X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bcftools%2Fcall1.c;h=9c0b498c18cb46e2dd673f0527f7c02d2e76dc3d;hb=f528e7da717a1a2a6ab2672c7c55f37d315452f3;hp=a5c03f452ed044cea0867812d33d95eb499a74c8;hpb=82e06ed4289c179c7a0da9b8ffa31b15836ae763;p=samtools.git diff --git a/bcftools/call1.c b/bcftools/call1.c index a5c03f4..9c0b498 100644 --- a/bcftools/call1.c +++ b/bcftools/call1.c @@ -13,23 +13,25 @@ KHASH_SET_INIT_INT64(set64) #include "kseq.h" KSTREAM_INIT(gzFile, gzread, 16384) -#define VC_NO_PL 1 #define VC_NO_GENO 2 #define VC_BCFOUT 4 #define VC_CALL 8 #define VC_VARONLY 16 #define VC_VCFIN 32 #define VC_UNCOMP 64 -#define VC_HWE 128 #define VC_KEEPALT 256 #define VC_ACGT_ONLY 512 #define VC_QCALL 1024 #define VC_CALL_GT 2048 +#define VC_ADJLD 4096 +#define VC_NO_INDEL 8192 +#define VC_ANNO_MAX 16384 typedef struct { - int flag, prior_type, n1; - char *fn_list, *prior_file; - double theta, pref; + int flag, prior_type, n1, n_sub, *sublist; + char *fn_list, *prior_file, **subsam, *fn_dict; + uint8_t *ploidy; + double theta, pref, indel_frac; } viewconf_t; khash_t(set64) *bcf_load_pos(const char *fn, bcf_hdr_t *_h) @@ -65,23 +67,6 @@ khash_t(set64) *bcf_load_pos(const char *fn, bcf_hdr_t *_h) return hash; } -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]; @@ -127,7 +112,7 @@ static int test16(bcf1_t *b, anno16_t *a) if ((p = strstr(b->info, "I16=")) == 0) return -1; p += 4; for (i = 0; i < 16; ++i) { - anno[i] = strtol(p, &p, 10); + errno = 0; anno[i] = strtol(p, &p, 10); if (anno[i] == 0 && (errno == EINVAL || errno == ERANGE)) return -2; ++p; } @@ -149,10 +134,9 @@ static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p { kstring_t s; int is_var = (pr->p_ref < pref); - double p_hwe, r = is_var? pr->p_ref : 1. - pr->p_ref; + double r = is_var? pr->p_ref : pr->p_var, fq; anno16_t a; - p_hwe = pr->g[0] >= 0.? test_hwe(pr->g) : 1.0; // only do HWE g[] is calculated test16(b, &a); rm_info(b, "I16="); @@ -161,20 +145,23 @@ static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p kputs(b->alt, &s); kputc('\0', &s); kputc('\0', &s); kputs(b->info, &s); if (b->info[0]) kputc(';', &s); - ksprintf(&s, "AF1=%.3lf;AFE=%.3lf", 1.-pr->f_em, 1.-pr->f_exp); +// 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", 1.-pr->f_em, pr->cil, pr->cih); 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 (a.is_tested) { - if (pr->pc[0] >= 0.) ksprintf(&s, ";PC4=%lg,%lg,%lg,%lg", pr->pc[0], pr->pc[1], pr->pc[2], pr->pc[3]); - ksprintf(&s, ";PV4=%.2lg,%.2lg,%.2lg,%.2lg", a.p[0], a.p[1], a.p[2], a.p[3]); + if (pr->pc[0] >= 0.) ksprintf(&s, ";PC4=%g,%g,%g,%g", pr->pc[0], pr->pc[1], pr->pc[2], pr->pc[3]); + ksprintf(&s, ";PV4=%.2g,%.2g,%.2g,%.2g", a.p[0], a.p[1], a.p[2], a.p[3]); } - 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); b->m_str = s.m; b->l_str = s.l; b->str = s.s; - b->qual = r < 1e-100? 