I am following a PRSice tutorial for PRS generation which recommends pruning and later clumping SNPs (removing those that are in high LD).

I am applying it to an imputed data set. My base data has about 8 million SNPs, the target data 400 000 SNP prior to imputation (presumably much more post imputation).

It seems like pruning then clumping would be non-sensical to perform on imputed data as the point is that the non-sequenced SNPs can only be confidently added to the list as they are highly correlated.

I saw another post that suggests that clumping is still worthwhile and won't lose useful information but I'm still not sure about pruning PRSice: Imputation and clumping

Thanks for any advice