Statistical models of gene-environment interactions (360G-Wellcome-109106_Z_15_A)
This PhD focuses on developing statistical methods to discover gene – environment (G-E) interactions. To date there has been some interest in testing for G-E interactions in animal models, but limited success in uncovering examples of G-E interactions in humans. This is in part due to the problem of exposure assessment, or rather, because representative data on the environment of a number of individuals over a lifetime has been hard to acquire. However the data recently made available by the UK BioBank, on over 500000 individuals and a wide array of environmental covariates, may now make it possible to detect these interactions. We aim to use a Bayesian methodology to first test a number of known models, such as a random effects model, against the dataset. We will then attempt to use a Gaussian Process Regression model to identify covariates involved in G-E interactions. This approach is advantageous as GPR is non-parametric, thus avoiding the curse of dimensionality, and places no assumptions on the order of interactions. However as this method currently scales in a cubic manner following the number of samples, significant computational challenges remain.