 1.-pr->f_em, 1.-pr->f_exp, pr->cil, pr->cih);
- ksprintf(&s, "AF1=%.4g;CI95=%.4g,%.4g;G3=%.4g,%.4g,%.4g", 1.-pr->f_em, pr->cil, pr->cih, pr->g[2], pr->g[1], pr->g[0]);
- if (n_smpl > 5) {
- double hwe = test_hwe(pr->g);
- if (hwe < 0.1) ksprintf(&s, ";HWE=%.4g", hwe);
+ { // print EM
+ if (em[0] >= 0) ksprintf(&s, "AF1=%.4g", 1 - em[0]);
+ if (em[4] >= 0 && em[4] <= 0.05) ksprintf(&s, ";G3=%.4g,%.4g,%.4g;HWE=%.3g", em[3], em[2], em[1], em[4]);
+ if (em[5] >= 0 && em[6] >= 0) ksprintf(&s, ";AF2=%.4g,%.4g", 1 - em[5], 1 - em[6]);
+ if (em[7] >= 0) ksprintf(&s, ";LRT=%.3g", em[7]);
+ if (em[8] >= 0) ksprintf(&s, ";LRT2=%.3g", em[8]);
+ }
+ if (cons_llr > 0) {
+ ksprintf(&s, ";CLR=%d", cons_llr);
+ if (cons_gt > 0)
+ ksprintf(&s, ";UGT=%c%c%c;CGT=%c%c%c", cons_gt&0xff, cons_gt>>8&0xff, cons_gt>>16&0xff,
+ cons_gt>>32&0xff, cons_gt>>40&0xff, cons_gt>>48&0xff);
+ }
+ if (pr == 0) { // if pr is unset, return
+ kputc('\0', &s); kputs(b->fmt, &s); kputc('\0', &s);
+ free(b->str);
+ b->m_str = s.m; b->l_str = s.l; b->str = s.s;
+ bcf_sync(b);
+ return 1;
}
+
+ is_var = (pr->p_ref < pref);
+ r = is_var? pr->p_ref : pr->p_var;
+
+// ksprintf(&s, ";CI95=%.4g,%.4g", pr->cil, pr->cih); // FIXME: when EM is not used, ";" should be omitted!
+ ksprintf(&s, ";AC1=%d", pr->ac);
if (has_I16) ksprintf(&s, ";DP4=%d,%d,%d,%d;MQ=%d", a.d[0], a.d[1], a.d[2], a.d[3], a.mq);
fq = pr->p_ref_folded < 0.5? -4.343 * log(pr->p_ref_folded) : 4.343 * log(pr->p_var_folded);
if (fq < -999) fq = -999;
if (fq > 999) fq = 999;
ksprintf(&s, ";FQ=%.3g", fq);
if (pr->cmp[0] >= 0.) { // two sample groups
- int i, q[3], pq;
+ int i, q[3];
for (i = 1; i < 3; ++i) {
double x = pr->cmp[i] + pr->cmp[0]/2.;
q[i] = x == 0? 255 : (int)(-4.343 * log(x) + .499);
if (q[i] > 255) q[i] = 255;
}
- pq = (int)(-4.343 * log(pr->p_chi2) + .499);
if (pr->perm_rank >= 0) ksprintf(&s, ";PR=%d", pr->perm_rank);
- ksprintf(&s, ";QCHI2=%d;PCHI2=%.3g;PC2=%d,%d", pq, q[1], q[2], pr->p_chi2);
- ksprintf(&s, ";AF2=%.4g,%.4g", 1.-pr->f_em2[0], 1.-pr->f_em2[1]);
-// ksprintf(&s, ",%g,%g,%g", pr->cmp[0], pr->cmp[1], pr->cmp[2]);
+ // ksprintf(&s, ";LRT3=%.3g", pr->lrt);
+ ksprintf(&s, ";PCHI2=%.3g;PC2=%d,%d", q[1], q[2], pr->p_chi2);
}
if (has_I16 && a.is_tested) ksprintf(&s, ";PV4=%.2g,%.2g,%.2g,%.2g", a.p[0], a.p[1], a.p[2], a.p[3]);
kputc('\0', &s);
b->m_str = s.m; b->l_str = s.l; b->str = s.s;
bcf_sync(b);
for (i = 0; i < b->n_smpl; ++i) {
- x = bcf_p1_call_gt(pa, pr->f_em, i);
+ x = bcf_p1_call_gt(pa, pr->f_exp, i);
((uint8_t*)b->gi[old_n_gi].data)[i] = (x&3) == 0? 1<<3|1 : (x&3) == 1? 1 : 0;
((uint8_t*)b->gi[old_n_gi+1].data)[i] = x>>2;
}
if (!strstr(str.s, "##INFO=<ID=FQ,"))