99 : -4.343 * log(r); - if (b->qual > 99) b->qual = 99; + b->qual = r < 1e-100? 999 : -4.343 * log(r); + if (b->qual > 999) b->qual = 999; bcf_sync(b); if (!is_var) bcf_shrink_alt(b, 1); else if (!(flag&VC_KEEPALT)) @@ -194,41 +181,129 @@ static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p return is_var; } +static char **read_samples(const char *fn, int *_n) +{ + gzFile fp; + kstream_t *ks; + kstring_t s; + int dret, n = 0, max = 0; + char **sam = 0; + *_n = 0; + s.l = s.m = 0; s.s = 0; + fp = gzopen(fn, "r"); + if (fp == 0) return 0; // fail to open file + ks = ks_init(fp); + while (ks_getuntil(ks, 0, &s, &dret) >= 0) { + int l; + if (max == n) { + max = max? max<<1 : 4; + sam = realloc(sam, sizeof(void*)*max); + } + l = s.l; + sam[n] = malloc(s.l + 2); + strcpy(sam[n], s.s); + sam[n][l+1] = 2; // by default, diploid + if (dret != '\n') { + if (ks_getuntil(ks, 0, &s, &dret) >= 0) { // read ploidy, 1 or 2 + int x = (int)s.s[0] - '0'; + if (x == 1 || x == 2) sam[n][l+1] = x; + else fprintf(stderr, "(%s) ploidy can only be 1 or 2; assume diploid\n", __func__); + } + if (dret != '\n') ks_getuntil(ks, '\n', &s, &dret); + } + ++n; + } + ks_destroy(ks); + gzclose(fp); + free(s.s); + *_n = n; + return sam; +} + +static void write_header(bcf_hdr_t *h) +{ + kstring_t str; + str.l = h->l_txt? h->l_txt - 1 : 0; + str.m = str.l + 1; str.s = h->txt; + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##INFO=\n", &str); + if (!strstr(str.s, "##FORMAT=\n", &str); + if (!strstr(str.s, "##FORMAT=\n", &str); + if (!strstr(str.s, "##FORMAT=\n", &str); + if (!strstr(str.s, "##FORMAT=\n", &str); + if (!strstr(str.s, "##FORMAT=\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]); + int bcfview(int argc, char *argv[]) { extern int bcf_2qcall(bcf_hdr_t *h, bcf1_t *b); + extern void bcf_p1_indel_prior(bcf_p1aux_t *ma, double x); + extern int bcf_fix_gt(bcf1_t *b); + extern int bcf_anno_max(bcf1_t *b); bcf_t *bp, *bout = 0; - bcf1_t *b; + bcf1_t *b, *blast; int c; uint64_t n_processed = 0; viewconf_t vc; bcf_p1aux_t *p1 = 0; - bcf_hdr_t *h; + bcf_hdr_t *hin, *hout; int tid, begin, end; char moder[4], modew[4]; khash_t(set64) *hash = 0; tid = begin = end = -1; memset(&vc, 0, sizeof(viewconf_t)); - vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.5; - while ((c = getopt(argc, argv, "N1:l:cHAGvLbSuP:t:p:Qg")) >= 0) { + vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.5; vc.indel_frac = -1.; + while ((c = getopt(argc, argv, "N1:l:cHAGvbSuP:t:p:QgLi:IMs:D:")) >= 0) { switch (c) { case '1': vc.n1 = atoi(optarg); break; case 'l': vc.fn_list = strdup(optarg); break; + case 'D': vc.fn_dict = strdup(optarg); break; case 'N': vc.flag |= VC_ACGT_ONLY; break; case 'G': vc.flag |= VC_NO_GENO; break; - case 'L': vc.flag |= VC_NO_PL; break; case 'A': vc.flag |= VC_KEEPALT; break; case 'b': vc.flag |= VC_BCFOUT; break; case 'S': vc.flag |= VC_VCFIN; break; case 'c': vc.flag |= VC_CALL; break; - case 'v': vc.flag |= VC_VARONLY; break; + case 'v': vc.flag |= VC_VARONLY | VC_CALL; break; case 'u': vc.flag |= VC_UNCOMP | VC_BCFOUT; break; - case 'H': vc.flag |= VC_HWE; break; - case 'g': vc.flag |= VC_CALL_GT; 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 '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 's': vc.subsam = read_samples(optarg, &vc.n_sub); + vc.ploidy = calloc(vc.n_sub + 1, 1); + for (tid = 0; tid < vc.n_sub; ++tid) vc.ploidy[tid] = vc.subsam[tid][strlen(vc.subsam[tid]) + 1]; + tid = -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; @@ -241,38 +316,56 @@ int bcfview(int argc, char *argv[]) fprintf(stderr, "\n"); fprintf(stderr, "Usage: bcftools view [options] [reg]\n\n"); fprintf(stderr, "Options: -c SNP calling\n"); + fprintf(stderr, " -v output potential variant sites only (force -c)\n"); + fprintf(stderr, " -g call genotypes at variant sites (force -c)\n"); fprintf(stderr, " -b output BCF instead of VCF\n"); - fprintf(stderr, " -u uncompressed BCF output\n"); + fprintf(stderr, " -u uncompressed BCF output (force -b)\n"); fprintf(stderr, " -S input is VCF\n"); fprintf(stderr, " -A keep all possible alternate alleles at variant sites\n"); fprintf(stderr, " -G suppress all individual genotype information\n"); - fprintf(stderr, " -g call genotypes for variant sites\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, " -N skip sites where REF is not A/C/G/T\n"); fprintf(stderr, " -Q output the QCALL likelihood format\n"); + fprintf(stderr, " -L calculate LD for adjacent sites\n"); + fprintf(stderr, " -I skip indels\n"); + fprintf(stderr, " -D FILE sequence dictionary for VCF->BCF conversion [null]\n"); fprintf(stderr, " -1 INT number of group-1 samples [0]\n"); fprintf(stderr, " -l FILE list of sites to output [all sites]\n"); - fprintf(stderr, " -t FLOAT scaled mutation rate [%.4lg]\n", vc.theta); - fprintf(stderr, " -p FLOAT variant if P(ref|D)BCF conversion please specify the sequence dictionary with -D\n", __func__); + return 1; + } b = calloc(1, sizeof(bcf1_t)); + blast = calloc(1, sizeof(bcf1_t)); strcpy(moder, "r"); if (!(vc.flag & VC_VCFIN)) strcat(moder, "b"); strcpy(modew, "w"); if (vc.flag & VC_BCFOUT) strcat(modew, "b"); if (vc.flag & VC_UNCOMP) strcat(modew, "u"); bp = vcf_open(argv[optind], moder); - h = vcf_hdr_read(bp); + hin = hout = vcf_hdr_read(bp); + if (vc.fn_dict && (vc.flag & VC_VCFIN)) + vcf_dictread(bp, hin, vc.fn_dict); bout = vcf_open("-", modew); - if (!(vc.flag & VC_QCALL)) vcf_hdr_write(bout, h); + if (!(vc.flag & VC_QCALL)) { + if (vc.n_sub) { + vc.sublist = calloc(vc.n_sub, sizeof(int)); + hout = bcf_hdr_subsam(hin, vc.n_sub, vc.subsam, vc.sublist); + } + if (vc.flag & VC_CALL) write_header(hout); + vcf_hdr_write(bout, hout); + } if (vc.flag & VC_CALL) { - p1 = bcf_p1_init(h->n_smpl); + p1 = bcf_p1_init(hout->n_smpl, vc.ploidy); if (vc.prior_file) { if (bcf_p1_read_prior(p1, vc.prior_file) < 0) { fprintf(stderr, "[%s] fail to read the prior AFS.