Bayesian Models for Genetic Pathways

• Activity Number: 222
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yuan Ji*+
• Companies: MD Anderson Cancer Center
• Keywords: Graphical models ; Mixture model ; Network ; Protein array ; Reciprocal graph

Abstract

We define a class of Bayesian graphical models for the inference of molecular pathways based on expression data. Under a Bayesian framework, our models update a prior consensus pathway using expression profiles and provide posterior estimates of protein-protein interactions that are specific to the study population. Specifically, we define mixture models for the observed expression data which further introduce latent states of the expression. We construct informative network prior models for the latent states to quantify the prior information of consensus pathways. For inference, the posterior network is presented as reciprocal graphs with directionality and strength for each protein-protein interaction. We demonstrate the models using a novel type of proteomics data based on reverse phase protein arrays (RPPA).