search

MRCIEU / TwoSampleMR

R package for performing 2-sample MR using MR-Base database
https://mrcieu.github.io/TwoSampleMR
Other
440 stars 177 forks source link

Clumping discrepancies #273

Open drobbie429 opened 3 years ago

drobbie429 commented 3 years ago

Hello,

I wanted to ask if you can advise on the following please.

I am performing a two sample MR within two cohorts. I have tried two methods to clump as both seem valid. 1) I clump the exposure first. 2) The other way I have tried is to find the genome-wide significant unclumped snps for the exposure that are present in the outcome cohorts and clump that set of SNPS. I get around 25 SNPs more using the second method. Can you advise why there’s quite a big discrepancy in the SNPs depending on which stage I clump and which may be the better way?

TIA

mightyphil2000 commented 3 years ago

Hi Drobbie

Possibly what's happening is that clumping method/stage 1 is leading to lots of index SNPs that are not present in the outcome dataset. How many SNPs do you get after clumping method/stage 1 and how many of those are missing from the outcome dataset? Another possibility is that some SNPs are missing from the LD reference panel. Perhaps the overlap between the reference panel and the outcome study SNPs is greater.

Shicheng-Guo commented 3 years ago

Where to find pre-clumped lead SNPs of GTEx-V8? Thanks.