I am trying to use PRSice. Unfortunately, my target sample has fewer SNPs than my discovery/base sample, as it comes from different platforms. My question is of a methodological nature: I could of course impute the missing SNPs, and PRSice explicitely mentions it can handle such information. However, PRSice also performs clumping. If I'm not mistaken, imputation fills in missing SNPs using linkage disequilibrium information, while clumping then tries to estimate LD and reduce the SNPs to an independent set. Wouldn't imputation and clumping pretty much cancel out then? I'm not sure what the best practice is in this situation.
The thing is, we only perform clumping on SNPs that were found in both the target and base file. By using the imputed data, you increase the coverage and might result in more post-clumped SNPs, therefore increase amount of information and possibly increase the performance of your PRS model