A Perspective on Interaction Tests in Genetic Association Studies

The identification of gene-gene and gene-environment interaction in human traits and diseases is an active area of research that generates high expectation, and most often lead to high disappointment. This is partly explained by a misunderstanding of some of the inherent characteristics of interaction effects. Here, I untangle several theoretical aspects of standard regression-based interaction tests in genetic association studies. In particular, I discuss variables coding scheme, interpretation of effect estimate, power, and estimation of variance explained in regard of various hypothetical interaction patterns. I show first that the simplest biological interaction models—in which the magnitude of a genetic effect depends on a common exposure—are among the most difficult to identify. Then, I demonstrate the demerits of the current strategy to evaluate the contribution of interaction effects to the variance of quantitative outcomes and argue for the use of new approaches to overcome these issues. Finally I explore the advantages and limitations of multivariate models when testing for interaction between multiple SNPs and/or multiple exposures, using either a joint test, or a test of interaction based on risk score. Theoretical and simulated examples presented along the manuscript demonstrate that the application of these methods can provide a new perspective on the role of interaction in multifactorial traits.