static int fake_PLs(args_t *args, bcf_hdr_t *hdr, bcf1_t *line)
{
// PLs not present, use GTs instead.
int fake_PL = args->no_PLs ? args->no_PLs : 99; // with 1, discordance is the number of non-matching GTs
int nsm_gt, i;
if ( (nsm_gt=bcf_get_genotypes(hdr, line, &args->tmp_arr, &args->ntmp_arr)) <= 0 )
error("GT not present at %s:%d?\n", hdr->id[BCF_DT_CTG][line->rid].key, line->pos+1);
nsm_gt /= bcf_hdr_nsamples(hdr);
int npl = line->n_allele*(line->n_allele+1)/2;
hts_expand(int,npl*bcf_hdr_nsamples(hdr),args->npl_arr,args->pl_arr);
for (i=0; i<bcf_hdr_nsamples(hdr); i++)
{
int *gt_ptr = args->tmp_arr + i*nsm_gt;
int j, *pl_ptr = args->pl_arr + i*npl;
if ( bcf_gt_is_missing(gt_ptr[0]) || bcf_gt_is_missing(gt_ptr[1]) ) // missing genotype
{
for (j=0; j<npl; j++) pl_ptr[j] = -1;
}
else
{
int a = bcf_gt_allele(gt_ptr[0]);
int b = bcf_gt_allele(gt_ptr[1]);
for (j=0; j<npl; j++) pl_ptr[j] = fake_PL;
int idx = bcf_alleles2gt(a,b);
pl_ptr[idx] = 0;
}
}
return npl;
}
int estimate_AF(args_t *args, bcf1_t *line, double *alt_freq)
{
if ( !args->nitmp )
{
args->nitmp = bcf_get_genotypes(args->hdr, line, &args->itmp, &args->mitmp);
if ( args->nitmp != 2*args->nsmpl ) return -1; // not diploid?
args->nitmp /= args->nsmpl;
}
int i, nalt = 0, nref = 0;
for (i=0; i<args->nsmpl; i++)
{
int32_t *gt = &args->itmp[i*args->nitmp];
if ( bcf_gt_is_missing(gt[0]) || bcf_gt_is_missing(gt[1]) ) continue;
if ( bcf_gt_allele(gt[0]) ) nalt++;
else nref++;
if ( bcf_gt_allele(gt[1]) ) nalt++;
else nref++;
}
if ( !nalt && !nref ) return -1;
*alt_freq = (double)nalt / (nalt + nref);
return 0;
}
int process_GT(args_t *args, bcf1_t *line, uint32_t *ntot, uint32_t *ndif)
{
int ngt = bcf_get_genotypes(args->sm_hdr, line, &args->tmp_arr, &args->ntmp_arr);
if ( ngt<=0 ) return 1; // GT not present
if ( ngt!=args->nsmpl*2 ) return 2; // not diploid
ngt /= args->nsmpl;
int i,j, idx = 0;
for (i=1; i<args->nsmpl; i++)
{
int32_t *a = args->tmp_arr + i*ngt;
if ( bcf_gt_is_missing(a[0]) || bcf_gt_is_missing(a[1]) || a[1]==bcf_int32_vector_end ) { idx+=i; continue; }
int agt = 1<<bcf_gt_allele(a[0]) | 1<<bcf_gt_allele(a[1]);
for (j=0; j<i; j++)
{
int32_t *b = args->tmp_arr + j*ngt;
if ( bcf_gt_is_missing(b[0]) || bcf_gt_is_missing(b[1]) || b[1]==bcf_int32_vector_end ) { idx++; continue; }
int bgt = 1<<bcf_gt_allele(b[0]) | 1<<bcf_gt_allele(b[1]);
ntot[idx]++;
if ( agt!=bgt ) ndif[idx]++;
idx++;
}
}
return 0;
}
static void set_observed_prob_unrelated(bcf1_t *rec)
{
float af = 0.5; // alternate allele frequency
int ngt = bcf_get_genotypes(args.hdr, rec, &args.gt_arr, &args.ngt_arr);
if ( ngt<0 ) return;
if ( ngt!=4 ) return; // chrX
int32_t a,b,c,d;
a = args.gt_arr[2*args.isample];
b = args.gt_arr[2*args.isample+1];
c = args.gt_arr[2*args.jsample];
d = args.gt_arr[2*args.jsample+1];
if ( bcf_gt_is_missing(a) || bcf_gt_is_missing(b) ) return;
if ( bcf_gt_is_missing(c) || bcf_gt_is_missing(d) ) return;
if ( !bcf_gt_is_phased(a) && !bcf_gt_is_phased(b) ) return; // only the second allele should be set when phased
if ( !bcf_gt_is_phased(c) && !bcf_gt_is_phased(d) ) return;
a = bcf_gt_allele(a);
b = bcf_gt_allele(b);
c = bcf_gt_allele(c);
d = bcf_gt_allele(d);
int m = args.msites;
args.nsites++;
hts_expand(uint32_t,args.nsites,args.msites,args.sites);
if ( m!=args.msites ) args.eprob = (double*) realloc(args.eprob, sizeof(double)*args.msites*args.nstates);
args.sites[args.nsites-1] = rec->pos;
double *prob = args.eprob + args.nstates*(args.nsites-1);
prob[UNRL_xxxx] = prob_not_shared(af,a,c) * prob_not_shared(af,a,d) * prob_not_shared(af,b,c) * prob_not_shared(af,b,d);
prob[UNRL_0x0x] = prob_shared(af,a,c) * prob_not_shared(af,b,d);
prob[UNRL_0xx0] = prob_shared(af,a,d) * prob_not_shared(af,b,c);
prob[UNRL_x00x] = prob_shared(af,b,c) * prob_not_shared(af,a,d);
prob[UNRL_x0x0] = prob_shared(af,b,d) * prob_not_shared(af,a,c);
prob[UNRL_0101] = prob_shared(af,a,c) * prob_shared(af,b,d);
prob[UNRL_0110] = prob_shared(af,a,d) * prob_shared(af,b,c);
#if 0
static int x = 0;
if ( !x++)
{
printf("p(0==0) .. %f\n", prob_shared(af,0,0));
printf("p(0!=0) .. %f\n", prob_not_shared(af,0,0));
printf("p(0==1) .. %f\n", prob_shared(af,0,1));
printf("p(0!=1) .. %f\n", prob_not_shared(af,0,1));
}
printf("%d|%d %d|%d x:%f 11:%f 12:%f 21:%f 22:%f 11,22:%f 12,21:%f %d\n", a,b,c,d,
prob[UNRL_xxxx], prob[UNRL_0x0x], prob[UNRL_0xx0], prob[UNRL_x00x], prob[UNRL_x0x0], prob[UNRL_0101], prob[UNRL_0110], rec->pos+1);
#endif
}
int process_region_precise(args_t *args, char *seq, regitr_t *itr)
{
int k = 1;
uint32_t start = itr->reg[itr->i].start, end = itr->reg[itr->i].end;
while ( itr->i+k<itr->n && start==itr->reg[itr->i+k].start && end==itr->reg[itr->i+k].end ) k++;
int ret = ploidy_query(args->ploidy, seq, start, args->sex2ploidy, NULL, NULL);
assert(ret);
memset(args->counts,0,args->ncounts*sizeof(int));
// Select 'nsites' sites spaced so that they evenly cover the whole region
// to get a representative sample. We index-jump as we should be checking
// a few sites only.
