static int _reader_seek(bcf_sr_t *reader, const char *seq, int start, int end)
{
if ( end>=MAX_CSI_COOR )
{
fprintf(stderr,"The coordinate is out of csi index limit: %d\n", end+1);
exit(1);
}
if ( reader->itr )
{
hts_itr_destroy(reader->itr);
reader->itr = NULL;
}
reader->nbuffer = 0;
if ( reader->tbx_idx )
{
int tid = tbx_name2id(reader->tbx_idx, seq);
if ( tid==-1 ) return -1; // the sequence not present in this file
reader->itr = tbx_itr_queryi(reader->tbx_idx,tid,start,end+1);
}
else
{
int tid = bcf_hdr_name2id(reader->header, seq);
if ( tid==-1 ) return -1; // the sequence not present in this file
reader->itr = bcf_itr_queryi(reader->bcf_idx,tid,start,end+1);
}
if ( !reader->itr ) fprintf(stderr,"Could not seek: %s:%d-%d\n",seq,start+1,end+1);
assert(reader->itr);
return 0;
}
static void phased_push(args_t *args, bcf1_t *arec, bcf1_t *brec)
{
if ( arec && arec->errcode )
error("Parse error at %s:%d, cannot proceed: %s\n", bcf_seqname(args->files->readers[0].header,arec),arec->pos+1, args->files->readers[0].fname);
if ( brec && brec->errcode )
error("Parse error at %s:%d, cannot proceed: %s\n", bcf_seqname(args->files->readers[1].header,brec),brec->pos+1, args->files->readers[1].fname);
int i, nsmpl = bcf_hdr_nsamples(args->out_hdr);
int chr_id = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[0].header,arec));
if ( args->prev_chr<0 || args->prev_chr!=chr_id )
{
if ( args->prev_chr>=0 ) phased_flush(args);
for (i=0; i<nsmpl; i++)
args->phase_set[i] = arec->pos+1;
args->phase_set_changed = 1;
if ( args->seen_seq[chr_id] ) error("The chromosome block %s is not contiguous\n", bcf_seqname(args->files->readers[0].header,arec));
args->seen_seq[chr_id] = 1;
args->prev_chr = chr_id;
args->prev_pos_check = -1;
}
if ( !brec )
{
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 in %s vs %d written [3]\n", bcf_seqname(args->files->readers[0].header,arec), arec->pos+1,args->files->readers[0].fname, args->prev_pos_check+1);
args->prev_pos_check = arec->pos;
return;
}
int m = args->mbuf;
args->nbuf += 2;
hts_expand(bcf1_t*,args->nbuf,args->mbuf,args->buf);
for (i=m; i<args->mbuf; i++)
args->buf[i] = bcf_init1();
SWAP(bcf1_t*, args->files->readers[0].buffer[0], args->buf[args->nbuf-2]);
SWAP(bcf1_t*, args->files->readers[1].buffer[0], args->buf[args->nbuf-1]);
}
static void init_region(args_t *args, char *line)
{
char *ss, *se = line;
while ( *se && !isspace(*se) && *se!=':' ) se++;
int from = 0, to = 0;
char tmp, *tmp_ptr = NULL;
if ( *se )
{
tmp = *se; *se = 0; tmp_ptr = se;
ss = ++se;
from = strtol(ss,&se,10);
if ( ss==se || !*se || *se!='-' ) from = 0;
else
{
from--;
ss = ++se;
to = strtol(ss,&se,10);
if ( ss==se || (*se && !isspace(*se)) ) { from = 0; to = 0; }
else to--;
}
}
args->rid = bcf_hdr_name2id(args->hdr,line);
if ( args->rid<0 ) fprintf(pysamerr,"Warning: Sequence \"%s\" not in %s\n", line,args->fname);
args->fa_buf.l = 0;
args->fa_length = 0;
args->fa_end_pos = to;
args->fa_ori_pos = from;
args->fa_src_pos = from;
args->fa_mod_off = 0;
args->fa_frz_pos = -1;
args->fa_case = -1;
args->vcf_rbuf.n = 0;
bcf_sr_seek(args->files,line,args->fa_ori_pos);
if ( tmp_ptr ) *tmp_ptr = tmp;
fprintf(args->fp_out,">%s\n",line);
if (args->chain_fname )
{
args->chain = init_chain(args->chain, args->fa_ori_pos);
} else {
args->chain = NULL;
}
}
static void print_missed_line(bcf_sr_regions_t *regs, void *data)
{
args_t *args = (args_t*) data;
call_t *call = &args->aux;
bcf1_t *missed = args->missed_line;
if ( args->flag & CF_GVCF ) error("todo: Combine --gvcf and --insert-missed\n");
char *ss = regs->line.s;
int i = 0;
while ( i<args->aux.srs->targets_als-1 && *ss )
{
if ( *ss=='\t' ) i++;
ss++;
}
if ( !*ss ) error("Could not parse: [%s] (%d)\n", regs->line.s,args->aux.srs->targets_als);
missed->rid = bcf_hdr_name2id(call->hdr,regs->seq_names[regs->prev_seq]);
missed->pos = regs->start;
