X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bcftools%2Fcall1.c;h=7aca1595cf7de74147a98492c08e34bd560c162f;hb=674ffee7adcfc928f5039777180206fdebc5539b;hp=e4ba879a64b3d6bdd4f9ff230457b6e4394c784b;hpb=c5d9a2277fc92480fb816fcc748f36213917250a;p=samtools.git diff --git a/bcftools/call1.c b/bcftools/call1.c index e4ba879..7aca159 100644 --- a/bcftools/call1.c +++ b/bcftools/call1.c @@ -6,9 +6,7 @@ #include "bcf.h" #include "prob1.h" #include "kstring.h" - -#include "khash.h" -KHASH_SET_INIT_INT64(set64) +#include "time.h" #include "kseq.h" KSTREAM_INIT(gzFile, gzread, 16384) @@ -19,7 +17,6 @@ KSTREAM_INIT(gzFile, gzread, 16384) #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 @@ -27,62 +24,19 @@ KSTREAM_INIT(gzFile, gzread, 16384) #define VC_ADJLD 4096 #define VC_NO_INDEL 8192 #define VC_ANNO_MAX 16384 +#define VC_FIX_PL 32768 typedef struct { - int flag, prior_type, n1, n_sub, *sublist; - char *fn_list, *prior_file, **subsam; - double theta, pref, indel_frac; + int flag, prior_type, n1, n_sub, *sublist, n_perm; + char *prior_file, **subsam, *fn_dict; + uint8_t *ploidy; + double theta, pref, indel_frac, min_perm_p, min_smpl_frac; + 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; -} - -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.); -} +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]; @@ -150,12 +104,11 @@ static void rm_info(bcf1_t *b, const char *key) static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p1rst_t *pr, double pref, int flag) { kstring_t s; - int is_var = (pr->p_ref < pref); - double p_hwe, r = is_var? pr->p_ref : pr->p_var, fq; + int has_I16, is_var = (pr->p_ref < pref); + double fq, r = is_var? pr->p_ref : pr->p_var; anno16_t a; - p_hwe = pr->g[0] >= 0.? test_hwe(pr->g) : 1.0; // only do HWE g[] is calculated - test16(b, &a); + has_I16 = test16(b, &a) >= 0? 1 : 0; rm_info(b, "I16="); memset(&s, 0, sizeof(kstring_t)); @@ -165,17 +118,25 @@ static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p 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); - ksprintf(&s, ";DP4=%d,%d,%d,%d;MQ=%d", a.d[0], a.d[1], a.d[2], a.d[3], a.mq); + 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 (a.is_tested) { - 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->cmp[0] >= 0.) { // two sample groups + int i, q[3], pq; + 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]); } - if (pr->g[0] >= 0. && p_hwe <= .2) - ksprintf(&s, ";GC=%.2f,%.2f,%.2f;HWE=%.3f", pr->g[2], pr->g[1], pr->g[0], p_hwe); + 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); kputs(b->fmt, &s); kputc('\0', &s); free(b->str); @@ -214,11 +175,24 @@ static char **read_samples(const char *fn, int *_n) 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); } - sam[n++] = strdup(s.s); + 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); @@ -239,7 +213,7 @@ static void write_header(bcf_hdr_t *h) if (!strstr(str.s, "##INFO=\n", &str); if (!strstr(str.s, "##INFO=\n", &str); + 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); + 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); @@ -271,25 +253,26 @@ int bcfview(int argc, char *argv[]) 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); + extern int bcf_shuffle(bcf1_t *b, int seed); bcf_t *bp, *bout = 0; bcf1_t *b, *blast; - int c; + int c, *seeds = 0; uint64_t n_processed = 0; viewconf_t vc; bcf_p1aux_t *p1 = 0; 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_sub = 0; vc.subsam = 0; vc.sublist = 0; - while ((c = getopt(argc, argv, "N1:l:cHAGvbSuP:t:p:QgLi:IMs:")) >= 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; + while ((c = getopt(argc, argv, "FN1:l:cHAGvbSuP: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; case 'G': vc.flag |= VC_NO_GENO; break; case 'A': vc.flag |= VC_KEEPALT; break; @@ -298,7 +281,6 @@ int bcfview(int argc, char *argv[]) case 'c': vc.flag |= VC_CALL; 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 | VC_CALL; break; case 'I': vc.flag |= VC_NO_INDEL; break; case 'M': vc.flag |= VC_ANNO_MAX; break; @@ -307,7 +289,14 @@ int bcfview(int argc, char *argv[]) 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); break; + case 'U': vc.n_perm = atoi(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); + 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; @@ -327,22 +316,36 @@ int bcfview(int argc, char *argv[]) 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, " -H perform Hardy-Weinberg test (slower)\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, " -1 INT number of group-1 samples [0]\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)BCF conversion please specify the sequence dictionary with -D\n", __func__); + return 1; + } + if (vc.n1 <= 0) vc.n_perm = 0; // TODO: give a warning here! + if (vc.n_perm > 0) { + seeds = malloc(vc.n_perm * sizeof(int)); + srand48(time(0)); + for (c = 0; c < vc.n_perm; ++c) seeds[c] = lrand48(); + } b = calloc(1, sizeof(bcf1_t)); blast = calloc(1, sizeof(bcf1_t)); strcpy(moder, "r"); @@ -352,6 +355,8 @@ int bcfview(int argc, char *argv[]) if (vc.flag & VC_UNCOMP) strcat(modew, "u"); bp = vcf_open(argv[optind], moder); 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)) { if (vc.n_sub) { @@ -362,7 +367,7 @@ int bcfview(int argc, char *argv[]) vcf_hdr_write(bout, hout); } if (vc.flag & VC_CALL) { - p1 = bcf_p1_init(hout->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__); @@ -375,7 +380,6 @@ int bcfview(int argc, char *argv[]) } 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) { @@ -395,7 +399,14 @@ int bcfview(int argc, char *argv[]) } while (vcf_read(bp, hin, b) > 0) { int is_indel; + 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); + int n = bcf_smpl_covered(b); + if ((double)n / b->n_smpl < vc.min_smpl_frac) continue; + } if (vc.n_sub) bcf_subsam(vc.n_sub, vc.sublist, b); + if (vc.flag & VC_FIX_PL) bcf_fix_pl(b); is_indel = bcf_is_indel(b); if ((vc.flag & VC_NO_INDEL) && is_indel) continue; if ((vc.flag & VC_ACGT_ONLY) && !is_indel) { @@ -404,13 +415,7 @@ 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); @@ -426,12 +431,21 @@ int bcfview(int argc, char *argv[]) if (vc.flag & VC_CALL) { // call variants bcf_p1rst_t 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 % 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; + if (vc.n_perm && vc.n1 > 0 && pr.p_chi2 < vc.min_perm_p) { // permutation test + bcf_p1rst_t r; + int i, n = 0; + for (i = 0; i < vc.n_perm; ++i) { + bcf_shuffle(b, seeds[i]); + bcf_p1_cal(b, p1, &r); + if (pr.p_chi2 >= r.p_chi2) ++n; + } + pr.perm_rank = n; + } update_bcf1(hout->n_smpl, b, p1, &pr, vc.pref, vc.flag); } if (vc.flag & VC_ADJLD) { // compute LD @@ -460,13 +474,15 @@ 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) { 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 (seeds) free(seeds); if (p1) bcf_p1_destroy(p1); return 0; }