11 KHASH_SET_INIT_INT64(set64)
14 KSTREAM_INIT(gzFile, gzread, 16384)
23 #define VC_KEEPALT 256
24 #define VC_ACGT_ONLY 512
26 #define VC_CALL_GT 2048
28 #define VC_NO_INDEL 8192
31 int flag, prior_type, n1;
32 char *fn_list, *prior_file;
33 double theta, pref, indel_frac;
36 khash_t(set64) *bcf_load_pos(const char *fn, bcf_hdr_t *_h)
41 int ret, dret, lineno = 1;
43 khash_t(set64) *hash = 0;
45 hash = kh_init(set64);
46 str2id = bcf_build_refhash(_h);
47 str = calloc(1, sizeof(kstring_t));
48 fp = strcmp(fn, "-")? gzopen(fn, "r") : gzdopen(fileno(stdin), "r");
50 while (ks_getuntil(ks, 0, str, &dret) >= 0) {
51 int tid = bcf_str2id(str2id, str->s);
52 if (tid >= 0 && dret != '\n') {
53 if (ks_getuntil(ks, 0, str, &dret) >= 0) {
54 uint64_t x = (uint64_t)tid<<32 | (atoi(str->s) - 1);
55 kh_put(set64, hash, x, &ret);
57 } else fprintf(stderr, "[%s] %s is not a reference name (line %d).\n", __func__, str->s, lineno);
58 if (dret != '\n') while ((dret = ks_getc(ks)) > 0 && dret != '\n');
62 bcf_str2id_destroy(str2id);
65 free(str->s); free(str);
69 static double test_hwe(const double g[3])
71 extern double kf_gammaq(double p, double x);
72 double fexp, chi2, f[3], n;
74 n = g[0] + g[1] + g[2];
75 fexp = (2. * g[2] + g[1]) / (2. * n);
76 if (fexp > 1. - 1e-10) fexp = 1. - 1e-10;
77 if (fexp < 1e-10) fexp = 1e-10;
78 f[0] = n * (1. - fexp) * (1. - fexp);
79 f[1] = n * 2. * fexp * (1. - fexp);
80 f[2] = n * fexp * fexp;
81 for (i = 0, chi2 = 0.; i < 3; ++i)
82 chi2 += (g[i] - f[i]) * (g[i] - f[i]) / f[i];
83 return kf_gammaq(.5, chi2 / 2.);
88 int mq, depth, is_tested, d[4];
91 static double ttest(int n1, int n2, int a[4])
93 extern double kf_betai(double a, double b, double x);
95 if (n1 == 0 || n2 == 0 || n1 + n2 < 3) return 1.0;
96 u1 = (double)a[0] / n1; u2 = (double)a[2] / n2;
97 if (u1 <= u2) return 1.;
98 t = (u1 - u2) / sqrt(((a[1] - n1 * u1 * u1) + (a[3] - n2 * u2 * u2)) / (n1 + n2 - 2) * (1./n1 + 1./n2));
100 // printf("%d,%d,%d,%d,%lf,%lf,%lf\n", a[0], a[1], a[2], a[3], t, u1, u2);
101 return t < 0.? 1. : .5 * kf_betai(.5*v, .5, v/(v+t*t));
104 static int test16_core(int anno[16], anno16_t *a)
106 extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
109 a->p[0] = a->p[1] = a->p[2] = a->p[3] = 1.;
110 memcpy(a->d, anno, 4 * sizeof(int));
111 a->depth = anno[0] + anno[1] + anno[2] + anno[3];
112 a->is_tested = (anno[0] + anno[1] > 0 && anno[2] + anno[3] > 0);
113 if (a->depth == 0) return -1;
114 a->mq = (int)(sqrt((anno[9] + anno[11]) / a->depth) + .499);
115 kt_fisher_exact(anno[0], anno[1], anno[2], anno[3], &left, &right, &a->p[0]);
116 for (i = 1; i < 4; ++i)
117 a->p[i] = ttest(anno[0] + anno[1], anno[2] + anno[3], anno+4*i);
121 static int test16(bcf1_t *b, anno16_t *a)
125 a->p[0] = a->p[1] = a->p[2] = a->p[3] = 1.;
126 a->d[0] = a->d[1] = a->d[2] = a->d[3] = 0.;
127 a->mq = a->depth = a->is_tested = 0;
