Joint Modeling with Bayesian Clustering and Bayesian Networks for Time Course Microarray Data

• Activity Number: 181
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yafeng Zhang*+ and Donatello Telesca and Steve Horvath
• Companies: University of California at Los Angeles and University of California at Los Angeles Fielding School of Public Health and University of California at Los Angeles
• Keywords: Nonparametric Bayesian methods ; Bayesian clustering ; Bayesian networks ; Functional data models ; Time course microarray data

Abstract

In this work, a model is proposed to perform clustering and network construction in a joint fashion for time course microarray data. Data profiles are modeled by a functional data model where the shape and time transformation functions are represented by Bsplines. The parameters of the shape functions are modeled by a product partition model which allows data profiles with a common shape but possibly different timing features to be clustered into the same group. Bayesian networks are adopted to construct directed networks among formed clusters. The model unifies clustering and network construction by representing the distribution of the parameters of shape functions using a matrix normal model. A Markov chain Monte Carlo algorithm is designed to carry out posterior simulation of the clustering and networks coherently. Simulation study shows that the proposed model is able to recover pre-specified clusters and their network structure in the simulated data. This model will be applied to genomics data (gene expression, DNA methylation) to study the relationships between genes and gene products.