GWAS SNPs LD Clumping

Question

[Peixe]

Hello,

I am trying to figure out how to know if GWAS SNPs tagging a same or near region do come from the same real association signal or not. In other words, if a later GWAS identifies a close SNP to the signal we have, I would like to know if its an independent loci or its just a replica.

I started by trying to code a custom script for it only based on a certain LD threshold including SNPs within a specified kb range, so all SNPs under these conditions belong to a same signal. The problem is that if later I want to add a new set of SNPs mapping near these ones, calculations have to be redone exponentially increasing the SNPs combinations to check for LD, and it could be highly resource consuming.

Right now I am trying with the --clump option from plink, but as this is based in LD and p-value, I don't know if this would be a correct approach for it.

Do you have any advice, suggestion or preferred way/software to do this?

Answer 1

[Larry_Parnell]

We use 1000 Genomes data for populations with ancestry relevant to the populations we study, e.g. CEU for pops of European ancestry. We use Haploview and routinely use r^2 cutoffs of 0.8. This will need to be done as one moves from one population to another because LD blocks are different in different populations. We work with a 3-way admixed population and this owes its own challenges, where we calculate LD within that population using the actual genotype data.

[Peixe]

Yes, I already took into account the population issue. I work with European populations, so CEU.

Thanks, Larry, for pointing that!