int i, rid = -1, pos, prev_pos = -1, ismpl;
for (i=0; i<args->nsites; i++)
{
rid = -1;
pos = ((i+1.0)/(args->nsites+1))*(end - start) + start;
if ( i>0 && pos <= prev_pos ) continue; // the vcf is too sparse
if ( bcf_sr_seek(args->sr,seq,pos)!=0 ) return k; // sequence not present
if ( !bcf_sr_next_line(args->sr) ) return k; // no sites found
bcf1_t *rec = bcf_sr_get_line(args->sr,0);
if ( rid==-1 ) rid = rec->rid;
if ( rid!=rec->rid || rec->pos > end ) break;
prev_pos = rec->pos;
int ngts = bcf_get_genotypes(args->hdr,rec,&args->gts,&args->ngts);
ngts /= args->nsample;
for (ismpl=0; ismpl<args->nsample; ismpl++)
{
int32_t *gts = args->gts + ngts*ismpl;
int igt, ploidy = 0;
for (igt=0; igt<ngts; igt++)
{
if ( gts[igt]==bcf_int32_vector_end || bcf_gt_is_missing(gts[igt]) ) break;
else ploidy++;
}
args->counts[ismpl*(args->max_ploidy+1) + ploidy]++;
if ( args->verbose )
fprintf(stderr,"%s:%d\t%s\tploidy=%d\n", seq,rec->pos+1,args->hdr->samples[ismpl],ploidy);
}
}
for (ismpl=0; ismpl<args->nsample; ismpl++)
{
float sum = 0, *probs = args->sex2prob + ismpl*args->nsex;
int *counts = args->counts + ismpl*(args->max_ploidy+1);
for (i=0; i<args->max_ploidy+1; i++) sum += counts[i];
if ( !sum ) continue;
for (i=0; i<args->nsex; i++)
{
int ploidy = args->sex2ploidy[i];
probs[i] *= counts[ploidy]/sum;
}
}
return k;
}
static void phase_update(args_t *args, bcf_hdr_t *hdr, bcf1_t *rec)
{
int i, nGTs = bcf_get_genotypes(hdr, rec, &args->GTa, &args->mGTa);
for (i=0; i<bcf_hdr_nsamples(hdr); i++)
{
if ( !args->swap_phase[i] ) continue;
int *gt = &args->GTa[i*2];
if ( bcf_gt_is_missing(gt[0]) || gt[1]==bcf_int32_vector_end ) continue;
SWAP(int, gt[0], gt[1]);
gt[1] |= 1;
}
bcf_update_genotypes(hdr,rec,args->GTa,nGTs);
}
static void set_observed_prob_trio(bcf1_t *rec)
{
int ngt = bcf_get_genotypes(args.hdr, rec, &args.gt_arr, &args.ngt_arr);
if ( ngt<0 ) return;
if ( ngt!=6 ) return; // chrX
int32_t a,b,c,d,e,f;
a = args.gt_arr[2*args.imother];
b = args.gt_arr[2*args.imother+1];
c = args.gt_arr[2*args.ifather];
d = args.gt_arr[2*args.ifather+1];
e = args.gt_arr[2*args.ichild];
f = args.gt_arr[2*args.ichild+1];
if ( bcf_gt_is_missing(a) || bcf_gt_is_missing(b) ) return;
if ( bcf_gt_is_missing(c) || bcf_gt_is_missing(d) ) return;
if ( bcf_gt_is_missing(e) || bcf_gt_is_missing(f) ) return;
if ( !bcf_gt_is_phased(a) && !bcf_gt_is_phased(b) ) return; // only the second allele should be set when phased
if ( !bcf_gt_is_phased(c) && !bcf_gt_is_phased(d) ) return;
if ( !bcf_gt_is_phased(e) && !bcf_gt_is_phased(f) ) return;
a = bcf_gt_allele(a);
b = bcf_gt_allele(b);
c = bcf_gt_allele(c);
d = bcf_gt_allele(d);
e = bcf_gt_allele(e);
f = bcf_gt_allele(f);
int mother = (1<<a) | (1<<b);
int father = (1<<c) | (1<<d);
int child = (1<<e) | (1<<f);
if ( !(mother&child) || !(father&child) ) return; // Mendelian-inconsistent site, skip
if ( a!=b ) args.nhet_mother++;
if ( c!=d ) args.nhet_father++;
int m = args.msites;
args.nsites++;
hts_expand(uint32_t,args.nsites,args.msites,args.sites);
if ( m!=args.msites ) args.eprob = (double*) realloc(args.eprob, sizeof(double)*args.msites*args.nstates);
args.sites[args.nsites-1] = rec->pos;
double *prob = args.eprob + args.nstates*(args.nsites-1);
prob[TRIO_AC] = prob_shared(0,e,a) * prob_shared(0,f,c);
prob[TRIO_AD] = prob_shared(0,e,a) * prob_shared(0,f,d);
prob[TRIO_BC] = prob_shared(0,e,b) * prob_shared(0,f,c);
prob[TRIO_BD] = prob_shared(0,e,b) * prob_shared(0,f,d);
prob[TRIO_CA] = prob_shared(0,e,c) * prob_shared(0,f,a);
prob[TRIO_DA] = prob_shared(0,e,d) * prob_shared(0,f,a);
prob[TRIO_CB] = prob_shared(0,e,c) * prob_shared(0,f,b);
prob[TRIO_DB] = prob_shared(0,e,d) * prob_shared(0,f,b);
}
static void phased_flush(args_t *args)
{
if ( !args->nbuf ) return;
bcf_hdr_t *ahdr = args->files->readers[0].header;
bcf_hdr_t *bhdr = args->files->readers[1].header;
int i, j, nsmpl = bcf_hdr_nsamples(args->out_hdr);
static int gt_absent_warned = 0;
for (i=0; i<args->nbuf; i+=2)
{
bcf1_t *arec = args->buf[i];
bcf1_t *brec = args->buf[i+1];
int nGTs = bcf_get_genotypes(ahdr, arec, &args->GTa, &args->mGTa);
if ( nGTs < 0 )
{
if ( !gt_absent_warned )
{
fprintf(stderr,"GT is not present at %s:%d. (This warning is printed only once.)\n", bcf_seqname(ahdr,arec), arec->pos+1);
gt_absent_warned = 1;
}
continue;
}
if ( nGTs != 2*nsmpl ) continue; // not diploid
nGTs = bcf_get_genotypes(bhdr, brec, &args->GTb, &args->mGTb);
if ( nGTs < 0 )
{
if ( !gt_absent_warned )
{
fprintf(stderr,"GT is not present at %s:%d. (This warning is printed only once.)\n", bcf_seqname(bhdr,brec), brec->pos+1);
gt_absent_warned = 1;
}
continue;
}