\n", __func__); @@ -283,10 +376,11 @@ int bcfview(int argc, char *argv[]) bcf_p1_set_n1(p1, vc.n1); bcf_p1_init_subprior(p1, vc.prior_type, vc.theta); } + if (vc.indel_frac > 0.) bcf_p1_indel_prior(p1, vc.indel_frac); // otherwise use the default indel_frac } - if (vc.fn_list) hash = bcf_load_pos(vc.fn_list, h); + if (vc.fn_list) hash = bcf_load_pos(vc.fn_list, hin); if (optind + 1 < argc && !(vc.flag&VC_VCFIN)) { - void *str2id = bcf_build_refhash(h); + void *str2id = bcf_build_refhash(hout); if (bcf_parse_region(str2id, argv[optind+1], &tid, &begin, &end) >= 0) { bcf_idx_t *idx; idx = bcf_idx_load(argv[optind]); @@ -302,8 +396,12 @@ int bcfview(int argc, char *argv[]) } } } - while (vcf_read(bp, h, b) > 0) { - if (vc.flag & VC_ACGT_ONLY) { + while (vcf_read(bp, hin, b) > 0) { + int is_indel; + if (vc.n_sub) bcf_subsam(vc.n_sub, vc.sublist, b); + is_indel = bcf_is_indel(b); + if ((vc.flag & VC_NO_INDEL) && is_indel) continue; + if ((vc.flag & VC_ACGT_ONLY) && !is_indel) { int x; if (b->ref[0] == 0 || b->ref[1] != 0) continue; x = toupper(b->ref[0]); @@ -324,35 +422,55 @@ int bcfview(int argc, char *argv[]) } ++n_processed; if (vc.flag & VC_QCALL) { // output QCALL format; STOP here - bcf_2qcall(h, b); + bcf_2qcall(hout, b); continue; } + if (vc.flag & (VC_CALL|VC_ADJLD)) bcf_gl2pl(b); if (vc.flag & VC_CALL) { // call variants bcf_p1rst_t pr; - bcf_gl2pl(b); 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 % 100000 == 0) { fprintf(stderr, "[%s] %ld sites processed.\n", __func__, (long)n_processed); 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(hout->n_smpl, b, p1, &pr, vc.pref, vc.flag); + } + if (vc.flag & VC_ADJLD) { // compute LD + double f[4], r2; + if ((r2 = bcf_ld_freq(blast, b, f)) >= 0) { + kstring_t s; + s.m = s.l = 0; s.s = 0; + if (*b->info) kputc(';', &s); + ksprintf(&s, "NEIR=%.3f;NEIF=%.3f,%.3f", r2, f[0]+f[2], f[0]+f[1]); + bcf_append_info(b, s.s, s.l); + free(s.s); + } + bcf_cpy(blast, b); } + if (vc.flag & VC_ANNO_MAX) bcf_anno_max(b); if (vc.flag & VC_NO_GENO) { // do not output GENO fields b->n_gi = 0; b->fmt[0] = '\0'; - } - vcf_write(bout, h, b); + b->l_str = b->fmt - b->str + 1; + } else bcf_fix_gt(b); + vcf_write(bout, hout, b); } if (vc.prior_file) free(vc.prior_file); if (vc.flag & VC_CALL) bcf_p1_dump_afs(p1); - bcf_hdr_destroy(h); - bcf_destroy(b); + if (hin != hout) bcf_hdr_destroy(hout); + bcf_hdr_destroy(hin); + bcf_destroy(b); bcf_destroy(blast); vcf_close(bp); vcf_close(bout); if (hash) kh_destroy(set64, hash); if (vc.fn_list) free(vc.fn_list); + if (vc.fn_dict) free(vc.fn_dict); + if (vc.ploidy) free(vc.ploidy); + 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 (p1) bcf_p1_destroy(p1); return 0; } -