bcf_update_alleles_str(call->hdr, missed,ss);
bcf_write1(args->out_fh, call->hdr, missed);
}
static void concat(args_t *args)
{
int i;
if ( args->phased_concat ) // phased concat
{
// keep only two open files at a time
while ( args->ifname < args->nfnames )
{
int new_file = 0;
while ( args->files->nreaders < 2 && args->ifname < args->nfnames )
{
if ( !bcf_sr_add_reader(args->files,args->fnames[args->ifname]) ) error("Failed to open %s: %s\n", args->fnames[args->ifname],bcf_sr_strerror(args->files->errnum));
new_file = 1;
args->ifname++;
if ( args->start_pos[args->ifname-1]==-1 ) break; // new chromosome, start with only one file open
if ( args->ifname < args->nfnames && args->start_pos[args->ifname]==-1 ) break; // next file starts on a different chromosome
}
// is there a line from the previous run? Seek the newly opened reader to that position
int seek_pos = -1;
int seek_chr = -1;
if ( bcf_sr_has_line(args->files,0) )
{
bcf1_t *line = bcf_sr_get_line(args->files,0);
bcf_sr_seek(args->files, bcf_seqname(args->files->readers[0].header,line), line->pos);
seek_pos = line->pos;
seek_chr = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[0].header,line));
}
else if ( new_file )
bcf_sr_seek(args->files,NULL,0); // set to start
int nret;
while ( (nret = bcf_sr_next_line(args->files)) )
{
if ( !bcf_sr_has_line(args->files,0) ) // no input from the first reader
{
// We are assuming that there is a perfect overlap, sites which are not present in both files are dropped
if ( ! bcf_sr_region_done(args->files,0) ) continue;
phased_flush(args);
bcf_sr_remove_reader(args->files, 0);
}
// Get a line to learn about current position
for (i=0; i<args->files->nreaders; i++)
if ( bcf_sr_has_line(args->files,i) ) break;
bcf1_t *line = bcf_sr_get_line(args->files,i);
// This can happen after bcf_sr_seek: indel may start before the coordinate which we seek to.
if ( seek_chr>=0 && seek_pos>line->pos && seek_chr==bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[i].header,line)) ) continue;
seek_pos = seek_chr = -1;
// Check if the position overlaps with the next, yet unopened, reader
int must_seek = 0;
while ( args->ifname < args->nfnames && args->start_pos[args->ifname]!=-1 && line->pos >= args->start_pos[args->ifname] )
{
must_seek = 1;
if ( !bcf_sr_add_reader(args->files,args->fnames[args->ifname]) ) error("Failed to open %s: %s\n", args->fnames[args->ifname],bcf_sr_strerror(args->files->errnum));
args->ifname++;
}
if ( must_seek )
{
bcf_sr_seek(args->files, bcf_seqname(args->files->readers[i].header,line), line->pos);
seek_pos = line->pos;
seek_chr = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[i].header,line));
continue;
}
// We are assuming that there is a perfect overlap, sites which are not present in both files are dropped
if ( args->files->nreaders>1 && !bcf_sr_has_line(args->files,1) && !bcf_sr_region_done(args->files,1) ) continue;
phased_push(args, bcf_sr_get_line(args->files,0), args->files->nreaders>1 ? bcf_sr_get_line(args->files,1) : NULL);
}
if ( args->files->nreaders )
{
phased_flush(args);
while ( args->files->nreaders )
bcf_sr_remove_reader(args->files, 0);
}
}
}
else if ( args->files ) // combining overlapping files, using synced reader
{
while ( bcf_sr_next_line(args->files) )
{
for (i=0; i<args->files->nreaders; i++)
{
bcf1_t *line = bcf_sr_get_line(args->files,i);
if ( !line ) continue;
bcf_translate(args->out_hdr, args->files->readers[i].header, line);
bcf_write1(args->out_fh, args->out_hdr, line);
if ( args->remove_dups ) break;
}
}
}
else // concatenating
{
kstring_t tmp = {0,0,0};
int prev_chr_id = -1, prev_pos;
bcf1_t *line = bcf_init();
for (i=0; i<args->nfnames; i++)
{
htsFile *fp = hts_open(args->fnames[i], "r"); if ( !fp ) error("Failed to open: %s\n", args->fnames[i]);
bcf_hdr_t *hdr = bcf_hdr_read(fp); if ( !hdr ) error("Failed to parse header: %s\n", args->fnames[i]);