if ( nGTs != 2*nsmpl ) continue; // not diploid
for (j=0; j<nsmpl; j++)
{
int *gta = &args->GTa[j*2];
int *gtb = &args->GTb[j*2];
if ( gta[1]==bcf_int32_vector_end || gtb[1]==bcf_int32_vector_end ) continue;
if ( bcf_gt_is_missing(gta[0]) || bcf_gt_is_missing(gta[1]) || bcf_gt_is_missing(gtb[0]) || bcf_gt_is_missing(gtb[1]) ) continue;
if ( !bcf_gt_is_phased(gta[1]) || !bcf_gt_is_phased(gtb[1]) ) continue;
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gta[1]) || bcf_gt_allele(gtb[0])==bcf_gt_allele(gtb[1]) ) continue;
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gtb[0]) && bcf_gt_allele(gta[1])==bcf_gt_allele(gtb[1]) )
{
if ( args->swap_phase[j] ) args->nmism[j]++; else args->nmatch[j]++;
}
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gtb[1]) && bcf_gt_allele(gta[1])==bcf_gt_allele(gtb[0]) )
{
if ( args->swap_phase[j] ) args->nmatch[j]++; else args->nmism[j]++;
}
}
}
for (i=0; i<args->nbuf/2; i+=2)
{
bcf1_t *arec = args->buf[i];
bcf_translate(args->out_hdr, args->files->readers[0].header, arec);
if ( args->nswap )
phase_update(args, args->out_hdr, arec);
if ( !args->compact_PS || args->phase_set_changed )
{
bcf_update_format_int32(args->out_hdr,arec,"PS",args->phase_set,nsmpl);
args->phase_set_changed = 0;
}
bcf_write(args->out_fh, args->out_hdr, arec);
if ( arec->pos < args->prev_pos_check ) error("FIXME, disorder: %s:%d vs %d [1]\n", bcf_seqname(args->files->readers[0].header,arec),arec->pos+1,args->prev_pos_check+1);
args->prev_pos_check = arec->pos;
}
args->nswap = 0;
for (j=0; j<nsmpl; j++)
{
if ( args->nmatch[j] >= args->nmism[j] )
args->swap_phase[j] = 0;
else
{
args->swap_phase[j] = 1;
args->nswap++;
}
if ( args->nmatch[j] && args->nmism[j] )
{
// Entropy-inspired quality. The factor 0.7 shifts and scales to (0,1)
double f = (double)args->nmatch[j]/(args->nmatch[j]+args->nmism[j]);
args->phase_qual[j] = 99*(0.7 + f*log(f) + (1-f)*log(1-f))/0.7;
}
else
args->phase_qual[j] = 99;
args->nmatch[j] = 0;
args->nmism[j] = 0;
}
int PQ_printed = 0;
for (; i<args->nbuf; i+=2)
{
bcf1_t *brec = args->buf[i+1];
bcf_translate(args->out_hdr, args->files->readers[1].header, brec);
if ( !PQ_printed )
{
bcf_update_format_int32(args->out_hdr,brec,"PQ",args->phase_qual,nsmpl);
PQ_printed = 1;
for (j=0; j<nsmpl; j++)
if ( args->phase_qual[j] < args->min_PQ )
{
args->phase_set[j] = brec->pos+1;
args->phase_set_changed = 1;
}
else if ( args->compact_PS ) args->phase_set[j] = bcf_int32_missing;
}
if ( args->nswap )
phase_update(args, args->out_hdr, brec);
if ( !args->compact_PS || args->phase_set_changed )
{
bcf_update_format_int32(args->out_hdr,brec,"PS",args->phase_set,nsmpl);
args->phase_set_changed = 0;
}
bcf_write(args->out_fh, args->out_hdr, brec);
if ( brec->pos < args->prev_pos_check ) error("FIXME, disorder: %s:%d vs %d [2]\n", bcf_seqname(args->files->readers[1].header,brec),brec->pos+1,args->prev_pos_check+1);
args->prev_pos_check = brec->pos;
}
args->nbuf = 0;
}
static void apply_variant(args_t *args, bcf1_t *rec)
{
if ( rec->n_allele==1 ) return;
if ( rec->pos <= args->fa_frz_pos )
{
fprintf(pysamerr,"The site %s:%d overlaps with another variant, skipping...\n", bcf_seqname(args->hdr,rec),rec->pos+1);
return;
}
if ( args->mask )
{
char *chr = (char*)bcf_hdr_id2name(args->hdr,args->rid);
int start = rec->pos;
int end = rec->pos + rec->rlen - 1;
if ( regidx_overlap(args->mask, chr,start,end,NULL) ) return;
}
int i, ialt = 1;
if ( args->isample >= 0 )
{
bcf_fmt_t *fmt = bcf_get_fmt(args->hdr, rec, "GT");
if ( !fmt ) return;
if ( args->haplotype )
{
if ( args->haplotype > fmt->n ) error("Can't apply %d-th haplotype at %s:%d\n", args->haplotype,bcf_seqname(args->hdr,rec),rec->pos+1);
uint8_t *ignore, *ptr = fmt->p + fmt->size*args->isample + args->haplotype - 1;
ialt = bcf_dec_int1(ptr, fmt->type, &ignore);
if ( bcf_gt_is_missing(ialt) || ialt==bcf_int32_vector_end ) return;
ialt = bcf_gt_allele(ialt);
}
else if ( args->output_iupac )
{
uint8_t *ignore, *ptr = fmt->p + fmt->size*args->isample;
ialt = bcf_dec_int1(ptr, fmt->type, &ignore);
if ( bcf_gt_is_missing(ialt) || ialt==bcf_int32_vector_end ) return;
ialt = bcf_gt_allele(ialt);
int jalt;
if ( fmt->n>1 )
{
ptr = fmt->p + fmt->size*args->isample + 1;
jalt = bcf_dec_int1(ptr, fmt->type, &ignore);
if ( bcf_gt_is_missing(jalt) || jalt==bcf_int32_vector_end ) jalt = ialt;
else jalt = bcf_gt_allele(jalt);
}
else jalt = ialt;
if ( rec->n_allele <= ialt || rec->n_allele <= jalt ) error("Broken VCF, too few alts at %s:%d\n", bcf_seqname(args->hdr,rec),rec->pos+1);
if ( ialt!=jalt && !rec->d.allele[ialt][1] && !rec->d.allele[jalt][1] ) // is this a het snp?
