Informed Bayesian Biclustering of Microarray Data

• Activity Number: 70
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Lisa Pham*+ and Mayetri Gupta and Surajit Ray and Eric Kolaczyk
• Companies: Boston University and Boston University and Boston University and Boston University
• Keywords: microarray ; clustering ; genes ; bayesian

Abstract

Clustering methods are frequently used in bioinformatics. In particular, standard clustering of microarray data allows us to find gene similarities across all experiments. However, genes rarely behave the same way under all conditions. Thus, it is important to cluster microarray data along both genes and experiments to study gene associations between conditions and condition associations between genes. Various methods exist to bicluster data. We extend the Bayesian plaid model proposed by Caldas et al, which uses non-informative priors on clusters, to incorporate informative priors that integrate biological information. Using our method, the formation of clusters can be guided by auxiliary data, such as specific biological pathways that are likely to be active within a group of genes sharing a general type of functionality. We show the benefits of this approach on a human data set.