Meta-imputation across reference panels of different genome builds

[xuj18]

Hi! I am planning to meta-impute across TOPMed, 1000G, and Genome Asia. However, the Genome Asia reference panel on the Michigan server only has the option of GRCh37. I was wondering whether there is a way to meta-impute across these 3 panels. I suppose I will have to make manual changes in the Genome Asia imputed files to lift it over to GRCh38. Do you have any suggestions on what columns I need to change (e.g., SNP names, positions) in the Genome Asia imputed files, and whether I need to change these columns in all .dose.vcf.gz and .empiricalDose.vcf.gz files for the meta-imputation. Thank you!

[yukt]

You will need to liftover both the Genome Asia imputed .dose.vcf.gz file and .empiricalDose.vcf.gz file.

The genotype information required for meta-imputation are HDS in .dose.vcf.gz and GT and LDS in .empiricalDose.vcf.gz, but I am not sure if current liftover tools support formats other than GT.

Not only the SNP name/position needs change. There are some switch between REF and ALT allele from GRCh37 to GRCh38, and when it happens you will need to make changes to genotype values (e.g. HDS_liftover = 1 - HDS) as well as INFO/AF if applicable.