{
char ial = rec->d.allele[ialt][0];
char jal = rec->d.allele[jalt][0];
rec->d.allele[ialt][0] = gt2iupac(ial,jal);
}
}
else
{
for (i=0; i<fmt->n; i++)
{
uint8_t *ignore, *ptr = fmt->p + fmt->size*args->isample + i;
ialt = bcf_dec_int1(ptr, fmt->type, &ignore);
if ( bcf_gt_is_missing(ialt) || ialt==bcf_int32_vector_end ) return;
ialt = bcf_gt_allele(ialt);
if ( ialt ) break;
}
}
if ( !ialt ) return; // ref allele
if ( rec->n_allele <= ialt ) error("Broken VCF, too few alts at %s:%d\n", bcf_seqname(args->hdr,rec),rec->pos+1);
}
else if ( args->output_iupac && !rec->d.allele[0][1] && !rec->d.allele[1][1] )
{
char ial = rec->d.allele[0][0];
char jal = rec->d.allele[1][0];
rec->d.allele[1][0] = gt2iupac(ial,jal);
}
int idx = rec->pos - args->fa_ori_pos + args->fa_mod_off;
if ( idx<0 || idx>=args->fa_buf.l )
error("FIXME: %s:%d .. idx=%d, ori_pos=%d, len=%d, off=%d\n",bcf_seqname(args->hdr,rec),rec->pos+1,idx,args->fa_ori_pos,args->fa_buf.l,args->fa_mod_off);
// sanity check the reference base
int len_diff = 0, alen = 0;
if ( rec->d.allele[ialt][0]=='<' )
{
if ( strcasecmp(rec->d.allele[ialt], "<DEL>") )
error("Symbolic alleles other than <DEL> are currently not supported: %s at %s:%d\n",rec->d.allele[ialt],bcf_seqname(args->hdr,rec),rec->pos+1);
assert( rec->d.allele[0][1]==0 ); // todo: for now expecting strlen(REF) = 1
len_diff = 1-rec->rlen;
rec->d.allele[ialt] = rec->d.allele[0]; // according to VCF spec, REF must precede the event
alen = strlen(rec->d.allele[ialt]);
}
else if ( strncasecmp(rec->d.allele[0],args->fa_buf.s+idx,rec->rlen) )
{
// fprintf(pysamerr,"%d .. [%s], idx=%d ori=%d off=%d\n",args->fa_ori_pos,args->fa_buf.s,idx,args->fa_ori_pos,args->fa_mod_off);
char tmp = 0;
if ( args->fa_buf.l - idx > rec->rlen )
{
tmp = args->fa_buf.s[idx+rec->rlen];
args->fa_buf.s[idx+rec->rlen] = 0;
}
error(
"The fasta sequence does not match the REF allele at %s:%d:\n"
" .vcf: [%s]\n"
" .vcf: [%s] <- (ALT)\n"
" .fa: [%s]%c%s\n",
bcf_seqname(args->hdr,rec),rec->pos+1, rec->d.allele[0], rec->d.allele[ialt], args->fa_buf.s+idx,
tmp?tmp:' ',tmp?args->fa_buf.s+idx+rec->rlen+1:""
);
}
else
{
alen = strlen(rec->d.allele[ialt]);
len_diff = alen - rec->rlen;
}
if ( args->fa_case )
for (i=0; i<alen; i++) rec->d.allele[ialt][i] = toupper(rec->d.allele[ialt][i]);
else
for (i=0; i<alen; i++) rec->d.allele[ialt][i] = tolower(rec->d.allele[ialt][i]);
if ( len_diff <= 0 )
{
// deletion or same size event
for (i=0; i<alen; i++)
args->fa_buf.s[idx+i] = rec->d.allele[ialt][i];
if ( len_diff )
memmove(args->fa_buf.s+idx+alen,args->fa_buf.s+idx+rec->rlen,args->fa_buf.l-idx-rec->rlen);
}
else
{
// insertion
ks_resize(&args->fa_buf, args->fa_buf.l + len_diff);
memmove(args->fa_buf.s + idx + rec->rlen + len_diff, args->fa_buf.s + idx + rec->rlen, args->fa_buf.l - idx - rec->rlen);
for (i=0; i<alen; i++)
args->fa_buf.s[idx+i] = rec->d.allele[ialt][i];
}
if (args->chain && len_diff != 0)
{
// If first nucleotide of both REF and ALT are the same... (indels typically include the nucleotide before the variant)
if ( strncasecmp(rec->d.allele[0],rec->d.allele[ialt],1) == 0)
{
// ...extend the block by 1 bp: start is 1 bp further and alleles are 1 bp shorter
push_chain_gap(args->chain, rec->pos + 1, rec->rlen - 1, rec->pos + 1 + args->fa_mod_off, alen - 1);
}
else
{
// otherwise, just the coordinates of the variant as given
push_chain_gap(args->chain, rec->pos, rec->rlen, rec->pos + args->fa_mod_off, alen);
}
}
args->fa_buf.l += len_diff;
args->fa_mod_off += len_diff;
args->fa_frz_pos = rec->pos + rec->rlen - 1;
}
static void check_gt(args_t *args)
{
int i,ret, *gt2ipl = NULL, m_gt2ipl = 0, *gt_arr = NULL, ngt_arr = 0;
int fake_pls = args->no_PLs;
// Initialize things: check which tags are defined in the header, sample names etc.
if ( bcf_hdr_id2int(args->gt_hdr, BCF_DT_ID, "GT")<0 ) error("[E::%s] GT not present in the header of %s?\n", __func__, args->files->readers[1].fname);
if ( bcf_hdr_id2int(args->sm_hdr, BCF_DT_ID, "PL")<0 )
{
if ( bcf_hdr_id2int(args->sm_hdr, BCF_DT_ID, "GT")<0 )
error("[E::%s] Neither PL nor GT present in the header of %s\n", __func__, args->files->readers[0].fname);
if ( !args->no_PLs )
fprintf(stderr,"Warning: PL not present in the header of %s, using GT instead\n", args->files->readers[0].fname);
fake_pls = 1;
}
FILE *fp = args->plot ? open_file(NULL, "w", "%s.tab", args->plot) : stdout;
print_header(args, fp);
int tgt_isample = -1, query_isample = 0;
if ( args->target_sample )
{
tgt_isample = bcf_hdr_id2int(args->gt_hdr, BCF_DT_SAMPLE, args->target_sample);
if ( tgt_isample<0 ) error("No such sample in %s: [%s]\n", args->files->readers[1].fname, args->target_sample);
}
if ( args->all_sites )
{
if ( tgt_isample==-1 )
{
fprintf(stderr,"No target sample selected for comparison, using the first sample in %s: %s\n", args->gt_fname,args->gt_hdr->samples[0]);
tgt_isample = 0;
}
}
if ( args->query_sample )
{
query_isample = bcf_hdr_id2int(args->sm_hdr, BCF_DT_SAMPLE, args->query_sample);
if ( query_isample<0 ) error("No such sample in %s: [%s]\n", args->files->readers[0].fname, args->query_sample);
}
if ( args->all_sites )
fprintf(fp, "# [1]SC, Site by Site Comparison\t[2]Chromosome\t[3]Position\t[4]-g alleles\t[5]-g GT (%s)\t[6]match log LK\t[7]Query alleles\t[8-]Query PLs (%s)\n",
args->gt_hdr->samples[tgt_isample],args->sm_hdr->samples[query_isample]);
// Main loop
float prev_lk = 0;
while ( (ret=bcf_sr_next_line(args->files)) )
{
if ( ret!=2 ) continue;
bcf1_t *sm_line = args->files->readers[0].buffer[0]; // the query file
bcf1_t *gt_line = args->files->readers[1].buffer[0]; // the -g target file
bcf_unpack(sm_line, BCF_UN_FMT);
bcf_unpack(gt_line, BCF_UN_FMT);
