Bayesian Inference of Multiple Gaussian Graphical Models

• Activity Number: 676
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Christine Peterson*+ and Francesco Stingo and Marina Vannucci
• Companies: Rice University and UT MD Anderson Cancer Center and Rice University
• Keywords: graphical model ; network inference ; informative prior

Abstract

We address the problem of inferring multiple Gaussian graphical models which may share common features while differing in scientifically important respects. In our approach, we place a Markov Random Field prior on the network structure for each sample group that both encourages similar structure between related groups and accounts for reference networks established by previous research. This formulation improves the reliability of the estimated networks by allowing us to borrow strength across related sample groups and encouraging similarity to a known network. In addition, we obtain a summary of network similarity which allows us to compare how strongly the network structures in various sample groups are related. Applications include the inference of protein-protein interaction networks for multiple cancer subtypes and inference of gene regulatory networks under different sample conditions.