TriLLIEM: Triad Log-Linear Modelling of Imprinting, Environmental Interactions, and Maternal Effects

Abstract

The case-parents design, where an affected child and their parents are genotyped, allows for estimation of disease risk due to either the child's or mother's genes through log-linear modelling. It is robust to the confounding effects of population subdivision, but cannot account for non-exchangeability of parental genotypes. Factors such as gene-environment interactions and imprinting, where risk depends on which parent contributed a genetic factor, can also be estimated. Existing analytical software are either deprecated (LEM) or have limited modelling capabilities (Haplin, EMIM). We introduce our R package, TriLLIEM, which implements the log-linear approach to address these limitations. The software includes options for gene-environment interactions, imprinting by environment interactions, and non-exchangeable parental genotypes if parents of controls are also available. We use these features in a simulation study to assess the accuracy of TriLLIEM and to show how population stratification confounds the model for gene-environment interactions when control-triads are included.