Effect Patterns in Genetic Mapping Studies

Abstract

In genetic mapping, regions of interest are generally first identified based on the suspected presence of marginal associations with the trait. It is natural to consider next the multigenic additive effects within these regions. Learning about the true multigenic effects is challenging since effect sizes are typically small relative to the uncertainty. Here we focus on patterns in the additive effect structure that can be captured as functions of the true effect parameters. Our major interest is in determining what functions can be estimated with reasonable accuracy in a given setting. We find that with careful estimation, some measures of centrality and dispersion can be estimated even when the predictive association is very weak. The ideas are demonstrated using a dataset of serum HDL cholesterol levels related to around 300 genetic variants in 14 genomic regions.