// Add a new region into a list sorted by start,end. On input the coordinates
// are 1-based, stored 0-based, inclusive.
static void _regions_add(bcf_sr_regions_t *reg, const char *chr, int start, int end)
{
if ( start==-1 && end==-1 )
{
start = 0; end = MAX_CSI_COOR-1;
}
else
{
start--; end--; // store 0-based coordinates
}
if ( !reg->seq_hash )
reg->seq_hash = khash_str2int_init();
int iseq;
if ( khash_str2int_get(reg->seq_hash, chr, &iseq)<0 )
{
// the chromosome block does not exist
iseq = reg->nseqs++;
reg->seq_names = (char**) realloc(reg->seq_names,sizeof(char*)*reg->nseqs);
reg->regs = (region_t*) realloc(reg->regs,sizeof(region_t)*reg->nseqs);
memset(®->regs[reg->nseqs-1],0,sizeof(region_t));
reg->seq_names[iseq] = strdup(chr);
reg->regs[iseq].creg = -1;
khash_str2int_set(reg->seq_hash,reg->seq_names[iseq],iseq);
}
region_t *creg = ®->regs[iseq];
// the regions may not be sorted on input: binary search
int i, min = 0, max = creg->nregs - 1;
while ( min<=max )
{
i = (max+min)/2;
if ( start < creg->regs[i].start ) max = i - 1;
else if ( start > creg->regs[i].start ) min = i + 1;
else break;
}
if ( min>max || creg->regs[i].start!=start || creg->regs[i].end!=end )
{
// no such region, insert a new one just after max
hts_expand(region1_t,creg->nregs+1,creg->mregs,creg->regs);
if ( ++max < creg->nregs )
memmove(&creg->regs[max+1],&creg->regs[max],(creg->nregs - max)*sizeof(region1_t));
creg->regs[max].start = start;
creg->regs[max].end = end;
creg->nregs++;
}
}
static void bsmpl_add_readgroup(bam_smpl_t *bsmpl, file_t *file, const char *rg_id, const char *smpl_name)
{
int ismpl = -1;
if ( smpl_name )
{
if ( khash_str2int_get(bsmpl->name2idx,smpl_name,&ismpl) < 0 )
{
// new sample
bsmpl->nsmpl++;
bsmpl->smpl = (char**) realloc(bsmpl->smpl,sizeof(char*)*bsmpl->nsmpl);
bsmpl->smpl[bsmpl->nsmpl-1] = strdup(smpl_name);
ismpl = khash_str2int_inc(bsmpl->name2idx,bsmpl->smpl[bsmpl->nsmpl-1]);
}
}
if ( !strcmp("*",rg_id) )
{
// all read groups in the bam treated as the same sample
file->default_idx = ismpl;
return;
}
if ( !file->rg2idx ) file->rg2idx = khash_str2int_init();
if ( khash_str2int_has_key(file->rg2idx,rg_id) ) return; // duplicate @RG:ID
khash_str2int_set(file->rg2idx, strdup(rg_id), ismpl);
}
static void bcf_sr_sort_set(bcf_srs_t *readers, sr_sort_t *srt, const char *chr, int min_pos)
{
if ( !srt->grp_str2int )
{
// first time here, initialize
if ( !srt->pair )
{
if ( readers->collapse==COLLAPSE_NONE ) readers->collapse = BCF_SR_PAIR_EXACT;
bcf_sr_set_opt(readers, BCF_SR_PAIR_LOGIC, readers->collapse);
}
bcf_sr_init_scores(srt);
srt->grp_str2int = khash_str2int_init();
srt->var_str2int = khash_str2int_init();
}
int k;
khash_t(str2int) *hash;
hash = srt->grp_str2int;
for (k=0; k < kh_end(hash); k++)
if ( kh_exist(hash,k) ) free((char*)kh_key(hash,k));
hash = srt->var_str2int;
for (k=0; k < kh_end(hash); k++)
if ( kh_exist(hash,k) ) free((char*)kh_key(hash,k));
kh_clear(str2int, srt->grp_str2int);
kh_clear(str2int, srt->var_str2int);
srt->ngrp = srt->nvar = srt->nvset = 0;
grp_t grp;
memset(&grp,0,sizeof(grp_t));
// group VCFs into groups, each with a unique combination of variants in the duplicate lines
int ireader,ivar,irec,igrp,ivset,iact;
for (ireader=0; ireader<readers->nreaders; ireader++) srt->vcf_buf[ireader].nrec = 0;