if ( !fp->is_bin && args->output_type&FT_VCF )
{
line->max_unpack = BCF_UN_STR;
// if VCF is on both input and output, avoid VCF to BCF conversion
while ( hts_getline(fp, KS_SEP_LINE, &fp->line) >=0 )
{
char *str = fp->line.s;
while ( *str && *str!='\t' ) str++;
tmp.l = 0;
kputsn(fp->line.s,str-fp->line.s,&tmp);
int chr_id = bcf_hdr_name2id(args->out_hdr, tmp.s);
if ( chr_id<0 ) error("The sequence \"%s\" not defined in the header: %s\n(Quick workaround: index the file.)\n", tmp.s, args->fnames[i]);
if ( prev_chr_id!=chr_id )
{
prev_pos = -1;
if ( args->seen_seq[chr_id] )
error("\nThe chromosome block %s is not contiguous, consider running with -a.\n", tmp.s);
}
char *end;
int pos = strtol(str+1,&end,10) - 1;
if ( end==str+1 ) error("Could not parse line: %s\n", fp->line.s);
if ( prev_pos > pos )
error("The chromosome block %s is not sorted, consider running with -a.\n", tmp.s);
args->seen_seq[chr_id] = 1;
prev_chr_id = chr_id;
if ( vcf_write_line(args->out_fh, &fp->line)!=0 ) error("Failed to write %d bytes\n", fp->line.l);
}
}
else
{
// BCF conversion is required
line->max_unpack = 0;
while ( bcf_read(fp, hdr, line)==0 )
{
bcf_translate(args->out_hdr, hdr, line);
if ( prev_chr_id!=line->rid )
{
prev_pos = -1;
if ( args->seen_seq[line->rid] )
error("\nThe chromosome block %s is not contiguous, consider running with -a.\n", bcf_seqname(args->out_hdr, line));
}
if ( prev_pos > line->pos )
error("The chromosome block %s is not sorted, consider running with -a.\n", bcf_seqname(args->out_hdr, line));
args->seen_seq[line->rid] = 1;
prev_chr_id = line->rid;
if ( bcf_write(args->out_fh, args->out_hdr, line)!=0 ) error("Failed to write\n");
}
}
bcf_hdr_destroy(hdr);
hts_close(fp);
}
bcf_destroy(line);
free(tmp.s);
}
}
int ingest1(const char *input,const char *output,char *ref,bool exit_on_mismatch=true) {
cerr << "Input: " << input << "\tOutput: "<<output<<endl;
kstream_t *ks;
kstring_t str = {0,0,0};
gzFile fp = gzopen(input, "r");
VarBuffer vbuf(1000);
int prev_rid = -1;
if(fp==NULL) {
fprintf(stderr,"problem opening %s\n",input);
exit(1);
}
char *out_fname = (char *)malloc(strlen(output)+5);
strcpy(out_fname,output);
strcat(out_fname,".tmp");
if(fileexists(out_fname)) {
fprintf(stderr,"%s file already exists. will not overwrite\n",out_fname);
exit(1);
}
printf("depth: %s\n",out_fname);
gzFile depth_fp = gzopen(out_fname, "wb1");
strcpy(out_fname,output);
strcat(out_fname,".bcf");
if(fileexists(out_fname)) {
fprintf(stderr,"%s file already exists. will not overwrite\n",out_fname);
exit(1);
}
printf("variants: %s\n",out_fname);
htsFile *variant_fp=hts_open(out_fname,"wb1");
if(variant_fp==NULL) {
fprintf(stderr,"problem opening %s\n",input);
exit(1);
}
ks = ks_init(fp);
htsFile *hfp=hts_open(input, "r");
bcf_hdr_t *hdr_in = bcf_hdr_read(hfp);
hts_close(hfp);
//this is a hack to fix gvcfs where AD is incorrectly defined in the header. (vcf4.2 does not technically allow Number=R)
bcf_hdr_remove(hdr_in,BCF_HL_FMT,"AD");
assert( bcf_hdr_append(hdr_in,"##FORMAT=<ID=AD,Number=R,Type=Integer,Description=\"Allelic depths for the ref and alt alleles in the order listed. For indels this value only includes reads which confidently support each allele (posterior prob 0.999 or higher that read contains indicated allele vs all other intersecting indel alleles)\">") == 0);
//this is a hack to fix broken gvcfs where GQ is incorrectly labelled as float (v4.3 spec says it should be integer)
bcf_hdr_remove(hdr_in,BCF_HL_FMT,"GQ");
assert( bcf_hdr_append(hdr_in,"##FORMAT=<ID=GQ,Number=1,Type=Integer,Description=\"Genotype Quality\">") == 0);
// bcf_hdr_t *hdr_out=hdr_in;
bcf_hdr_t *hdr_out = bcf_hdr_dup(hdr_in);
remove_hdr_lines(hdr_out,BCF_HL_INFO);
remove_hdr_lines(hdr_out,BCF_HL_FLT);
bcf_hdr_sync(hdr_out);
//here we add FORMAT/PF. which is the pass filter flag for alts.
