Relatedness with KING vs heterogeneous data

Question

[German.M.Demidov]

Hi community members,

I am working on KING relatedness inference having >20K samples sequenced with >30 exome enrichment kits. When I worked on less heterogeneous data, I was getting nice results - IBD1/2 were centered where they supposed to be. But now the IBD1/2 are shifted and I guess this is due to SNVs covered/genotyped well for some kits and not so well for others.

Could you recommend me how to choose SNVs well-genotyped in all the kits? Currently I use intersection of exome target files for filtering, but it does not seem very optimal.

I have gVCF files so I can say quality of genotype for all positions.