Robust Principal Interactions Analysis for Gene-Gene, Gene-Environment Interactions with Repeated Measures Data

• Activity Number: 575
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Yi-An Ko*+ and Bhramar Mukherjee and Sung Kyun Park
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: gene-environment interactions ; singular value decomposition ; non-additivity ; mixed model ; longitudinal data ; profile likelihood

Abstract

There has been extensive literature targeted towards gene-gene and gene-environment interaction modeling in case-control studies but not much in longitudinal studies on quantitative traits. We borrow the ideas from the classical two-way ANOVA literature to address the issue of efficient modeling of interactions in longitudinal cohort studies. Additive main effects and multiplicative interaction (AMMI) models, which entail a singular value decomposition of the cell residual matrix after fitting additive main effects, have been shown to perform well across a spectrum of interaction models. We propose a novel method based on profile likelihood to test interactions from AMMI models for unbalanced longitudinal data. We compare the performance of AMMI models with Tukey's one degree-of-freedom non-additivity test and Mandel's row/column-regression models in which the interaction structure depends on main effects. Simulation results show that principal interaction analysis for repeated measures data is very powerful and robust with respective to misspecified interaction structures. We apply the proposed method to the Normative Aging Study.