Gene-environment interaction in association analysis

点击翻译按钮获取中文摘要

DESCRIPTION (provided by Applicant): Gene x environment interaction (GxE) in complex human phenotypes is likely to be common, important and difficult to detect. In particular, for the understanding of behavioral and psychiatric phenotypes, approaches are needed that bring together multiple levels of genetic, biological, psychological and social factors, allowing for the interaction and correlation of these factors. However, statistical tests for interaction are prone to both false positive (as the presence of statistical interaction is dependent on how the main effects are defined) and false negative results (as tests for interaction suffer from low power). Furthermore, in the context of large scale association analysis (in which multiple testing is already a critical issue) incorporation of GxE, if lacking the appropriate statistical rigor, could potentially only exacerbate the considerable set of existing problems. This application aims to develop and evaluate new methodology with specific emphasis on robustness to nonlinear effects; to extend current methods to fit causal models to multiple genes and multiple environments; to develop and evaluate a framework for incorporating GxE in large-scale association studies in a way that refines provisional positive associations rather than increasing the multiple testing burden; to perform an exploratory simulation study within a developmental framework to investigate the likelihood and impact of nonlinear effects; to create freely-distributed software implementations. Expanding on preliminary studies, the objective is to develop and implement a maximum likelihood approach to provide a general framework for testing for GxE in families and unrelated individuals, for dichotomous and continuous traits, for allelic, genotypic or haplotypic association. Simulation studies (using both standard approaches based on asymptotic theory as well as a dynamic, nonlinear developmental approach) have been designed that will assess the performance of the methods across a broad range of genetic models.