// Init mapping from target genotype index to the sample's PL fields
int n_gt2ipl = gt_line->n_allele*(gt_line->n_allele + 1)/2;
if ( n_gt2ipl > m_gt2ipl )
{
m_gt2ipl = n_gt2ipl;
gt2ipl = (int*) realloc(gt2ipl, sizeof(int)*m_gt2ipl);
}
if ( !init_gt2ipl(args, gt_line, sm_line, gt2ipl, n_gt2ipl) ) continue;
// Target genotypes
int ngt, npl;
if ( (ngt=bcf_get_genotypes(args->gt_hdr, gt_line, >_arr, &ngt_arr)) <= 0 )
error("GT not present at %s:%d?", args->gt_hdr->id[BCF_DT_CTG][gt_line->rid].key, gt_line->pos+1);
ngt /= bcf_hdr_nsamples(args->gt_hdr);
if ( ngt!=2 ) continue; // checking only diploid genotypes
// Sample PLs
if ( !fake_pls )
{
if ( (npl=bcf_get_format_int32(args->sm_hdr, sm_line, "PL", &args->pl_arr, &args->npl_arr)) <= 0 )
error("PL not present at %s:%d?", args->sm_hdr->id[BCF_DT_CTG][sm_line->rid].key, sm_line->pos+1);
npl /= bcf_hdr_nsamples(args->sm_hdr);
}
else
npl = fake_PLs(args, args->sm_hdr, sm_line);
// Calculate likelihoods for all samples, assuming diploid genotypes
// For faster access to genotype likelihoods (PLs) of the query sample
int max_ipl, *pl_ptr = args->pl_arr + query_isample*npl;
double sum_pl = 0; // for converting PLs to probs
for (max_ipl=0; max_ipl<npl; max_ipl++)
{
if ( pl_ptr[max_ipl]==bcf_int32_vector_end ) break;
if ( pl_ptr[max_ipl]==bcf_int32_missing ) continue;
sum_pl += pow(10, -0.1*pl_ptr[max_ipl]);
}
if ( sum_pl==0 ) continue; // no PLs present
if ( fake_pls && args->no_PLs==1 ) sum_pl = -1;
// The main stats: concordance of the query sample with the target -g samples
for (i=0; i<bcf_hdr_nsamples(args->gt_hdr); i++)
{
int *gt_ptr = gt_arr + i*ngt;
if ( gt_ptr[1]==bcf_int32_vector_end ) continue; // skip haploid genotypes
if ( bcf_gt_is_missing(gt_ptr[0]) || bcf_gt_is_missing(gt_ptr[1]) ) continue;
int a = bcf_gt_allele(gt_ptr[0]);
int b = bcf_gt_allele(gt_ptr[1]);
if ( args->hom_only && a!=b ) continue; // heterozygous genotype
int igt_tgt = igt_tgt = bcf_alleles2gt(a,b); // genotype index in the target file
int igt_qry = gt2ipl[igt_tgt]; // corresponding genotype in query file
if ( igt_qry>=max_ipl || pl_ptr[igt_qry]<0 ) continue; // genotype not present in query sample: haploid or missing
args->lks[i] += sum_pl<0 ? -pl_ptr[igt_qry] : log(pow(10, -0.1*pl_ptr[igt_qry])/sum_pl);
args->sites[i]++;
}
if ( args->all_sites )
{
// Print LKs at all sites for debugging
int *gt_ptr = gt_arr + tgt_isample*ngt;
if ( gt_ptr[1]==bcf_int32_vector_end ) continue; // skip haploid genotypes
int a = bcf_gt_allele(gt_ptr[0]);
int b = bcf_gt_allele(gt_ptr[1]);
if ( args->hom_only && a!=b ) continue; // heterozygous genotype
fprintf(fp, "SC\t%s\t%d", args->gt_hdr->id[BCF_DT_CTG][gt_line->rid].key, gt_line->pos+1);
for (i=0; i<gt_line->n_allele; i++) fprintf(fp, "%c%s", i==0?'\t':',', gt_line->d.allele[i]);
fprintf(fp, "\t%s/%s", a>=0 ? gt_line->d.allele[a] : ".", b>=0 ? gt_line->d.allele[b] : ".");
fprintf(fp, "\t%f", args->lks[query_isample]-prev_lk);
prev_lk = args->lks[query_isample];
int igt, *pl_ptr = args->pl_arr + query_isample*npl; // PLs of the query sample
for (i=0; i<sm_line->n_allele; i++) fprintf(fp, "%c%s", i==0?'\t':',', sm_line->d.allele[i]);
for (igt=0; igt<npl; igt++)
if ( pl_ptr[igt]==bcf_int32_vector_end ) break;
else if ( pl_ptr[igt]==bcf_int32_missing ) fprintf(fp, ".");
else fprintf(fp, "\t%d", pl_ptr[igt]);
fprintf(fp, "\n");
}
}
free(gt2ipl);
free(gt_arr);
free(args->pl_arr);
free(args->tmp_arr);
// To be able to plot total discordance (=number of mismatching GTs with -G1) in the same
// plot as discordance per site, the latter must be scaled to the same range
int nsamples = bcf_hdr_nsamples(args->gt_hdr);
double extreme_lk = 0, extreme_lk_per_site = 0;
for (i=0; i<nsamples; i++)
{
if ( args->lks[i] < extreme_lk ) extreme_lk = args->lks[i];
if ( args->sites[i] && args->lks[i]/args->sites[i] < extreme_lk_per_site ) extreme_lk_per_site = args->lks[i]/args->sites[i];
}
// Sorted output
double **p = (double**) malloc(sizeof(double*)*nsamples);
for (i=0; i<nsamples; i++) p[i] = &args->lks[i];
qsort(p, nsamples, sizeof(int*), cmp_doubleptr);
fprintf(fp, "# [1]CN\t[2]Discordance with %s (total)\t[3]Discordance (score per site)\t[4]Number of sites compared\t[5]Sample\t[6]Sample ID\n", args->sm_hdr->samples[query_isample]);
for (i=0; i<nsamples; i++)
{
int idx = p[i] - args->lks;
double per_site = 0;
if ( args->sites[idx] )
{
if ( args->sites[idx] && extreme_lk_per_site )
{
per_site = args->lks[idx]/args->sites[idx];
per_site *= extreme_lk / extreme_lk_per_site;
}
else
per_site = 0;
}
fprintf(fp, "CN\t%e\t%e\t%.0f\t%s\t%d\n", fabs(args->lks[idx]), fabs(per_site), args->sites[idx], args->gt_hdr->samples[idx], i);
}
if ( args->plot )
{
fclose(fp);
plot_check(args, args->target_sample ? args->target_sample : "", args->sm_hdr->samples[query_isample]);
}
}
void bcf_remove_alleles(const bcf_hdr_t *header, bcf1_t *line, int rm_mask)
{
int *map = (int*) calloc(line->n_allele, sizeof(int));
// create map of indexes from old to new ALT numbering and modify ALT
kstring_t str = {0,0,0};
kputs(line->d.allele[0], &str);
int nrm = 0, i,j; // i: ori alleles, j: new alleles
for (i=1, j=1; i<line->n_allele; i++)
{
if ( rm_mask & 1<<i )
{
// remove this allele
line->d.allele[i] = NULL;
nrm++;
continue;
}
kputc(',', &str);
kputs(line->d.allele[i], &str);
map[i] = j;
j++;
}
if ( !nrm ) { free(map); free(str.s); return; }
int nR_ori = line->n_allele;
int nR_new = line->n_allele-nrm;
assert(nR_new > 0); // should not be able to remove reference allele
int nA_ori = nR_ori-1;
int nA_new = nR_new-1;
int nG_ori = nR_ori*(nR_ori + 1)/2;
int nG_new = nR_new*(nR_new + 1)/2;
bcf_update_alleles_str(header, line, str.s);
// remove from Number=G, Number=R and Number=A INFO fields.