assert( bcf_hdr_append(hdr_out,"##FORMAT=<ID=PF,Number=A,Type=Integer,Description=\"variant was PASS filter in original sample gvcf\">") == 0);
args_t *norm_args = init_vcfnorm(hdr_out,ref);
norm_args->check_ref |= CHECK_REF_WARN;
bcf1_t *bcf_rec = bcf_init();
bcf_hdr_write(variant_fp, hdr_out);
kstring_t work1 = {0,0,0};
int buf[5];
ks_tokaux_t aux;
int ndec=0;
int ref_len,alt_len;
while( ks_getuntil(ks, '\n', &str, 0) >=0) {
// fprintf(stderr,"%s\n",str.s);
if(str.s[0]!='#') {
char *ptr = kstrtok(str.s,"\t",&aux);//chrom
ptr = kstrtok(NULL,NULL,&aux);//pos
work1.l=0;
kputsn(str.s,ptr-str.s-1, &work1);
buf[0] = bcf_hdr_name2id(hdr_in, work1.s);
assert( buf[0]>=0);
buf[1]=atoi(ptr)-1;
ptr = kstrtok(NULL,NULL,&aux);//ID
ptr = kstrtok(NULL,NULL,&aux);//REF
ref_len=0;
while(ptr[ref_len]!='\t') ref_len++;
ptr = kstrtok(NULL,NULL,&aux);//ALT
bool is_variant=false;
alt_len=0;
while(ptr[alt_len]!='\t') alt_len++;
if(ptr[0]!='.')
is_variant=true;
char * QUAL_ptr = kstrtok(NULL, NULL, &aux);
assert (QUAL_ptr != NULL);
for(int i=0;i<2;i++) ptr = kstrtok(NULL,NULL,&aux);// gets us to INFO
//find END if it is there
char *end_ptr=strstr(ptr,"END=") ;
if(end_ptr!=NULL)
buf[2]=atoi(end_ptr+4)-1;
else
buf[2]=buf[1]+alt_len-1;
ptr = kstrtok(NULL,NULL,&aux);//FORMAT
//find index of DP (if present)
//if not present, dont output anything (indels ignored)
char *DP_ptr = find_format(ptr,"DP");
int GQX = 0;
int QUAL = 0;
// AH: change code to use the minimum of GQ and QUAL fields if
// GQX is not defined. See here:
// https://support.basespace.illumina.com/knowledgebase/articles/144844-vcf-file
// "GQXGenotype quality. GQX is the minimum of the GQ value
// and the QUAL column. In general, these are similar values;
// taking the minimum makes GQX the more conservative measure of
// genotype quality."
if(DP_ptr!=NULL) {
buf[3]=atoi(DP_ptr);
char *GQX_ptr = find_format(ptr,"GQX");
if (GQX_ptr == NULL)
{
GQX_ptr = find_format(ptr,"GQ");
GQX = atoi(GQX_ptr);
if (QUAL_ptr[0] != '.')
{
QUAL = atoi(QUAL_ptr);
if (QUAL < GQX)
GQX = QUAL;
}
}
else
{
GQX = atoi(GQX_ptr);
}
//trying to reduce entropy on GQ to get better compression performance.
//1. rounds down to nearest 10.
