Statistics Colloquium, Dr. Chuanhua Julia Xing

Title: Nonparametric Bayesian Methods for Identifying Joint Effects of Disease Risk Factors from Genetic Interaction Networks.

Abstract:

It is widely acknowledged that diseases are typically triggered by the confluence of multiple genetic and external factors, rather than a single gene. We propose a nonparametric Bayes model that is based on tensor decomposition, which can capture high order interactions. The proposed model expands from probabilistic factorization for mixed scale variables, and utilizes conditional mutual information-based methods for evaluating interactions. Our proposed approach has considerable advantages in simultaneously considering high dimensional single nucleotide polymorphisms and environmental factors. We evaluate our methods using simulation studies and apply to real data for illustration.