uint8_t *dat = NULL;
int mdat = 0, ndat = 0, mdat_bytes = 0, nret;
for (i=0; i<line->n_info; i++)
{
bcf_info_t *info = &line->d.info[i];
int vlen = bcf_hdr_id2length(header,BCF_HL_INFO,info->key);
if ( vlen!=BCF_VL_A && vlen!=BCF_VL_G && vlen!=BCF_VL_R ) continue; // no need to change
int type = bcf_hdr_id2type(header,BCF_HL_INFO,info->key);
if ( type==BCF_HT_FLAG ) continue;
int size = 1;
if ( type==BCF_HT_REAL || type==BCF_HT_INT ) size = 4;
mdat = mdat_bytes / size;
nret = bcf_get_info_values(header, line, bcf_hdr_int2id(header,BCF_DT_ID,info->key), (void**)&dat, &mdat, type);
mdat_bytes = mdat * size;
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not access INFO/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,info->key), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
if ( type==BCF_HT_STR )
{
str.l = 0;
char *ss = (char*) dat, *se = (char*) dat;
if ( vlen==BCF_VL_A || vlen==BCF_VL_R )
{
int nexp, inc = 0;
if ( vlen==BCF_VL_A )
{
nexp = nA_ori;
inc = 1;
}
else
nexp = nR_ori;
for (j=0; j<nexp; j++)
{
if ( !*se ) break;
while ( *se && *se!=',' ) se++;
if ( rm_mask & 1<<(j+inc) )
{
if ( *se ) se++;
ss = se;
continue;
}
if ( str.l ) kputc(',',&str);
kputsn(ss,se-ss,&str);
if ( *se ) se++;
ss = se;
}
assert( j==nexp );
}
else // Number=G, assuming diploid genotype
{
int k = 0, n = 0;
for (j=0; j<nR_ori; j++)
{
for (k=0; k<=j; k++)
{
if ( !*se ) break;
while ( *se && *se!=',' ) se++;
n++;
if ( rm_mask & 1<<j || rm_mask & 1<<k )
{
if ( *se ) se++;
ss = se;
continue;
}
if ( str.l ) kputc(',',&str);
kputsn(ss,se-ss,&str);
if ( *se ) se++;
ss = se;
}
if ( !*se ) break;
}
assert( n=nG_ori );
}
nret = bcf_update_info(header, line, bcf_hdr_int2id(header,BCF_DT_ID,info->key), (void*)str.s, str.l, type);
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not update INFO/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,info->key), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
continue;
}
if ( vlen==BCF_VL_A || vlen==BCF_VL_R )
{
int inc = 0, ntop;
if ( vlen==BCF_VL_A )
{
assert( nret==nA_ori );
ntop = nA_ori;
ndat = nA_new;
inc = 1;
}
else
{
assert( nret==nR_ori );
ntop = nR_ori;
ndat = nR_new;
}
int k = 0;
#define BRANCH(type_t,is_vector_end) \
{ \
type_t *ptr = (type_t*) dat; \
int size = sizeof(type_t); \
for (j=0; j<ntop; j++) /* j:ori, k:new */ \
{ \
if ( is_vector_end ) { memcpy(dat+k*size, dat+j*size, size); break; } \
if ( rm_mask & 1<<(j+inc) ) continue; \
if ( j!=k ) memcpy(dat+k*size, dat+j*size, size); \
k++; \
} \
}
switch (type)
{
case BCF_HT_INT: BRANCH(int32_t,ptr[j]==bcf_int32_vector_end); break;
case BCF_HT_REAL: BRANCH(float,bcf_float_is_vector_end(ptr[j])); break;
}
#undef BRANCH
}
else // Number=G
{
assert( nret==nG_ori );
int k, l_ori = -1, l_new = 0;
ndat = nG_new;
#define BRANCH(type_t,is_vector_end) \
{ \
type_t *ptr = (type_t*) dat; \
int size = sizeof(type_t); \
for (j=0; j<nR_ori; j++) \
{ \
for (k=0; k<=j; k++) \
{ \
l_ori++; \
if ( is_vector_end ) { memcpy(dat+l_new*size, dat+l_ori*size, size); break; } \
if ( rm_mask & 1<<j || rm_mask & 1<<k ) continue; \
if ( l_ori!=l_new ) memcpy(dat+l_new*size, dat+l_ori*size, size); \
l_new++; \
} \
} \
}
switch (type)
{
case BCF_HT_INT: BRANCH(int32_t,ptr[l_ori]==bcf_int32_vector_end); break;
case BCF_HT_REAL: BRANCH(float,bcf_float_is_vector_end(ptr[l_ori])); break;
}
#undef BRANCH
}
nret = bcf_update_info(header, line, bcf_hdr_int2id(header,BCF_DT_ID,info->key), (void*)dat, ndat, type);
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not update INFO/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,info->key), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
}
// Update GT fields, the allele indexes might have changed
for (i=1; i<line->n_allele; i++) if ( map[i]!=i ) break;
if ( i<line->n_allele )
{
mdat = mdat_bytes / 4; // sizeof(int32_t)
nret = bcf_get_genotypes(header,line,(void**)&dat,&mdat);
mdat_bytes = mdat * 4;
if ( nret>0 )
{
nret /= line->n_sample;
int32_t *ptr = (int32_t*) dat;
for (i=0; i<line->n_sample; i++)
{
for (j=0; j<nret; j++)
{
if ( bcf_gt_is_missing(ptr[j]) ) continue;
if ( ptr[j]==bcf_int32_vector_end ) break;
int al = bcf_gt_allele(ptr[j]);
assert( al<nR_ori && map[al]>=0 );
ptr[j] = (map[al]+1)<<1 | (ptr[j]&1);
}
ptr += nret;
}
bcf_update_genotypes(header, line, (void*)dat, nret*line->n_sample);
}
}
// Remove from Number=G, Number=R and Number=A FORMAT fields.
// Assuming haploid or diploid GTs
for (i=0; i<line->n_fmt; i++)
{
bcf_fmt_t *fmt = &line->d.fmt[i];
int vlen = bcf_hdr_id2length(header,BCF_HL_FMT,fmt->id);
if ( vlen!=BCF_VL_A && vlen!=BCF_VL_G && vlen!=BCF_VL_R ) continue; // no need to change
int type = bcf_hdr_id2type(header,BCF_HL_FMT,fmt->id);
if ( type==BCF_HT_FLAG ) continue;
int size = 1;
if ( type==BCF_HT_REAL || type==BCF_HT_INT ) size = 4;
mdat = mdat_bytes / size;
nret = bcf_get_format_values(header, line, bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), (void**)&dat, &mdat, type);
mdat_bytes = mdat * size;
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not access FORMAT/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
if ( type==BCF_HT_STR )
{
int size = nret/line->n_sample; // number of bytes per sample
str.l = 0;
if ( vlen==BCF_VL_A || vlen==BCF_VL_R )
{
int nexp, inc = 0;
if ( vlen==BCF_VL_A )
{
nexp = nA_ori;
inc = 1;
}
else
nexp = nR_ori;
for (j=0; j<line->n_sample; j++)
{
char *ss = ((char*)dat) + j*size, *se = ss + size, *ptr = ss;
int k_src = 0, k_dst = 0, l = str.l;
for (k_src=0; k_src<nexp; k_src++)
{
if ( ptr>=se || !*ptr) break;
while ( ptr<se && *ptr && *ptr!=',' ) ptr++;
if ( rm_mask & 1<<(k_src+inc) )
{
ss = ++ptr;
continue;
}
if ( k_dst ) kputc(',',&str);
kputsn(ss,ptr-ss,&str);
ss = ++ptr;
k_dst++;
}
assert( k_src==nexp );
l = str.l - l;
for (; l<size; l++) kputc(0, &str);
}
}
else // Number=G, diploid or haploid
{
for (j=0; j<line->n_sample; j++)
{
char *ss = ((char*)dat) + j*size, *se = ss + size, *ptr = ss;
int k_src = 0, k_dst = 0, l = str.l;
int nexp = 0; // diploid or haploid?