//2. sets gq to min(gq,100).
buf[4]=GQX/10;
buf[4]*=10;
if(buf[4]>100) buf[4]=100;
// printf("%d\t%d\t%d\t%d\t%d\n",buf[0],buf[1],buf[2],buf[3],buf[4]);
if(gzwrite(depth_fp,buf,5*sizeof(int))!=(5*sizeof(int)))
die("ERROR: problem writing "+(string)out_fname+".tmp");
}
if(is_variant) {//wass this a variant? if so write it out to the bcf
norm_args->ntotal++;
vcf_parse(&str,hdr_in,bcf_rec);
// cerr<<bcf_rec->rid<<":"<<bcf_rec->pos<<endl;
if(prev_rid!=bcf_rec->rid)
vbuf.flush(variant_fp,hdr_out);
else
vbuf.flush(bcf_rec->pos,variant_fp,hdr_out);
prev_rid=bcf_rec->rid;
int32_t pass = bcf_has_filter(hdr_in, bcf_rec, ".");
bcf_update_format_int32(hdr_out,bcf_rec,"PF",&pass,1);
bcf_update_filter(hdr_out,bcf_rec,NULL,0);
if(bcf_rec->n_allele>2) {//split multi-allelics (using vcfnorm.c from bcftools1.3
norm_args->nsplit++;
split_multiallelic_to_biallelics(norm_args,bcf_rec );
for(int i=0;i<norm_args->ntmp_lines;i++){
remove_info(norm_args->tmp_lines[i]);
if(realign(norm_args,norm_args->tmp_lines[i]) != ERR_REF_MISMATCH)
ndec+=decompose(norm_args->tmp_lines[i],hdr_out,vbuf);
else
if(exit_on_mismatch)
die("vcf did not match the reference");
else
norm_args->nskipped++;
}
}
else {
remove_info(bcf_rec);
if( realign(norm_args,bcf_rec) != ERR_REF_MISMATCH)
ndec+=decompose(bcf_rec,hdr_out,vbuf);
else
if(exit_on_mismatch)
die("vcf did not match the reference");
else
norm_args->nskipped++;
}
vbuf.flush(bcf_rec->pos,variant_fp,hdr_out);
}
}
}
vbuf.flush(variant_fp,hdr_out);
bcf_hdr_destroy(hdr_in);
bcf_hdr_destroy(hdr_out);
bcf_destroy1(bcf_rec);
ks_destroy(ks);
gzclose(fp);
gzclose(depth_fp);
free(str.s);
free(work1.s);
hts_close(variant_fp);
destroy_data(norm_args);
fprintf(stderr,"Variant lines total/split/realigned/skipped:\t%d/%d/%d/%d\n", norm_args->ntotal,norm_args->nsplit,norm_args->nchanged,norm_args->nskipped);
fprintf(stderr,"Decomposed %d MNPs\n", ndec);
fprintf(stderr,"Indexing %s\n",out_fname);
bcf_index_build(out_fname, BCF_LIDX_SHIFT);
free(out_fname);
return 0;
}
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;
}
void abcWriteBcf::print(funkyPars *pars){
if(doBcf==0)
return;
kstring_t buf;
if(fp==NULL){
buf.s=NULL;buf.l=buf.m=0;
fp=aio::openFileHts(outfiles,".bcf");
hdr = bcf_hdr_init("w");
rec = bcf_init1();
print_bcf_header(fp,hdr,args,buf,header);
}
lh3struct *lh3 = (lh3struct*) pars->extras[5];
freqStruct *freq = (freqStruct *) pars->extras[6];
genoCalls *geno = (genoCalls *) pars->extras[10];
for(int s=0;s<pars->numSites;s++){
if(pars->keepSites[s]==0)
continue;
rec->rid = bcf_hdr_name2id(hdr,header->target_name[pars->refId]);
rec->pos = pars->posi[s];//<- maybe one index?
