Bayesian Mixture Models for Complex High-dimensional Count Data

• Activity Number: 82
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yuan Ji*+
• Companies: University of Texas M.D. Anderson Cancer Center
• Keywords: MCMC ; Gibbs sampler ; Metropololis-Hastings ; Bayesian inference

Abstract

Phage display is a very useful method to study the behavior of a very large number of peptides and proteins on the surface of a small bacterial virus, called a phage. The resulting count data from phage display experiments usually possess high dimensionality and a complex correlated structure. Statistical modeling of these data are therefore challenging. The main issues involve multiple comparisons and, probably more importantly, modeling the complex correlation structure in the data, which is of major interest. We develop a class of Bayesian mixture models for such complex high-dimensional count data and propose a selection methodology for identifying peptides with distinct ascending display patterns in their counts. We construct Bayesian hierarchical priors for the parameters that are specifically designed for this type of data. Our simulation results indicate that the proposed mixture models and priors are very suitable for the count data. We present a case study in detail to demonstrate the proposed methodology.