128 if ((p = strstr(b->info, "I16=")) == 0) return -1;
130 for (i = 0; i < 16; ++i) {
131 errno = 0; anno[i] = strtol(p, &p, 10);
132 if (anno[i] == 0 && (errno == EINVAL || errno == ERANGE)) return -2;
135 return test16_core(anno, a);
138 static void rm_info(bcf1_t *b, const char *key)
141 if ((p = strstr(b->info, key)) == 0) return;
142 for (q = p; *q && *q != ';'; ++q);
143 if (p > b->info && *(p-1) == ';') --p;
144 memmove(p, q, b->l_str - (q - b->str));
149 static int update_bcf1(int n_smpl, bcf1_t *b, const bcf_p1aux_t *pa, const bcf_p1rst_t *pr, double pref, int flag)
152 int is_var = (pr->p_ref < pref);
153 double p_hwe, r = is_var? pr->p_ref : pr->p_var, fq;
156 p_hwe = pr->g[0] >= 0.? test_hwe(pr->g) : 1.0; // only do HWE g[] is calculated
160 memset(&s, 0, sizeof(kstring_t));
161 kputc('\0', &s); kputs(b->ref, &s); kputc('\0', &s);
162 kputs(b->alt, &s); kputc('\0', &s); kputc('\0', &s);
164 if (b->info[0]) kputc(';', &s);
165 // ksprintf(&s, "AF1=%.4lg;AFE=%.4lg;CI95=%.4lg,%.4lg", 1.-pr->f_em, 1.-pr->f_exp, pr->cil, pr->cih);
166 ksprintf(&s, "AF1=%.4g;CI95=%.4g,%.4g", 1.-pr->f_em, pr->cil, pr->cih);
167 ksprintf(&s, ";DP4=%d,%d,%d,%d;MQ=%d", a.d[0], a.d[1], a.d[2], a.d[3], a.mq);
168 fq = pr->p_ref_folded < 0.5? -4.343 * log(pr->p_ref_folded) : 4.343 * log(pr->p_var_folded);
169 if (fq < -999) fq = -999;
170 if (fq > 999) fq = 999;
171 ksprintf(&s, ";FQ=%.3g", fq);
173 if (pr->pc[0] >= 0.) ksprintf(&s, ";PC4=%g,%g,%g,%g", pr->pc[0], pr->pc[1], pr->pc[2], pr->pc[3]);
174 ksprintf(&s, ";PV4=%.2g,%.2g,%.2g,%.2g", a.p[0], a.p[1], a.p[2], a.p[3]);
176 if (pr->g[0] >= 0. && p_hwe <= .2)
177 ksprintf(&s, ";GC=%.2f,%.2f,%.2f;HWE=%.3f", pr->g[2], pr->g[1], pr->g[0], p_hwe);
179 kputs(b->fmt, &s); kputc('\0', &s);
181 b->m_str = s.m; b->l_str = s.l; b->str = s.s;
182 b->qual = r < 1e-100? 999 : -4.343 * log(r);
183 if (b->qual > 999) b->qual = 999;
185 if (!is_var) bcf_shrink_alt(b, 1);
186 else if (!(flag&VC_KEEPALT))
187 bcf_shrink_alt(b, pr->rank0 < 2? 2 : pr->rank0+1);
188 if (is_var && (flag&VC_CALL_GT)) { // call individual genotype
189 int i, x, old_n_gi = b->n_gi;
190 s.m = b->m_str; s.l = b->l_str - 1; s.s = b->str;
191 kputs(":GT:GQ", &s); kputc('\0', &s);
192 b->m_str = s.m; b->l_str = s.l; b->str = s.s;
194 for (i = 0; i < b->n_smpl; ++i) {
195 x = bcf_p1_call_gt(pa, pr->f_em, i);
196 ((uint8_t*)b->gi[old_n_gi].data)[i] = (x&3) == 0? 1<<3|1 : (x&3) == 1? 1 : 0;
197 ((uint8_t*)b->gi[old_n_gi+1].data)[i] = x>>2;
203 static void write_header(bcf_hdr_t *h)
206 str.l = h->l_txt? h->l_txt - 1 : 0;
207 str.m = str.l + 1; str.s = h->txt;
208 if (!strstr(str.s, "##INFO=<ID=DP,"))
209 kputs("##INFO=<ID=DP,Number=1,Type=Integer,Description=\"Raw read depth\">\n", &str);
210 if (!strstr(str.s, "##INFO=<ID=DP4,"))
211 kputs("##INFO=<ID=DP4,Number=4,Type=Integer,Description=\"# high-quality ref-forward bases, ref-reverse, alt-forward and alt-reverse bases\">\n", &str);