kputs("##INFO=<ID=FQ,Number=1,Type=Float,Description=\"Phred probability of all samples being the same\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=AF1,"))
- kputs("##INFO=<ID=AF1,Number=1,Type=Float,Description=\"Max-likelihood estimate of the site allele frequency of the first ALT allele\">\n", &str);
+ kputs("##INFO=<ID=AF1,Number=1,Type=Float,Description=\"Max-likelihood estimate of the first ALT allele frequency (assuming HWE)\">\n", &str);
+ if (!strstr(str.s, "##INFO=<ID=AC1,"))
+ kputs("##INFO=<ID=AC1,Number=1,Type=Float,Description=\"Max-likelihood estimate of the first ALT allele count (no HWE assumption)\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=G3,"))
kputs("##INFO=<ID=G3,Number=3,Type=Float,Description=\"ML estimate of genotype frequencies\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=HWE,"))
kputs("##INFO=<ID=HWE,Number=1,Type=Float,Description=\"Chi^2 based HWE test P-value based on G3\">\n", &str);
- if (!strstr(str.s, "##INFO=<ID=CI95,"))
- kputs("##INFO=<ID=CI95,Number=2,Type=Float,Description=\"Equal-tail Bayesian credible interval of the site allele frequency at the 95% level\">\n", &str);
+ if (!strstr(str.s, "##INFO=<ID=CLR,"))
+ kputs("##INFO=<ID=CLR,Number=1,Type=Integer,Description=\"Log ratio of genotype likelihoods with and without the constraint\">\n", &str);
+ if (!strstr(str.s, "##INFO=<ID=UGT,"))
+ kputs("##INFO=<ID=UGT,Number=1,Type=String,Description=\"The most probable unconstrained genotype configuration in the trio\">\n", &str);
+ if (!strstr(str.s, "##INFO=<ID=CGT,"))
+ kputs("##INFO=<ID=CGT,Number=1,Type=String,Description=\"The most probable constrained genotype configuration in the trio\">\n", &str);
+// if (!strstr(str.s, "##INFO=<ID=CI95,"))
+// kputs("##INFO=<ID=CI95,Number=2,Type=Float,Description=\"Equal-tail Bayesian credible interval of the site allele frequency at the 95% level\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=PV4,"))
kputs("##INFO=<ID=PV4,Number=4,Type=Float,Description=\"P-values for strand bias, baseQ bias, mapQ bias and tail distance bias\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=INDEL,"))
kputs("##INFO=<ID=QCHI2,Number=1,Type=Integer,Description=\"Phred scaled PCHI2.\">\n", &str);
if (!strstr(str.s, "##INFO=<ID=RP,"))
kputs("##INFO=<ID=PR,Number=1,Type=Integer,Description=\"# permutations yielding a smaller PCHI2.\">\n", &str);
+ if (!strstr(str.s, "##INFO=<ID=VDB,"))
+ kputs("##INFO=<ID=VDB,Number=1,Type=Float,Description=\"Variant Distance Bias\">\n", &str);
if (!strstr(str.s, "##FORMAT=<ID=GT,"))
kputs("##FORMAT=<ID=GT,Number=1,Type=String,Description=\"Genotype\">\n", &str);
if (!strstr(str.s, "##FORMAT=<ID=GQ,"))