for (iact=0; iact<srt->nactive; iact++)
{
ireader = srt->active[iact];
bcf_sr_t *reader = &readers->readers[ireader];
int rid = bcf_hdr_name2id(reader->header, chr);
grp.nvar = 0;
hts_expand(int,reader->nbuffer,srt->moff,srt->off);
srt->noff = 0;
srt->str.l = 0;
for (irec=1; irec<=reader->nbuffer; irec++)
{
bcf1_t *line = reader->buffer[irec];
if ( line->rid!=rid || line->pos!=min_pos ) break;
if ( srt->str.l ) kputc(';',&srt->str);
srt->off[srt->noff++] = srt->str.l;
size_t beg = srt->str.l;
for (ivar=1; ivar<line->n_allele; ivar++)
{
if ( ivar>1 ) kputc(',',&srt->str);
kputs(line->d.allele[0],&srt->str);
kputc('>',&srt->str);
kputs(line->d.allele[ivar],&srt->str);
}
if ( line->n_allele==1 )
{
kputs(line->d.allele[0],&srt->str);
kputsn(">.",2,&srt->str);
}
// Create new variant or attach to existing one. But careful, there can be duplicate
// records with the same POS,REF,ALT (e.g. in dbSNP-b142)
char *var_str = beg + srt->str.s;
int ret, var_idx = 0, var_end = srt->str.l;
while ( 1 )
{
ret = khash_str2int_get(srt->var_str2int, var_str, &ivar);
if ( ret==-1 ) break;
var_t *var = &srt->var[ivar];
if ( var->vcf[var->nvcf-1] != ireader ) break;
srt->str.l = var_end;
kputw(var_idx, &srt->str);
var_str = beg + srt->str.s;
var_idx++;
}
if ( ret==-1 )
{
ivar = srt->nvar++;
hts_expand0(var_t,srt->nvar,srt->mvar,srt->var);
srt->var[ivar].nvcf = 0;
khash_str2int_set(srt->var_str2int, strdup(var_str), ivar);
free(srt->var[ivar].str); // possible left-over from the previous position
}
var_t *var = &srt->var[ivar];
var->nalt = line->n_allele - 1;
var->type = bcf_get_variant_types(line);
srt->str.s[var_end] = 0;
if ( ret==-1 )
var->str = strdup(var_str);
int mvcf = var->mvcf;
var->nvcf++;
hts_expand0(int*, var->nvcf, var->mvcf, var->vcf);
if ( mvcf != var->mvcf ) var->rec = (bcf1_t **) realloc(var->rec,sizeof(bcf1_t*)*var->mvcf);
var->vcf[var->nvcf-1] = ireader;
var->rec[var->nvcf-1] = line;
grp.nvar++;
hts_expand(var_t,grp.nvar,grp.mvar,grp.var);
grp.var[grp.nvar-1] = ivar;
}
char *grp_key = grp_create_key(srt);
int ret = khash_str2int_get(srt->grp_str2int, grp_key, &igrp);
if ( ret==-1 )
{
igrp = srt->ngrp++;
hts_expand0(grp_t, srt->ngrp, srt->mgrp, srt->grp);
free(srt->grp[igrp].var);
srt->grp[igrp] = grp;
srt->grp[igrp].key = grp_key;
khash_str2int_set(srt->grp_str2int, grp_key, igrp);
memset(&grp,0,sizeof(grp_t));
}
else
free(grp_key);
srt->grp[igrp].nvcf++;
}
free(grp.var);
// initialize bitmask - which groups is the variant present in
for (ivar=0; ivar<srt->nvar; ivar++)
{
srt->var[ivar].mask = kbs_resize(srt->var[ivar].mask, srt->ngrp);
kbs_clear(srt->var[ivar].mask);
}
for (igrp=0; igrp<srt->ngrp; igrp++)
{
for (ivar=0; ivar<srt->grp[igrp].nvar; ivar++)
{
int i = srt->grp[igrp].var[ivar];
kbs_insert(srt->var[i].mask, igrp);
}
}
// create the initial list of variant sets
for (ivar=0; ivar<srt->nvar; ivar++)
{
ivset = srt->nvset++;
hts_expand0(varset_t, srt->nvset, srt->mvset, srt->vset);
varset_t *vset = &srt->vset[ivset];
vset->nvar = 1;
hts_expand0(var_t, vset->nvar, vset->mvar, vset->var);
vset->var[vset->nvar-1] = ivar;
var_t *var = &srt->var[ivar];
vset->cnt = var->nvcf;
vset->mask = kbs_resize(vset->mask, srt->ngrp);
kbs_clear(vset->mask);
kbs_bitwise_or(vset->mask, var->mask);
int type = 0;
