Hardy Weinberg Equilibrium in GWAS

Question

[evoclive]

Hello,

When identifying population structure for GWAS studies, I assume it's correct to initially remove markers not in HWE. However, after determining pop structure - is it the case that markers that have been initially thrown out maybe of interest as their departure from HWE may indicate that they are under selection? Therefore, they should be considered in the actual GWAS analysis?

Is this a valid strategy or should they be removed for the entire study?

Thanks

Answer 1

[Kevin Blighe]

Ask two people and you will likely receive two different answers.

For things such as the determination of population structure and, e.g., performing PCA, MDS, and deriving covariates for population stratification adjustment, I filter ('prune') the variants based on linkage disequilibrium (LD) and MAF.

For the actual main GWAS, I do not filter for these, and neither do I filter for HWE. It had never felt 100% 'right' to me to eliminate variants based on HWE, and, indeed, this is backed up by this well-cited review by Stephen Turner:

• Quality Control Procedures for Genome Wide Association Studies (jump to the section entitled Hardy-Weinberg Equilibrium.

Kevin