bcftools POS comparison seems to ignore CHROM

[fgvieira]

Dear all,

I am using bcftools v1.10.2 and got an error while filtering a VCF file with SVs:

$ cat FILE.vcf | bcftools filter --soft-filter GM_PoN --mode + --exclude 'INFO/GM_CNV_AF > 0.05' --output-type z
[W::vcf_parse] INFO/END=19212036 is smaller than POS at 2:33141291

Looking at that position:

$ fgrep '33141291' FILE.vcf
2	33141291	BND00002716	T	[1:19212036[T	.	LowQual	IMPRECISE;SVTYPE=BND;SVMETHOD=EMBL.DELLYv0.8.1;CHR2=1;END=19212036;PE=2;MAPQ=33;CT=5to5;CIPOS=-460,460;CIEND=-460,460;GM_CNV_AC=47;GM_CNV_AF=0.10829493087557604	GT:GL:GQ:FT:RCL:RC:RCR:CN:DR:DV:RR:RV	0/1:-673.579,0,-70.7019:10000:PASS:78:996:171:8:30:379:0:0

It seems that bcftools is only comparing the start/end positions but not the chromosome. Is this intended?
Thanks,

[jmarshall]

The intention in VCF has always been that INFO/END represents the end of the region starting from CHROM:POS that the variant spans, thus is necessarily considered to be on the same chromosome. This was clarified in the VCF specification last year; see samtools/hts-specs#425 and samtools/hts-specs#436.

This is a bug in Delly, which started using END for a different purpose when the specification was less clear as to END's standard purpose. See dellytools/delly#159.