insertion became deletion after liftover

[nzhun]

here is an example, an insertion was liftovered to a deletion. hg37 chr19:54754990_C/CGG, was liftovered to be: hg38 chr19:54251125:CGG/C

[freeseek]

BCFtools/liftover correctly handles indels swaps between references, as opposed to Picard/LiftoverVcf or CrossMap/VCF who cannot do this. You can use the tool to observe this behavior. This is what the variant looks like in hg19:

$ echo -e "##fileformat=VCFv4.2\n##contig=<ID=chr19>\n#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\nchr19\t54754990\t.\tC\tCGG\t.\t.\t." | bcftools +liftover -- -c hg19ToHg38.over.chain.gz -s hg19.fa | bcftools view -H
chr19	54754990	.	CGGC	CGGGGC	.	.	.

In hg19 the reference allele is two G's and the alternate allele is four G's. When you liftover to hg38:

$ echo -e "##fileformat=VCFv4.2\n##contig=<ID=chr19>\n#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\nchr19\t54754990\t.\tC\tCGG\t.\t.\t." | bcftools +liftover -- -c hg19ToHg38.over.chain.gz -s hg19.fa -f hg38.fa --no-left-align | bcftools view -H
Lines   total/swapped/reference added/rejected:	1/1/0/0
chr19	54251125	.	CGGGGC	CGGC	.	.	SWAP=1

In hg38 the reference allele is four G's and the alternate allele is two G's so that what was before the alternate allele is now the reference allele, as it should be