if (!strstr(str.s, "##FORMAT=<ID=SP,"))
kputs("##FORMAT=<ID=SP,Number=1,Type=Integer,Description=\"Phred-scaled strand bias P-value\">\n", &str);
if (!strstr(str.s, "##FORMAT=<ID=PL,"))
- kputs("##FORMAT=<ID=PL,Number=-1,Type=Integer,Description=\"List of Phred-scaled genotype likelihoods, number of values is (#ALT+1)*(#ALT+2)/2\">\n", &str);
+ kputs("##FORMAT=<ID=PL,Number=G,Type=Integer,Description=\"List of Phred-scaled genotype likelihoods\">\n", &str);
h->l_txt = str.l + 1; h->txt = str.s;
}
-double bcf_ld_freq(const bcf1_t *b0, const bcf1_t *b1, double f[4]);
+double bcf_pair_freq(const bcf1_t *b0, const bcf1_t *b1, double f[4]);
int bcfview(int argc, char *argv[])
{
extern int bcf_fix_gt(bcf1_t *b);
extern int bcf_anno_max(bcf1_t *b);
extern int bcf_shuffle(bcf1_t *b, int seed);
+ extern uint32_t *bcf_trio_prep(int is_x, int is_son);
+ extern int bcf_trio_call(uint32_t *prep, const bcf1_t *b, int *llr, int64_t *gt);
+ extern int bcf_pair_call(const bcf1_t *b);
+ extern int bcf_min_diff(const bcf1_t *b);
+
bcf_t *bp, *bout = 0;
bcf1_t *b, *blast;
int c, *seeds = 0;
- uint64_t n_processed = 0;
+ uint64_t n_processed = 0, qcnt[256];
viewconf_t vc;
bcf_p1aux_t *p1 = 0;
bcf_hdr_t *hin, *hout;
tid = begin = end = -1;
memset(&vc, 0, sizeof(viewconf_t));
- vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.5; vc.indel_frac = -1.; vc.n_perm = 0; vc.min_perm_p = 0.01; vc.min_smpl_frac = 0;
- while ((c = getopt(argc, argv, "FN1:l:cHAGvbSuP:t:p:QgLi:IMs:D:U:X:d:")) >= 0) {
+ vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.5; vc.indel_frac = -1.; vc.n_perm = 0; vc.min_perm_p = 0.01; vc.min_smpl_frac = 0; vc.min_lrt = 1;
+ memset(qcnt, 0, 8 * 256);
+ while ((c = getopt(argc, argv, "FN1:l:cC:eHAGvbSuP:t:p:QgLi:IMs:D:U:X:d:T:Y")) >= 0) {
switch (c) {
case '1': vc.n1 = atoi(optarg); break;
case 'l': vc.bed = bed_read(optarg); break;
case 'b': vc.flag |= VC_BCFOUT; break;
case 'S': vc.flag |= VC_VCFIN; break;
case 'c': vc.flag |= VC_CALL; break;
+ case 'e': vc.flag |= VC_EM; break;
case 'v': vc.flag |= VC_VARONLY | VC_CALL; break;
case 'u': vc.flag |= VC_UNCOMP | VC_BCFOUT; break;
case 'g': vc.flag |= VC_CALL_GT | VC_CALL; break;
case 'I': vc.flag |= VC_NO_INDEL; break;
case 'M': vc.flag |= VC_ANNO_MAX; break;
+ case 'Y': vc.flag |= VC_QCNT; break;
case 't': vc.theta = atof(optarg); break;
case 'p': vc.pref = atof(optarg); break;
case 'i': vc.indel_frac = atof(optarg); break;
case 'Q': vc.flag |= VC_QCALL; break;
case 'L': vc.flag |= VC_ADJLD; break;
case 'U': vc.n_perm = atoi(optarg); break;
+ case 'C': vc.min_lrt = atof(optarg); break;
case 'X': vc.min_perm_p = atof(optarg); break;
case 'd': vc.min_smpl_frac = atof(optarg); break;