212 if (!strstr(str.s, "##INFO=<ID=MQ,"))
213 kputs("##INFO=<ID=MQ,Number=1,Type=Integer,Description=\"Root-mean-square mapping quality of covering reads\">\n", &str);
214 if (!strstr(str.s, "##INFO=<ID=FQ,"))
215 kputs("##INFO=<ID=FQ,Number=1,Type=Integer,Description=\"Phred probability that sample chromosomes are not all the same\">\n", &str);
216 if (!strstr(str.s, "##INFO=<ID=AF1,"))
217 kputs("##INFO=<ID=AF1,Number=1,Type=Float,Description=\"Max-likelihood estimate of the site allele frequency of the first ALT allele\">\n", &str);
218 if (!strstr(str.s, "##INFO=<ID=CI95,"))
219 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);
220 if (!strstr(str.s, "##INFO=<ID=PV4,"))
221 kputs("##INFO=<ID=PV4,Number=4,Type=Float,Description=\"P-values for strand bias, baseQ bias, mapQ bias and tail distance bias\">\n", &str);
222 if (!strstr(str.s, "##FORMAT=<ID=DP,"))
223 kputs("##FORMAT=<ID=DP,Number=1,Type=Integer,Description=\"# high-quality bases\">\n", &str);
224 if (!strstr(str.s, "##FORMAT=<ID=SP,"))
225 kputs("##FORMAT=<ID=SP,Number=1,Type=Integer,Description=\"Phred-scaled strand bias P-value\">\n", &str);
226 if (!strstr(str.s, "##FORMAT=<ID=PL,"))
227 kputs("##FORMAT=<ID=PL,Number=(#ALT+1)*(#ALT+2)/2,Type=Integer,Description=\"List of Phred-scaled genotype likelihoods\">\n", &str);
228 h->l_txt = str.l + 1; h->txt = str.s;
231 double bcf_ld_freq(const bcf1_t *b0, const bcf1_t *b1, double f[4]);
233 int bcfview(int argc, char *argv[])
235 extern int bcf_2qcall(bcf_hdr_t *h, bcf1_t *b);
236 extern void bcf_p1_indel_prior(bcf_p1aux_t *ma, double x);
237 extern int bcf_fix_gt(bcf1_t *b);
238 bcf_t *bp, *bout = 0;
241 uint64_t n_processed = 0;
246 char moder[4], modew[4];
247 khash_t(set64) *hash = 0;
249 tid = begin = end = -1;
250 memset(&vc, 0, sizeof(viewconf_t));
251 vc.prior_type = vc.n1 = -1; vc.theta = 1e-3; vc.pref = 0.5; vc.indel_frac = -1.;
252 while ((c = getopt(argc, argv, "N1:l:cHAGvbSuP:t:p:QgLi:I")) >= 0) {
254 case '1': vc.n1 = atoi(optarg); break;
255 case 'l': vc.fn_list = strdup(optarg); break;
256 case 'N': vc.flag |= VC_ACGT_ONLY; break;
257 case 'G': vc.flag |= VC_NO_GENO; break;
258 case 'A': vc.flag |= VC_KEEPALT; break;
259 case 'b': vc.flag |= VC_BCFOUT; break;
260 case 'S': vc.flag |= VC_VCFIN; break;
261 case 'c': vc.flag |= VC_CALL; break;
262 case 'v': vc.flag |= VC_VARONLY | VC_CALL; break;
263 case 'u': vc.flag |= VC_UNCOMP | VC_BCFOUT; break;
264 case 'H': vc.flag |= VC_HWE; break;
265 case 'g': vc.flag |= VC_CALL_GT | VC_CALL; break;
266 case 'I': vc.flag |= VC_NO_INDEL; break;
267 case 't': vc.theta = atof(optarg); break;
268 case 'p': vc.pref = atof(optarg); break;
269 case 'i': vc.indel_frac = atof(optarg); break;
270 case 'Q': vc.flag |= VC_QCALL; break;
271 case 'L': vc.flag |= VC_ADJLD; break;
273 if (strcmp(optarg, "full") == 0) vc.prior_type = MC_PTYPE_FULL;
