Biological Pathway Selection Through Bayesian Integrative Modeling

• 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): Lingling Zheng*+ and Joseph Lucas
• Companies: duke university and Duke University
• Keywords: Joint Bayesian factor model ; beta-Bernoulli ; pathway analysis ; gene expression data integration ; radiation

Abstract

Pathway analysis has become central in understanding the underlying biology of differentially expressed genes. As large amount of microarray data has been accumulated in public repositories, it provides a comprehensive way to generate biological hypothesis by integrating large-scale information and exploring pathway activities across diverse experimental conditions. In this study, we developed a two-level model, consisting of 1) a joint Bayesian factor model that integrates multiple microarray experiments and assumes each factor a predefined pathway 2) a beta-Bernoulli distribution that infers factors relevant/irrelevant to each dataset. Efficient computational methods are developed upon MCMC. Without the use of a phenotype, our method identifies pathways specific to a particular experiment or shared among different interventions. This provides a very powerful and open-ended analysis. Simulation and application to a radiation study on C57BL/6 mice demonstrate the efficacy of our approach uncovering irradiation-induced pathways, which shed valuable insight into the molecular basis of time- and dose- dependent response to radiation.