case 's': vc.subsam = read_samples(optarg, &vc.n_sub);
for (tid = 0; tid < vc.n_sub; ++tid) vc.ploidy[tid] = vc.subsam[tid][strlen(vc.subsam[tid]) + 1];
tid = -1;
break;
+ case 'T':
+ if (strcmp(optarg, "trioauto") == 0) vc.trio_aux = bcf_trio_prep(0, 0);
+ else if (strcmp(optarg, "trioxd") == 0) vc.trio_aux = bcf_trio_prep(1, 0);
+ else if (strcmp(optarg, "trioxs") == 0) vc.trio_aux = bcf_trio_prep(1, 1);
+ else if (strcmp(optarg, "pair") == 0) vc.flag |= VC_PAIRCALL;
+ else {
+ fprintf(stderr, "[%s] Option '-T' can only take value trioauto, trioxd or trioxs.\n", __func__);
+ return 1;
+ }
+ break;
case 'P':
if (strcmp(optarg, "full") == 0) vc.prior_type = MC_PTYPE_FULL;
else if (strcmp(optarg, "cond2") == 0) vc.prior_type = MC_PTYPE_COND2;
fprintf(stderr, " -S input is VCF\n");
fprintf(stderr, " -u uncompressed BCF output (force -b)\n");
fprintf(stderr, "\nConsensus/variant calling options:\n\n");
- fprintf(stderr, " -c SNP calling\n");
+ fprintf(stderr, " -c SNP calling (force -e)\n");
fprintf(stderr, " -d FLOAT skip loci where less than FLOAT fraction of samples covered [0]\n");
+ fprintf(stderr, " -e likelihood based analyses\n");
fprintf(stderr, " -g call genotypes at variant sites (force -c)\n");
fprintf(stderr, " -i FLOAT indel-to-substitution ratio [%.4g]\n", vc.indel_frac);
fprintf(stderr, " -I skip indels\n");
fprintf(stderr, " -p FLOAT variant if P(ref|D)<FLOAT [%.3g]\n", vc.pref);
fprintf(stderr, " -P STR type of prior: full, cond2, flat [full]\n");
fprintf(stderr, " -t FLOAT scaled substitution mutation rate [%.4g]\n", vc.theta);
+ fprintf(stderr, " -T STR constrained calling; STR can be: pair, trioauto, trioxd and trioxs (see manual) [null]\n");
fprintf(stderr, " -v output potential variant sites only (force -c)\n");
fprintf(stderr, "\nContrast calling and association test options:\n\n");
fprintf(stderr, " -1 INT number of group-1 samples [0]\n");
+ fprintf(stderr, " -C FLOAT posterior constrast for LRT<FLOAT and P(ref|D)<0.5 [%g]\n", vc.min_lrt);
fprintf(stderr, " -U INT number of permutations for association testing (effective with -1) [0]\n");
fprintf(stderr, " -X FLOAT only perform permutations for P(chi^2)<FLOAT [%g]\n", vc.min_perm_p);
fprintf(stderr, "\n");
return 1;
}
+ if (vc.flag & VC_CALL) vc.flag |= VC_EM;
if ((vc.flag & VC_VCFIN) && (vc.flag & VC_BCFOUT) && vc.fn_dict == 0) {
fprintf(stderr, "[%s] For VCF->BCF conversion please specify the sequence dictionary with -D\n", __func__);
return 1;
return 1;
}
} else bcf_p1_init_prior(p1, vc.prior_type, vc.theta);
- if (vc.n1 > 0) {
+ if (vc.n1 > 0 && vc.min_lrt > 0.) { // set n1
bcf_p1_set_n1(p1, vc.n1);
bcf_p1_init_subprior(p1, vc.prior_type, vc.theta);