274 else if (strcmp(optarg, "cond2") == 0) vc.prior_type = MC_PTYPE_COND2;
275 else if (strcmp(optarg, "flat") == 0) vc.prior_type = MC_PTYPE_FLAT;
276 else vc.prior_file = strdup(optarg);
280 if (argc == optind) {
281 fprintf(stderr, "\n");
282 fprintf(stderr, "Usage: bcftools view [options] <in.bcf> [reg]\n\n");
283 fprintf(stderr, "Options: -c SNP calling\n");
284 fprintf(stderr, " -v output potential variant sites only (force -c)\n");
285 fprintf(stderr, " -g call genotypes at variant sites (force -c)\n");
286 fprintf(stderr, " -b output BCF instead of VCF\n");
287 fprintf(stderr, " -u uncompressed BCF output (force -b)\n");
288 fprintf(stderr, " -S input is VCF\n");
289 fprintf(stderr, " -A keep all possible alternate alleles at variant sites\n");
290 fprintf(stderr, " -G suppress all individual genotype information\n");
291 fprintf(stderr, " -H perform Hardy-Weinberg test (slower)\n");
292 fprintf(stderr, " -N skip sites where REF is not A/C/G/T\n");
293 fprintf(stderr, " -Q output the QCALL likelihood format\n");
294 fprintf(stderr, " -L calculate LD for adjacent sites\n");
295 fprintf(stderr, " -I skip indels\n");
296 fprintf(stderr, " -1 INT number of group-1 samples [0]\n");
297 fprintf(stderr, " -l FILE list of sites to output [all sites]\n");
298 fprintf(stderr, " -t FLOAT scaled substitution mutation rate [%.4g]\n", vc.theta);
299 fprintf(stderr, " -i FLOAT indel-to-substitution ratio [%.4g]\n", vc.indel_frac);
300 fprintf(stderr, " -p FLOAT variant if P(ref|D)<FLOAT [%.3g]\n", vc.pref);
301 fprintf(stderr, " -P STR type of prior: full, cond2, flat [full]\n");
302 fprintf(stderr, "\n");
306 b = calloc(1, sizeof(bcf1_t));
307 blast = calloc(1, sizeof(bcf1_t));
309 if (!(vc.flag & VC_VCFIN)) strcat(moder, "b");
311 if (vc.flag & VC_BCFOUT) strcat(modew, "b");
312 if (vc.flag & VC_UNCOMP) strcat(modew, "u");
313 bp = vcf_open(argv[optind], moder);
314 h = vcf_hdr_read(bp);
315 bout = vcf_open("-", modew);
316 if (!(vc.flag & VC_QCALL)) {
317 if (vc.flag & VC_CALL) write_header(h);
318 vcf_hdr_write(bout, h);
320 if (vc.flag & VC_CALL) {
321 p1 = bcf_p1_init(h->n_smpl);
323 if (bcf_p1_read_prior(p1, vc.prior_file) < 0) {
324 fprintf(stderr, "[%s] fail to read the prior AFS.\n", __func__);
327 } else bcf_p1_init_prior(p1, vc.prior_type, vc.theta);
329 bcf_p1_set_n1(p1, vc.n1);
330 bcf_p1_init_subprior(p1, vc.prior_type, vc.theta);
332 if (vc.indel_frac > 0.) bcf_p1_indel_prior(p1, vc.indel_frac); // otherwise use the default indel_frac
334 if (vc.fn_list) hash = bcf_load_pos(vc.fn_list, h);
335 if (optind + 1 < argc && !(vc.flag&VC_VCFIN)) {
336 void *str2id = bcf_build_refhash(h);
337 if (bcf_parse_region(str2id, argv[optind+1], &tid, &begin, &end) >= 0) {
339 idx = bcf_idx_load(argv[optind]);
342 off = bcf_idx_query(idx, tid, begin);
344 fprintf(stderr, "[%s] no records in the query region.\n", __func__);
345 return 1; // FIXME: a lot of memory leaks...