// bcf_update_id(hdr, rec, "rs6054257");
char majmin[4]={intToRef[pars->major[s]],',',intToRef[pars->minor[s]],'\0'};
bcf_update_alleles_str(hdr, rec, majmin);
rec->qual = 29;
// .. FILTER
int32_t tmpi = bcf_hdr_id2int(hdr, BCF_DT_ID, "PASS");
bcf_update_filter(hdr, rec, &tmpi, 1);
// .. INFO
tmpi = pars->keepSites[s];
bcf_update_info_int32(hdr, rec, "NS", &tmpi, 1);
if(pars->counts){
int depth = 0;
for(int i=0; i<4*pars->nInd; i++)
depth += pars->counts[s][i];
tmpi = depth;
bcf_update_info_int32(hdr, rec, "DP", &tmpi, 1);
}
if(freq){
float tmpf = freq->freq_EM[s];
bcf_update_info_float(hdr, rec, "AF", &tmpf, 1);
}
// .. FORMAT
assert(geno);
if(geno){
int32_t *tmpia = (int*)malloc(bcf_hdr_nsamples(hdr)*2*sizeof(int32_t));
for(int i=0; i<pars->nInd;i++){
if(geno->dat[s][i]==0){
tmpia[2*i+0] = bcf_gt_unphased(0);
tmpia[2*i+1] = bcf_gt_unphased(0);
}else if(geno->dat[s][i]==1){
tmpia[2*i+0] = bcf_gt_unphased(0);
tmpia[2*i+1] = bcf_gt_unphased(1);
} else{
tmpia[2*i+0] = bcf_gt_unphased(1);
tmpia[2*i+1] = bcf_gt_unphased(1);
}
}
bcf_update_genotypes(hdr, rec, tmpia, bcf_hdr_nsamples(hdr)*2);
free(tmpia);
}
if(pars->counts){
int32_t *tmpfa = (int32_t*)malloc(sizeof(int32_t)*bcf_hdr_nsamples(hdr));
suint *ary=pars->counts[s];
for(int i=0;i<bcf_hdr_nsamples(hdr);i++)
tmpfa[i] = ary[0]+ary[1]+ary[2]+ary[3];
bcf_update_format_int32(hdr, rec, "DP", tmpfa,bcf_hdr_nsamples(hdr) );
free(tmpfa);
}
assert(lh3);
if(lh3){
float *tmpfa = (float*)malloc(3*bcf_hdr_nsamples(hdr)*sizeof(float ));
int32_t *tmpi = (int32_t*)malloc(3*bcf_hdr_nsamples(hdr)*sizeof(int32_t));
double *ary = lh3->lh3[s];
for(int i=0;i<bcf_hdr_nsamples(hdr);i++)
for(int j=0;j<3;j++){
tmpfa[i*3+j] = ary[i*3+j]/M_LN10;
tmpi[i*3+j] =(int) -log10(exp(ary[i*3+j]))*10.0;
// fprintf(stderr,"pl:%d raw:%f\n",tmpi[i*3+j],ary[i*3+j]);
}
bcf_update_format_float(hdr, rec, "GL", tmpfa,3*bcf_hdr_nsamples(hdr) );
bcf_update_format_int32(hdr, rec, "PL", tmpi,3*bcf_hdr_nsamples(hdr) );
free(tmpfa);
free(tmpi);
}
if ( bcf_write1(fp, hdr, rec)!=0 ){
fprintf(stderr,"Failed to write to \n");
exit(0);
}
// fprintf(stderr,"------\n");
bcf_clear1(rec);
}
}
int main(int argc, char **argv)
{
char *fname = argc>1 ? argv[1] : "/dev/null";
htsFile *fp = hts_open(fname, "w");
bcf_hdr_t *hdr1, *hdr2;
hdr1 = bcf_hdr_init("w");
hdr2 = bcf_hdr_init("w");
// Add two shared and two private annotations
bcf_hdr_append(hdr1, "##contig=<ID=1>");
bcf_hdr_append(hdr1, "##contig=<ID=2>");
bcf_hdr_append(hdr2, "##contig=<ID=2>");
bcf_hdr_append(hdr2, "##contig=<ID=1>");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT1,Description=\"Filter 1\">");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT2,Description=\"Filter 2\">");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT3,Description=\"Filter 3\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT4,Description=\"Filter 4\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT3,Description=\"Filter 3\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT2,Description=\"Filter 2\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF1,Number=.,Type=Integer,Description=\"Info 1\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF2,Number=.,Type=Integer,Description=\"Info 2\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF3,Number=.,Type=Integer,Description=\"Info 3\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF4,Number=.,Type=Integer,Description=\"Info 4\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF3,Number=.,Type=Integer,Description=\"Info 3\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF2,Number=.,Type=Integer,Description=\"Info 