Entering edit mode
3.8 years ago
Fadoua
▴
20
Hi Dear all,
I am trying to create a weighted genetic risk score model.
However, The GWAS from where I got the SNP associated with the exposure factor, contained some SNPs that have a negative Beta coefficient; when I compare to my data the coded allele is reverse. how can I correct this before starting to establish my genetic score?
Any help is much appreciated.
Thank you,
Best regards,
It's a known problem, especially if you're doing meta-analysis, you have to make sure your betas are in the same direction. Easiest way is to take the negative or make sure the major/minor alleles are the same.
Thank you for your reply. Actually I am willing to conduct an MR, using genetics variants as instrumental variables in order to infer causality between my exposure of interest and my target outcome. but first, I need to construct a PRS using summary data from a GWAS. the issue here is that in the original GWAS some betas of some SNPs are negative, when I have compared it to the target data I have found that the coded allele is reverse.so I am not sure how should I fix this problem? calculating the genetic risk score will be a mistake based on the revered allele? I am sorry I don't know if I am being clear.thank you in advance Regards
There are software (e.g. PRSice) that takes care of this automatically.
Have you tried multiplying the estimates for reverse SNPs by -1? This is the simplest and suggested solution before running MR. You need to do it in one of the two (exposure and outcome) datasets only.
It is generally not recommend to use PRS as an instrument for MR as PRS likely contains pleiotropic effects, which breaks the basic assumption of MR.
On the other hand, you can use many different software for PRS calculation which usually handles the mismatch coding alleles and GWAS effective alleles automatically such as PRSice as suggested by Lemire, lassosum, LDPred 2 and also PRS-CS
yes, Thank you for your suggestion. I was a little bit confused back then. its already done.