347 bgzf_seek(bp->fp, off, SEEK_SET);
348 bcf_idx_destroy(idx);
352 while (vcf_read(bp, h, b) > 0) {
353 int is_indel = bcf_is_indel(b);
354 if ((vc.flag & VC_NO_INDEL) && is_indel) continue;
355 if ((vc.flag & VC_ACGT_ONLY) && !is_indel) {
357 if (b->ref[0] == 0 || b->ref[1] != 0) continue;
358 x = toupper(b->ref[0]);
359 if (x != 'A' && x != 'C' && x != 'G' && x != 'T') continue;
362 uint64_t x = (uint64_t)b->tid<<32 | b->pos;
363 khint_t k = kh_get(set64, hash, x);
364 if (kh_size(hash) == 0) break;
365 if (k == kh_end(hash)) continue;
366 kh_del(set64, hash, k);
369 int l = strlen(b->ref);
370 l = b->pos + (l > 0? l : 1);
371 if (b->tid != tid || b->pos >= end) break;
372 if (!(l > begin && end > b->pos)) continue;
375 if (vc.flag & VC_QCALL) { // output QCALL format; STOP here
379 if (vc.flag & (VC_CALL|VC_ADJLD)) bcf_gl2pl(b);
380 if (vc.flag & VC_CALL) { // call variants
382 bcf_p1_cal(b, p1, &pr); // pr.g[3] is not calculated here
383 if (vc.flag&VC_HWE) bcf_p1_cal_g3(p1, pr.g);
384 if (n_processed % 100000 == 0) {
385 fprintf(stderr, "[%s] %ld sites processed.\n", __func__, (long)n_processed);
388 if (pr.p_ref >= vc.pref && (vc.flag & VC_VARONLY)) continue;
389 update_bcf1(h->n_smpl, b, p1, &pr, vc.pref, vc.flag);
391 if (vc.flag & VC_ADJLD) { // compute LD
393 if ((r2 = bcf_ld_freq(blast, b, f)) >= 0) {
395 s.m = s.l = 0; s.s = 0;
396 if (*b->info) kputc(';', &s);
397 ksprintf(&s, "NEIR=%.3f;NEIF=%.3f,%.3f", r2, f[0]+f[2], f[0]+f[1]);
398 bcf_append_info(b, s.s, s.l);
403 if (vc.flag & VC_NO_GENO) { // do not output GENO fields
408 vcf_write(bout, h, b);
410 if (vc.prior_file) free(vc.prior_file);
411 if (vc.flag & VC_CALL) bcf_p1_dump_afs(p1);
413 bcf_destroy(b); bcf_destroy(blast);
414 vcf_close(bp); vcf_close(bout);
415 if (hash) kh_destroy(set64, hash);
416 if (vc.fn_list) free(vc.fn_list);
417 if (p1) bcf_p1_destroy(p1);