X-Git-Url: http://woldlab.caltech.edu/gitweb/?p=samtools.git;a=blobdiff_plain;f=bcftools%2Fcall1.c;h=8e57aa1d6da3c57bd9c5dbae39385816af163773;hp=e074fb5cac2fec5bf2129aec0c2c736764dd0ef5;hb=b7c06ea4740153b8f27c7c2374131dbd607b6113;hpb=fa217aa47313e2535cbd2d4bb034cfd405162662 diff --git a/bcftools/call1.c b/bcftools/call1.c index e074fb5..8e57aa1 100644 --- a/bcftools/call1.c +++ b/bcftools/call1.c @@ -8,9 +8,6 @@ #include "kstring.h" #include "time.h" -#include "khash.h" -KHASH_SET_INIT_INT64(set64) - #include "kseq.h" KSTREAM_INIT(gzFile, gzread, 16384) @@ -28,46 +25,19 @@ KSTREAM_INIT(gzFile, gzread, 16384) #define VC_NO_INDEL 8192 #define VC_ANNO_MAX 16384 #define VC_FIX_PL 32768 +#define VC_EM 0x10000 typedef struct { int flag, prior_type, n1, n_sub, *sublist, n_perm; - char *fn_list, *prior_file, **subsam, *fn_dict; + 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; -khash_t(set64) *bcf_load_pos(const char *fn, bcf_hdr_t *_h) -{ - void *str2id; - gzFile fp; - kstream_t *ks; - int ret, dret, lineno = 1; - kstring_t *str; - khash_t(set64) *hash = 0; - - hash = kh_init(set64); - str2id = bcf_build_refhash(_h); - str = calloc(1, sizeof(kstring_t)); - fp = strcmp(fn, "-")? gzopen(fn, "r") : gzdopen(fileno(stdin), "r"); - ks = ks_init(fp); - while (ks_getuntil(ks, 0, str, &dret) >= 0) { - int tid = bcf_str2id(str2id, str->s); - if (tid >= 0 && dret != '\n') { - if (ks_getuntil(ks, 0, str, &dret) >= 0) { - uint64_t x = (uint64_t)tid<<32 | (atoi(str->s) - 1); - kh_put(set64, hash, x, &ret); - } else break; - } else fprintf(stderr, "[%s] %s is not a reference name (line %d).\n", __func__, str->s, lineno); - if (dret != '\n') while ((dret = ks_getc(ks)) > 0 && dret != '\n'); - if (dret < 0) break; - ++lineno; - } - bcf_str2id_destroy(str2id); - ks_destroy(ks); - gzclose(fp); - free(str->s); free(str); - return hash; -} +void *bed_read(const char *fn); +void bed_destroy(void *_h); +int bed_overlap(const void *_h, const char *chr, int beg, int end); typedef struct { double p[4]; @@ -132,39 +102,53 @@ static void rm_info(bcf1_t *b, const char *key) 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[9]) { 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", 1.-pr->f_em, pr->cil, pr->cih); + { // 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 (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! 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.) { - int i, q[3], pq; + if (pr->cmp[0] >= 0.) { // two sample groups + 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, ",%g,%g,%g", pr->cmp[0], pr->cmp[1], pr->cmp[2]); + 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); @@ -184,7 +168,7 @@ static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p 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; } @@ -246,6 +230,10 @@ static void write_header(bcf_hdr_t *h) kputs("##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); + kputs("##INFO=\n", &str); if (!strstr(str.s, "##FORMAT=\n", &str); if (!strstr(str.s, "##FORMAT=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[]) { @@ -293,15 +281,14 @@ int bcfview(int argc, char *argv[]) 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; 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; + while ((c = getopt(argc, argv, "FN1:l:cC:eHAGvbSuP:t:p:QgLi:IMs:D:U:X:d:")) >= 0) { switch (c) { case '1': vc.n1 = atoi(optarg); break; - case 'l': vc.fn_list = strdup(optarg); break; + case 'l': vc.bed = bed_read(optarg); break; case 'D': vc.fn_dict = strdup(optarg); break; case 'F': vc.flag |= VC_FIX_PL; break; case 'N': vc.flag |= VC_ACGT_ONLY; break; @@ -310,6 +297,7 @@ int bcfview(int argc, char *argv[]) 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; @@ -321,6 +309,7 @@ int bcfview(int argc, char *argv[]) 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); @@ -338,35 +327,41 @@ int bcfview(int argc, char *argv[]) } if (argc == optind) { 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 (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, " -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, " -F PL generated by r921 or before\n"); - fprintf(stderr, " -d FLOAT skip loci where less than FLOAT fraction of samples covered [0]\n"); - fprintf(stderr, " -D FILE sequence dictionary for VCF->BCF conversion [null]\n"); - fprintf(stderr, " -l FILE list of sites to output [all sites]\n"); - fprintf(stderr, " -t FLOAT scaled substitution mutation rate [%.4g]\n", vc.theta); - fprintf(stderr, " -i FLOAT indel-to-substitution ratio [%.4g]\n", vc.indel_frac); - fprintf(stderr, " -p FLOAT variant if P(ref|D) [reg]\n\n"); + fprintf(stderr, "Input/output options:\n\n"); + fprintf(stderr, " -A keep all possible alternate alleles at variant sites\n"); + fprintf(stderr, " -b output BCF instead of VCF\n"); + fprintf(stderr, " -D FILE sequence dictionary for VCF->BCF conversion [null]\n"); + fprintf(stderr, " -F PL generated by r921 or before (which generate old ordering)\n"); + fprintf(stderr, " -G suppress all individual genotype information\n"); + fprintf(stderr, " -l FILE list of sites (chr pos) or regions (BED) to output [all sites]\n"); + fprintf(stderr, " -L calculate LD for adjacent sites\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, " -s FILE list of samples to use [all samples]\n"); + 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 (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)BCF conversion please specify the sequence dictionary with -D\n", __func__); return 1; @@ -405,13 +400,12 @@ int bcfview(int argc, char *argv[]) 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); } 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, hin); if (optind + 1 < argc && !(vc.flag&VC_VCFIN)) { void *str2id = bcf_build_refhash(hout); if (bcf_parse_region(str2id, argv[optind+1], &tid, &begin, &end) >= 0) { @@ -431,6 +425,7 @@ int bcfview(int argc, char *argv[]) } while (vcf_read(bp, hin, b) > 0) { int is_indel; + double em[9]; 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); @@ -447,13 +442,7 @@ int bcfview(int argc, char *argv[]) x = toupper(b->ref[0]); if (x != 'A' && x != 'C' && x != 'G' && x != 'T') continue; } - if (hash) { - uint64_t x = (uint64_t)b->tid<<32 | b->pos; - khint_t k = kh_get(set64, hash, x); - if (kh_size(hash) == 0) break; - if (k == kh_end(hash)) continue; - kh_del(set64, hash, k); - } + if (vc.bed && !bed_overlap(vc.bed, hin->ns[b->tid], b->pos, b->pos + strlen(b->ref))) continue; if (tid >= 0) { int l = strlen(b->ref); l = b->pos + (l > 0? l : 1); @@ -465,10 +454,15 @@ int bcfview(int argc, char *argv[]) bcf_2qcall(hout, b); continue; } + if (vc.flag & VC_EM) bcf_em1(b, vc.n1, 0xff, em); + else { + int i; + for (i = 0; i < 9; ++i) em[i] = -1.; + } if (vc.flag & (VC_CALL|VC_ADJLD)) bcf_gl2pl(b); 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); @@ -478,21 +472,28 @@ int bcfview(int argc, char *argv[]) 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[9]; 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); + } else if (vc.flag & VC_EM) update_bcf1(b, 0, 0, 0, vc.flag, em); 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); - ksprintf(&s, "NEIR=%.3f;NEIF=%.3f,%.3f", r2, f[0]+f[2], f[0]+f[1]); + ksprintf(&s, "NEIR=%.3f;NEIF4=%.3f,%.3f,%.3f,%.3f", r2, f[0], f[1], f[2], f[3]); bcf_append_info(b, s.s, s.l); free(s.s); } @@ -512,8 +513,6 @@ int bcfview(int argc, char *argv[]) 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) { @@ -521,6 +520,7 @@ int bcfview(int argc, char *argv[]) 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 (seeds) free(seeds); if (p1) bcf_p1_destroy(p1); return 0;