A Semiparametric Bayesian Hierarchical Frailty Model for Evaluating Pathway Effects on Survival Time

• Activity Number: 608
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Lin Zhang* and Inyoung Kim
• Companies: Virginia Tech and Virginia Tech
• Keywords: Semiparametric ; Frailty model ; Bayesian ; Gene pathway

Abstract

Frailty model has been widely used in dealing with survival data with unobserved random effect. However, the strong assumptions it requires are often not satisfied in real application. Motivated by our breast cancer gene expression data (with small n and large p), we propose a semiparametric Bayesian hierarchical frailty model to study the effects of clinical covariates and expression levels of genes in a pathway on survival time. Pathway-based analysis has the ability to detect subtle changes of survival time that could be missed using gene-based analysis. The unknown high dimensional function for pathways is modeled nonparametrically via a kernel machine. Since the relationship between survival time and pathway effects is complex, our model allow the possibility of gene interaction within the same pathway. Clinical covariates are parametrically modeled. Hence this framework allows evaluating the joint effect of genes within a gene pathway simultaneously, as well as modeling potential nonlinear gene expression effects and gene-gene interaction within the same pathway. The advantages of our approach is illustrated with simulated data and a breast cancer gene expression data.