}
}
}
while (vcf_read(bp, hin, b) > 0) {
- int is_indel;
+ int is_indel, cons_llr = -1;
+ int64_t cons_gt = -1;
+ double em[10];
if ((vc.flag & VC_VARONLY) && strcmp(b->alt, "X") == 0) continue;
if ((vc.flag & VC_VARONLY) && vc.min_smpl_frac > 0.) {
extern int bcf_smpl_covered(const bcf1_t *b);
if (!(l > begin && end > b->pos)) continue;
}
++n_processed;
+ if ((vc.flag & VC_QCNT) && !is_indel) { // summarize the difference
+ int x = bcf_min_diff(b);
+ if (x > 255) x = 255;
+ if (x >= 0) ++qcnt[x];
+ }
if (vc.flag & VC_QCALL) { // output QCALL format; STOP here
bcf_2qcall(hout, b);
continue;
}
- if (vc.flag & (VC_CALL|VC_ADJLD)) bcf_gl2pl(b);
+ if (vc.trio_aux) // do trio calling
+ bcf_trio_call(vc.trio_aux, b, &cons_llr, &cons_gt);
+ else if (vc.flag & VC_PAIRCALL)
+ cons_llr = bcf_pair_call(b);
+ if (vc.flag & (VC_CALL|VC_ADJLD|VC_EM)) bcf_gl2pl(b);
+ if (vc.flag & VC_EM) bcf_em1(b, vc.n1, 0x1ff, em);
+ else {
+ int i;
+ for (i = 0; i < 9; ++i) em[i] = -1.;
+ }
if (vc.flag & VC_CALL) { // call variants
bcf_p1rst_t pr;
- bcf_p1_cal(b, p1, &pr); // pr.g[3] is not calculated here
+ int calret = bcf_p1_cal(b, (em[7] >= 0 && em[7] < vc.min_lrt), p1, &pr);
if (n_processed % 100000 == 0) {
fprintf(stderr, "[%s] %ld sites processed.\n", __func__, (long)n_processed);
bcf_p1_dump_afs(p1);
bcf_p1rst_t r;
int i, n = 0;
for (i = 0; i < vc.n_perm; ++i) {
+#ifdef BCF_PERM_LRT // LRT based permutation is much faster but less robust to artifacts
+ double x[10];
bcf_shuffle(b, seeds[i]);
- bcf_p1_cal(b, p1, &r);
+ bcf_em1(b, vc.n1, 1<<7, x);
+ if (x[7] < em[7]) ++n;
+#else
+ bcf_shuffle(b, seeds[i]);
+ bcf_p1_cal(b, 1, p1, &r);
if (pr.p_chi2 >= r.p_chi2) ++n;
+#endif
}
pr.perm_rank = n;
}
- update_bcf1(hout->n_smpl, b, p1, &pr, vc.pref, vc.flag);
- }
+ if (calret >= 0) update_bcf1(b, p1, &pr, vc.pref, vc.flag, em, cons_llr, cons_gt);
+ } else if (vc.flag & VC_EM) update_bcf1(b, 0, 0, 0, vc.flag, em, cons_llr, cons_gt);
if (vc.flag & VC_ADJLD) { // compute LD
double f[4], r2;
- if ((r2 = bcf_ld_freq(blast, b, f)) >= 0) {
+ if ((r2 = bcf_pair_freq(blast, b, f)) >= 0) {
kstring_t s;
s.m = s.l = 0; s.s = 0;
if (*b->info) kputc(';', &s);
vcf_close(bp); vcf_close(bout);
if (vc.fn_dict) free(vc.fn_dict);
if (vc.ploidy) free(vc.ploidy);
+ if (vc.trio_aux) free(vc.trio_aux);
if (vc.n_sub) {
int i;
for (i = 0; i < vc.n_sub; ++i) free(vc.subsam[i]);
free(vc.subsam); free(vc.sublist);
}
if (vc.bed) bed_destroy(vc.bed);
+ if (vc.flag & VC_QCNT)
+ for (c = 0; c < 256; ++c)
+ fprintf(stderr, "QT\t%d\t%lld\n", c, (long long)qcnt[c]);
if (seeds) free(seeds);
if (p1) bcf_p1_destroy(p1);
return 0;
projects / samtools.git / blobdiff