Bayesian Kernel-Based Modeling and Selection of Genetic Pathways and Genes for Cancer

• Activity Number: 694
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Zhenyu Wang*+ and Sounak Chakraborty and Jianguo Sun
• Companies: University of Missouri, Columbia and University of Missouri-Columbia and University of Missouri-Columbia
• Keywords: Bayesian variable selection ; Reproducing Kernel Hilbert Space ; Markov chain Monte Carlo ; pathways selection ; genes expression

Abstract

Much attention has been given to the development of methods that utilize the large quantity of genetic information available in online databases. Recently a new philosophy emerged which considers the genetic pathways, which contain sets of genes, combined effect on a disease. Under the new philosophy the goal is to identify the significant genetic pathways and the corresponding influential genes in regards to different diseases.

In this research, a Bayesian kernel machine model which incorporates existing information on pathways and gene networks in the analysis of DNA microarray data is developed. Each pathway is modeled nonparametrically using a reproducing kernel Hilbert space. Mixture priors on the pathway indicator variable and the gene indicator variable are assigned. This approach can be used to model both linear and non-linear pathway effects and can pinpoint the important pathways along with the active genes within each pathway. An efficient Markov Chain Monte Carlo (MCMC) algorithm is developed to fit our model. Simulation studies and a real data analysis, using, van't Veer et al. (2002) breast cancer microarray data, are used to illustrate the proposed method.