GWAS principal components

Question

[Raygozak]

Hi, I have a question regarding the three genetic components that are used for GWAS as covariates. Given that I'm interested in a particular population, I compute the pcs from independent SNPs in my dataset and then remove outliers by visual inspection. My question is whether I should use the principal components calculated during this step or should I recompute the PCs used in the actual analysis after the removal of the outliers?

Thanks

Answer 1

[Kevin Blighe]

Is population stratification an issue in your cohort? I have come across people who 'blindly' adjust for PC covariates when it may not be needed. In any case, I feel that it would make more sense to re-calculate the eigenvectors / PCs are you remove the outlier(s).

[Raygozak]

That's a good question, I'm taking over the project and I'm relatively new to GWAs, but thanks for the tip I will check this. This is how it has been done before I started, but I think it is time to revisit some of these basic assumptions about the population.

[Kevin Blighe]

Okay, cool.