2\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT1,Number=.,Type=Integer,Description=\"FMT 1\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT2,Number=.,Type=Integer,Description=\"FMT 2\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT3,Number=.,Type=Integer,Description=\"FMT 3\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT4,Number=.,Type=Integer,Description=\"FMT 4\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT3,Number=.,Type=Integer,Description=\"FMT 3\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT2,Number=.,Type=Integer,Description=\"FMT 2\">");
bcf_hdr_add_sample(hdr1,"SMPL1");
bcf_hdr_add_sample(hdr1,"SMPL2");
bcf_hdr_add_sample(hdr2,"SMPL1");
bcf_hdr_add_sample(hdr2,"SMPL2");
bcf_hdr_sync(hdr1);
bcf_hdr_sync(hdr2);
hdr2 = bcf_hdr_merge(hdr2,hdr1);
bcf_hdr_sync(hdr2);
if ( bcf_hdr_write(fp, hdr2)!=0 ) error("Failed to write to %s\n", fname);
bcf1_t *rec = bcf_init1();
rec->rid = bcf_hdr_name2id(hdr1, "1");
rec->pos = 0;
bcf_update_alleles_str(hdr1, rec, "G,A");
int32_t tmpi[3];
tmpi[0] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT1");
tmpi[1] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT2");
tmpi[2] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT3");
bcf_update_filter(hdr1, rec, tmpi, 3);
tmpi[0] = 1; bcf_update_info_int32(hdr1, rec, "INF1", tmpi, 1);
tmpi[0] = 2; bcf_update_info_int32(hdr1, rec, "INF2", tmpi, 1);
tmpi[0] = 3; bcf_update_info_int32(hdr1, rec, "INF3", tmpi, 1);
tmpi[0] = tmpi[1] = 1; bcf_update_format_int32(hdr1, rec, "FMT1", tmpi, 2);
tmpi[0] = tmpi[1] = 2; bcf_update_format_int32(hdr1, rec, "FMT2", tmpi, 2);
tmpi[0] = tmpi[1] = 3; bcf_update_format_int32(hdr1, rec, "FMT3", tmpi, 2);
bcf_remove_filter(hdr1, rec, bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT2"), 0);
bcf_update_info_int32(hdr1, rec, "INF2", NULL, 0);
bcf_update_format_int32(hdr1, rec, "FMT2", NULL, 0);
bcf_translate(hdr2, hdr1, rec);
if ( bcf_write(fp, hdr2, rec)!=0 ) error("Faild to write to %s\n", fname);
// Clean
bcf_destroy1(rec);
bcf_hdr_destroy(hdr1);
bcf_hdr_destroy(hdr2);
int ret;
if ( (ret=hts_close(fp)) )
{
fprintf(stderr,"hts_close(%s): non-zero status %d\n",fname,ret);
exit(ret);
}
return 0;
}
int process_region_guess(args_t *args, char *seq, regitr_t *itr)
{
int kitr = 1;
uint32_t start = 0, end = INT_MAX;
reg_stats_t *stats = NULL;
// set the start and the end position
if ( itr )
{
start = itr->reg[itr->i].start;
end = itr->reg[itr->i].end;
// flush all records with the same coordinates
while ( itr->i+kitr<itr->n && start==itr->reg[itr->i+kitr].start && end==itr->reg[itr->i+kitr].end ) kitr++;
int min,max,ret = ploidy_query(args->ploidy, seq, start, args->sex2ploidy, &min, &max);
assert(ret);
stats = expand_regs(args, seq,start,end);
}
else
{
// background region
int spos, epos;
const char *ptr = hts_parse_reg(args->background, &spos, &epos);
if ( !ptr )
error("Could not parse the region: %s\n", args->background);
seq = (char*) malloc(ptr - args->background + 1);
memcpy(seq,args->background,ptr-args->background);
seq[ptr-args->background] = 0;
start = spos;
end = epos;
}
if ( bcf_sr_seek(args->sr,seq,start)!=0 )
{
// sequence not present
if ( !itr ) free(seq);
return kitr;
}
int ismpl, rid = bcf_hdr_name2id(args->hdr,seq);
if ( !itr ) free(seq);
while ( bcf_sr_next_line(args->sr) )
{
bcf1_t *rec = bcf_sr_get_line(args->sr,0);
if ( rec->rid!=rid || rec->pos > end ) break;
if ( args->guess & GUESS_GT ) // use GTs to guess the ploidy
{
bcf_fmt_t *fmt = bcf_get_fmt(args->hdr, rec, "GT");
if ( !fmt ) continue;
for (ismpl=0; ismpl<args->nsample; ismpl++)
{