while ( ptr<se )
{
if ( !*ptr ) break;
if ( *ptr==',' ) nexp++;
ptr++;
}
if ( ptr!=ss ) nexp++;
assert( nexp==nG_ori || nexp==nR_ori );
ptr = ss;
if ( nexp==nG_ori ) // diploid
{
int ia, ib;
for (ia=0; ia<nR_ori; ia++)
{
for (ib=0; ib<=ia; ib++)
{
if ( ptr>=se || !*ptr ) break;
while ( ptr<se && *ptr && *ptr!=',' ) ptr++;
if ( rm_mask & 1<<ia || rm_mask & 1<<ib )
{
ss = ++ptr;
continue;
}
if ( k_dst ) kputc(',',&str);
kputsn(ss,ptr-ss,&str);
ss = ++ptr;
k_dst++;
}
if ( ptr>=se || !*ptr ) break;
}
}
else // haploid
{
for (k_src=0; k_src<nR_ori; k_src++)
{
if ( ptr>=se || !*ptr ) break;
while ( ptr<se && *ptr && *ptr!=',' ) ptr++;
if ( rm_mask & 1<<k_src )
{
ss = ++ptr;
continue;
}
if ( k_dst ) kputc(',',&str);
kputsn(ss,ptr-ss,&str);
ss = ++ptr;
k_dst++;
}
assert( k_src==nR_ori );
l = str.l - l;
for (; l<size; l++) kputc(0, &str);
}
}
}
nret = bcf_update_format(header, line, bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), (void*)str.s, str.l, type);
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not update FORMAT/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
continue;
}
int nori = nret / line->n_sample;
if ( vlen==BCF_VL_A || vlen==BCF_VL_R || (vlen==BCF_VL_G && nori==nR_ori) ) // Number=A, R or haploid Number=G
{
int inc = 0, nnew;
if ( vlen==BCF_VL_A )
{
assert( nori==nA_ori ); // todo: will fail if all values are missing
ndat = nA_new*line->n_sample;
nnew = nA_new;
inc = 1;
}
else
{
assert( nori==nR_ori ); // todo: will fail if all values are missing
ndat = nR_new*line->n_sample;
nnew = nR_new;
}
#define BRANCH(type_t,is_vector_end) \
{ \
for (j=0; j<line->n_sample; j++) \
{ \
type_t *ptr_src = ((type_t*)dat) + j*nori; \
type_t *ptr_dst = ((type_t*)dat) + j*nnew; \
int size = sizeof(type_t); \
int k_src, k_dst = 0; \
for (k_src=0; k_src<nori; k_src++) \
{ \
if ( is_vector_end ) { memcpy(ptr_dst+k_dst, ptr_src+k_src, size); break; } \
if ( rm_mask & 1<<(k_src+inc) ) continue; \
memcpy(ptr_dst+k_dst, ptr_src+k_src, size); \
k_dst++; \
} \
} \
}
switch (type)
{
case BCF_HT_INT: BRANCH(int32_t,ptr_src[k_src]==bcf_int32_vector_end); break;
case BCF_HT_REAL: BRANCH(float,bcf_float_is_vector_end(ptr_src[k_src])); break;
}
#undef BRANCH
}
else // Number=G, diploid or mixture of haploid+diploid
{
assert( nori==nG_ori );
ndat = nG_new*line->n_sample;
#define BRANCH(type_t,is_vector_end) \
{ \
for (j=0; j<line->n_sample; j++) \
{ \
type_t *ptr_src = ((type_t*)dat) + j*nori; \
type_t *ptr_dst = ((type_t*)dat) + j*nG_new; \
int size = sizeof(type_t); \
int ia, ib, k_dst = 0, k_src; \
int nset = 0; /* haploid or diploid? */ \
for (k_src=0; k_src<nG_ori; k_src++) { if ( is_vector_end ) break; nset++; } \
if ( nset==nR_ori ) /* haploid */ \
{ \
for (k_src=0; k_src<nR_ori; k_src++) \
{ \
if ( rm_mask & 1<<k_src ) continue; \
memcpy(ptr_dst+k_dst, ptr_src+k_src, size); \
k_dst++; \
} \
memcpy(ptr_dst+k_dst, ptr_src+k_src, size); \
} \
else /* diploid */ \
{ \
k_src = -1; \
for (ia=0; ia<nR_ori; ia++) \
{ \
for (ib=0; ib<=ia; ib++) \
{ \
k_src++; \
if ( is_vector_end ) { memcpy(ptr_dst+k_dst, ptr_src+k_src, size); ia = nR_ori; break; } \
if ( rm_mask & 1<<ia || rm_mask & 1<<ib ) continue; \
memcpy(ptr_dst+k_dst, ptr_src+k_src, size); \
k_dst++; \
} \
} \
} \
} \
}
switch (type)
{
case BCF_HT_INT: BRANCH(int32_t,ptr_src[k_src]==bcf_int32_vector_end); break;
case BCF_HT_REAL: BRANCH(float,bcf_float_is_vector_end(ptr_src[k_src])); break;
}
#undef BRANCH
}
nret = bcf_update_format(header, line, bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), (void*)dat, ndat, type);
if ( nret<0 )
{
fprintf(stderr,"[%s:%d %s] Could not update FORMAT/%s at %s:%d [%d]\n", __FILE__,__LINE__,__FUNCTION__,
bcf_hdr_int2id(header,BCF_DT_ID,fmt->id), bcf_seqname(header,line), line->pos+1, nret);
exit(1);
}
}
free(dat);
free(str.s);
free(map);
}
bcf1_t *process(bcf1_t *rec)
{
bcf1_t *dflt = args.mode&MODE_LIST_GOOD ? rec : NULL;
args.nrec++;
if ( rec->n_allele > 63 ) return dflt; // we use 64bit bitmask below
int ngt = bcf_get_genotypes(args.hdr, rec, &args.gt_arr, &args.ngt_arr);
if ( ngt<0 ) return dflt;
if ( ngt!=2*bcf_hdr_nsamples(args.hdr) && ngt!=bcf_hdr_nsamples(args.hdr) ) return dflt;
ngt /= bcf_hdr_nsamples(args.hdr);
int itr_set = regidx_overlap(args.rules, bcf_seqname(args.hdr,rec),rec->pos,rec->pos, args.itr_ori);
int i, has_bad = 0, needs_update = 0;
for (i=0; i<args.ntrios; i++)
{
int32_t a,b,c,d,e,f;
trio_t *trio = &args.trios[i];
a = args.gt_arr[ngt*trio->imother];
b = ngt==2 ? args.gt_arr[ngt*trio->imother+1] : bcf_int32_vector_end;
c = args.gt_arr[ngt*trio->ifather];
d = ngt==2 ? args.gt_arr[ngt*trio->ifather+1] : bcf_int32_vector_end;
e = args.gt_arr[ngt*trio->ichild];
f = ngt==2 ? args.gt_arr[ngt*trio->ichild+1] : bcf_int32_vector_end;
// skip sites with missing data in child
if ( bcf_gt_is_missing(e) || bcf_gt_is_missing(f) ) continue;
uint64_t mother = 0, father = 0,child1,child2;
int is_ok = 0;
if ( !itr_set )
{
if ( f==bcf_int32_vector_end ) { warn_ploidy(rec); continue; }
// All M,F,C genotypes are diploid. Missing data are considered consistent.
