Hi All! Apologies for a somewhat silly question. I want to do some basic analyses with polygenic scores and whether they predict psychiatric outcomes, but a big chunk of the sample are individuals who are related (not on purpose, not all of them are twin pairs or trios, degrees of relatedness vary). I briefly looked at the within-sibling GWAS but I wasn’t entirely sure if I can use its output/summary stats to calculate PRS? Or should I just exclude all the related samples (I was wondering if I can analyse them separately and apply the w-s GWAS data to them).
Hi Ada,
This is a great question. I would recommend you use the largest GWAS you can find with your trait (that does not include your target/application sample), this would maximise the quality of your PRS.
Then, you can model the relatedness in your sample, while evaluating your PRS. We ran a tutorial a couple years ago, specifically on this topic, you will find all the details and code here: https://ibg.colorado.edu/cdrom2022/day7/Answers/RR_WorkshopDay7.html#openmx_model.
I hope this helps.
To be a bit more specific, you say you have a wide range of relatedness (not only twins), so I assume you have genotyped your individuals? If this is the case, have a look at the GCTA section of the code, we have showed how you can model genetic relatedness, as well as familial shared environment, if you also wanted to control for it.