count_t *counts = stats ? &stats->counts[ismpl] : &args->bg_counts[ismpl];
int gt = bcf_gt_type(fmt, ismpl, NULL,NULL);
if ( gt==GT_UNKN ) counts->nmiss++;
else if ( gt==GT_HET_RA || gt==GT_HET_AA ) counts->nhet++;
else counts->nhom++;
}
}
else // use PLs to guess the ploidy
{
int gl2pl = args->guess & GUESS_PL ? 1 : -1;
int npl = bcf_get_format_int32(args->hdr,rec,args->guess&GUESS_PL?"PL":"GL",&args->pls,&args->npls);
if ( npl<=0 ) continue;
npl /= args->nsample;
for (ismpl=0; ismpl<args->nsample; ismpl++)
{
int32_t *ptr = args->pls + ismpl*npl;
int phom = INT_MAX, phet = INT_MAX, ial, jal, k = 0;
for (ial=0; ial<rec->n_allele; ial++)
{
for (jal=0; jal<ial; jal++)
{
if ( ptr[k] == bcf_int32_missing || ptr[k] == bcf_int32_vector_end ) break;
ptr[k] *= gl2pl;
if ( phet > ptr[k] ) phet = ptr[k];
k++;
}
if ( ptr[k] == bcf_int32_missing || ptr[k] == bcf_int32_vector_end ) break;
ptr[k] *= gl2pl;
if ( phom > ptr[k] ) phom = ptr[k];
k++;
}
count_t *counts = stats ? &stats->counts[ismpl] : &args->bg_counts[ismpl];
if ( k == rec->n_allele ) counts->nhom++; // haploid
else if ( phet == phom || k != rec->n_allele*(rec->n_allele+1)/2 ) counts->nmiss++;
else if ( phet < phom ) counts->nhet++;
else counts->nhom++;
}
}
}
return kitr;
}
void merge_chrom2qual(args_t *args, bcf1_t *out)
{
bcf_srs_t *files = args->files;
bcf_hdr_t *out_hdr = args->out_hdr;
int i, ret;
khiter_t kitr;
strdict_t *tmph = args->tmph;
kh_clear(strdict, tmph);
kstring_t *tmps = &args->tmps;
tmps->l = 0;
maux_t *ma = args->maux;
int *al_idxs = (int*) calloc(ma->nals,sizeof(int));
out->qual = 0;
// CHROM, POS, ID, QUAL
out->pos = -1;
for (i=0; i<files->nreaders; i++)
{
if ( !ma->has_line[i] ) continue;
bcf_sr_t *reader = &files->readers[i];
bcf1_t *line = reader->buffer[0];
bcf_hdr_t *hdr = reader->header;
// alleles
int j;
for (j=1; j<line->n_allele; j++)
al_idxs[ ma->d[i][0].map[j] ] = 1;
// position
if ( out->pos==-1 )
{
const char *chr = hdr->id[BCF_DT_CTG][line->rid].key;
out->rid = bcf_hdr_name2id(out_hdr, chr);
if ( strcmp(chr,out_hdr->id[BCF_DT_CTG][out->rid].key) ) error("Uh\n");
out->pos = line->pos;
}
// ID
if ( line->d.id[0]!='.' || line->d.id[1] )
{
kitr = kh_get(strdict, tmph, line->d.id);
if ( kitr == kh_end(tmph) )
{
if ( tmps->l ) kputc(';', tmps);
kputs(line->d.id, tmps);
kh_put(strdict, tmph, line->d.id, &ret);
}
}
// set QUAL to the max qual value. Not exactly correct, but good enough for now
if ( out->qual < files->readers[i].buffer[0]->qual ) out->qual = files->readers[i].buffer[0]->qual;
}
// set ID
if ( !tmps->l ) kputs(".", tmps);
if ( out->d.id ) free(out->d.id);
out->d.id = strdup(tmps->s);
// set alleles
ma->nout_als = 0;
for (i=1; i<ma->nals; i++)
{
if ( !al_idxs[i] ) continue;
ma->nout_als++;
// Adjust the indexes, the allele map could be created for multiple collapsed records,
// some of which might be unused for this output line
int ir, j;
for (ir=0; ir<files->nreaders; ir++)
{
if ( !ma->has_line[ir] ) continue;
bcf1_t *line = files->readers[ir].buffer[0];
for (j=1; j<line->n_allele; j++)
if ( ma->d[ir][0].map[j]==i ) ma->d[ir][0].map[j] = ma->nout_als;
}
}
// Expand the arrays and realloc the alleles string. Note that all alleles are in a single allocated block.
ma->nout_als++;
hts_expand0(char*, ma->nout_als, ma->mout_als, ma->out_als);
int k = 0;
for (i=0; i<ma->nals; i++)
if ( i==0 || al_idxs[i] ) ma->out_als[k++] = strdup(ma->als[i]);
assert( k==ma->nout_als );
normalize_alleles(ma->out_als, ma->nout_als);
bcf_update_alleles(out_hdr, out, (const char**) ma->out_als, ma->nout_als);
free(al_idxs);
for (i=0; i<ma->nout_als; i++) free(ma->out_als[i]);
}