if ( var->type==VCF_REF ) type |= SR_REF;
else
{
if ( var->type & VCF_SNP ) type |= SR_SNP;
if ( var->type & VCF_MNP ) type |= SR_SNP;
if ( var->type & VCF_INDEL ) type |= SR_INDEL;
if ( var->type & VCF_OTHER ) type |= SR_OTHER;
}
var->type = type;
}
#if DEBUG_VSETS
debug_vsets(srt);
#endif
// initialize the pairing matrix
hts_expand(int, srt->ngrp*srt->nvset, srt->mpmat, srt->pmat);
hts_expand(int, srt->nvset, srt->mcnt, srt->cnt);
memset(srt->pmat, 0, sizeof(*srt->pmat)*srt->ngrp*srt->nvset);
for (ivset=0; ivset<srt->nvset; ivset++)
{
varset_t *vset = &srt->vset[ivset];
for (igrp=0; igrp<srt->ngrp; igrp++) srt->pmat[ivset*srt->ngrp+igrp] = 0;
srt->cnt[ivset] = vset->cnt;
}
// pair the lines
while ( srt->nvset )
{
#if DEBUG_VSETS
fprintf(stderr,"\n");
debug_vsets(srt);
#endif
int imax = 0;
for (ivset=1; ivset<srt->nvset; ivset++)
if ( srt->cnt[imax] < srt->cnt[ivset] ) imax = ivset;
int ipair = -1;
uint32_t max_score = 0;
for (ivset=0; ivset<srt->nvset; ivset++)
{
if ( kbs_logical_and(srt->vset[imax].mask,srt->vset[ivset].mask) ) continue; // cannot be merged
uint32_t score = pairing_score(srt, imax, ivset);
// fprintf(stderr,"score: %d %d, logic=%d \t..\t %u\n", imax,ivset,srt->pair,score);
if ( max_score < score ) { max_score = score; ipair = ivset; }
}
// merge rows creating a new variant set this way
if ( ipair!=-1 && ipair!=imax )
{
imax = merge_vsets(srt, imax, ipair);
continue;
}
push_vset(srt, imax);
}
srt->chr = chr;
srt->pos = min_pos;
}
bcf_sr_regions_t *bcf_sr_regions_init(const char *regions, int is_file, int ichr, int ifrom, int ito)
{
bcf_sr_regions_t *reg;
if ( !is_file ) return _regions_init_string(regions);
reg = (bcf_sr_regions_t *) calloc(1, sizeof(bcf_sr_regions_t));
reg->start = reg->end = -1;
reg->prev_start = reg->prev_seq = -1;
reg->file = hts_open(regions, "rb");
if ( !reg->file )
{
fprintf(stderr,"[%s:%d %s] Could not open file: %s\n", __FILE__,__LINE__,__FUNCTION__,regions);
free(reg);
return NULL;
}
reg->tbx = tbx_index_load(regions);
if ( !reg->tbx )
{
int len = strlen(regions);
int is_bed = strcasecmp(".bed",regions+len-4) ? 0 : 1;
if ( !is_bed && !strcasecmp(".bed.gz",regions+len-7) ) is_bed = 1;
if ( reg->file->format.format==vcf ) ito = 1;
// read the whole file, tabix index is not present
while ( hts_getline(reg->file, KS_SEP_LINE, ®->line) > 0 )
{
char *chr, *chr_end;
int from, to, ret;
ret = _regions_parse_line(reg->line.s, ichr,ifrom,abs(ito), &chr,&chr_end,&from,&to);
if ( ret < 0 )
{
if ( ito<0 )
ret = _regions_parse_line(reg->line.s, ichr,ifrom,ifrom, &chr,&chr_end,&from,&to);
if ( ret<0 )
{
fprintf(stderr,"[%s:%d] Could not parse the file %s, using the columns %d,%d[,%d]\n", __FILE__,__LINE__,regions,ichr+1,ifrom+1,ito+1);
hts_close(reg->file); reg->file = NULL; free(reg);
return NULL;
}
}
if ( !ret ) continue;
if ( is_bed ) from++;
*chr_end = 0;
_regions_add(reg, chr, from, to);
*chr_end = '\t';
}
hts_close(reg->file); reg->file = NULL;
if ( !reg->nseqs ) { free(reg); return NULL; }
return reg;
}
reg->seq_names = (char**) tbx_seqnames(reg->tbx, ®->nseqs);
if ( !reg->seq_hash )
reg->seq_hash = khash_str2int_init();
int i;
for (i=0; i<reg->nseqs; i++)
{
khash_str2int_set(reg->seq_hash,reg->seq_names[i],i);
}
reg->fname = strdup(regions);
reg->is_bin = 1;
return reg;
}