child1 = 1<<bcf_gt_allele(e);
child2 = 1<<bcf_gt_allele(f);
mother = bcf_gt_is_missing(a) ? child1|child2 : 1<<bcf_gt_allele(a);
mother |= bcf_gt_is_missing(b) || b==bcf_int32_vector_end ? child1|child2 : 1<<bcf_gt_allele(b);
father = bcf_gt_is_missing(c) ? child1|child2 : 1<<bcf_gt_allele(c);
father |= bcf_gt_is_missing(d) || d==bcf_int32_vector_end ? child1|child2 : 1<<bcf_gt_allele(d);
if ( (mother&child1 && father&child2) || (mother&child2 && father&child1) ) is_ok = 1;
}
else
{
child1 = 1<<bcf_gt_allele(e);
child2 = bcf_gt_is_missing(f) || f==bcf_int32_vector_end ? 0 : 1<<bcf_gt_allele(f);
mother |= bcf_gt_is_missing(a) ? 0 : 1<<bcf_gt_allele(a);
mother |= bcf_gt_is_missing(b) || b==bcf_int32_vector_end ? 0 : 1<<bcf_gt_allele(b);
father |= bcf_gt_is_missing(c) ? 0 : 1<<bcf_gt_allele(c);
father |= bcf_gt_is_missing(d) || d==bcf_int32_vector_end ? 0 : 1<<bcf_gt_allele(d);
regitr_copy(args.itr, args.itr_ori);
while ( !is_ok && regitr_overlap(args.itr) )
{
rule_t *rule = ®itr_payload(args.itr,rule_t);
if ( child1 && child2 )
{
if ( !rule->mal || !rule->fal ) continue; // wrong rule (haploid), but this is a diploid GT
if ( !mother ) mother = child1|child2;
if ( !father ) father = child1|child2;
if ( (mother&child1 && father&child2) || (mother&child2 && father&child1) ) is_ok = 1;
continue;
}
if ( rule->mal )
{
if ( mother && !(child1&mother) ) continue;
}
if ( rule->fal )
{
if ( father && !(child1&father) ) continue;
}
is_ok = 1;
}
}
if ( is_ok )
{
trio->nok++;
}
else
{
trio->nbad++;
has_bad = 1;
if ( args.mode&MODE_DELETE )
{
args.gt_arr[ngt*trio->imother] = bcf_gt_missing;
if ( b!=bcf_int32_vector_end ) args.gt_arr[ngt*trio->imother+1] = bcf_gt_missing; // should be always true
args.gt_arr[ngt*trio->ifather] = bcf_gt_missing;
if ( d!=bcf_int32_vector_end ) args.gt_arr[ngt*trio->ifather+1] = bcf_gt_missing;
args.gt_arr[ngt*trio->ichild] = bcf_gt_missing;
if ( f!=bcf_int32_vector_end ) args.gt_arr[ngt*trio->ichild+1] = bcf_gt_missing;
needs_update = 1;
}
}
}
if ( needs_update && bcf_update_genotypes(args.hdr,rec,args.gt_arr,ngt*bcf_hdr_nsamples(args.hdr)) )
error("Could not update GT field at %s:%d\n", bcf_seqname(args.hdr,rec),rec->pos+1);
if ( args.mode&MODE_DELETE ) return rec;
if ( args.mode&MODE_LIST_GOOD ) return has_bad ? NULL : rec;
if ( args.mode&MODE_LIST_BAD ) return has_bad ? rec : NULL;
return NULL;
}
int parse_line(args_t *args, bcf1_t *line, double *alt_freq, double *pdg)
{
args->nitmp = 0;
// Set allele frequency
int ret;
if ( args->af_tag )
{
// Use an INFO tag provided by the user
ret = bcf_get_info_float(args->hdr, line, args->af_tag, &args->AFs, &args->mAFs);
if ( ret==1 )
*alt_freq = args->AFs[0];
if ( ret==-2 )
error("Type mismatch for INFO/%s tag at %s:%d\n", args->af_tag, bcf_seqname(args->hdr,line), line->pos+1);
}
else if ( args->af_fname )
{
// Read AF from a file
ret = read_AF(args->files->targets, line, alt_freq);
}
else
{
// Use GTs or AC/AN: GTs when AC/AN not present or when GTs explicitly requested by --estimate-AF
ret = -1;
if ( !args->estimate_AF )
{
int AC = -1, AN = 0;
ret = bcf_get_info_int32(args->hdr, line, "AN", &args->itmp, &args->mitmp);
if ( ret==1 )
{
AN = args->itmp[0];
ret = bcf_get_info_int32(args->hdr, line, "AC", &args->itmp, &args->mitmp);
if ( ret>0 )
AC = args->itmp[0];
}
if ( AN<=0 || AC<0 )
ret = -1;
else
*alt_freq = (double) AC/AN;
}
if ( ret==-1 )
ret = estimate_AF(args, line, alt_freq); // reads GTs into args->itmp
}
if ( ret<0 ) return ret;
if ( *alt_freq==0.0 )
{
if ( args->dflt_AF==0 ) return -1; // we skip sites with AF=0
*alt_freq = args->dflt_AF;
}
// Set P(D|G)
if ( args->fake_PLs )
{
if ( !args->nitmp )
{
args->nitmp = bcf_get_genotypes(args->hdr, line, &args->itmp, &args->mitmp);
if ( args->nitmp != 2*args->nsmpl ) return -1; // not diploid?
args->nitmp /= args->nsmpl;
}
int32_t *gt = &args->itmp[args->ismpl*args->nitmp];
if ( bcf_gt_is_missing(gt[0]) || bcf_gt_is_missing(gt[1]) ) return -1;
int a = bcf_gt_allele(gt[0]);
int b = bcf_gt_allele(gt[1]);
if ( a!=b )
{
pdg[0] = pdg[2] = args->unseen_PL;
pdg[1] = 1 - 2*args->unseen_PL;
}
else if ( a==0 )
{
pdg[0] = 1 - 2*args->unseen_PL;
pdg[1] = pdg[2] = args->unseen_PL;
}
else
{
pdg[0] = pdg[1] = args->unseen_PL;
pdg[2] = 1 - 2*args->unseen_PL;
}
}
else
{
args->nitmp = bcf_get_format_int32(args->hdr, line, "PL", &args->itmp, &args->mitmp);
if ( args->nitmp != args->nsmpl*line->n_allele*(line->n_allele+1)/2. ) return -1; // not diploid?
args->nitmp /= args->nsmpl;
int32_t *pl = &args->itmp[args->ismpl*args->nitmp];
pdg[0] = pl[0] < 256 ? args->pl2p[ pl[0] ] : 1.0;
pdg[1] = pl[1] < 256 ? args->pl2p[ pl[1] ] : 1.0;
pdg[2] = pl[2] < 256 ? args->pl2p[ pl[2] ] : 1.0;
double sum = pdg[0] + pdg[1] + pdg[2];
if ( !sum ) return -1;
pdg[0] /= sum;
pdg[1] /= sum;
pdg[2] /= sum;
}
return 0;
}
