A Bayesian Extension of the Hypergeometric Test for Functional Enrichment Analysis

• Activity Number: 604
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: International Society for Bayesian Analysis (ISBA)
• Author(s): Jing Cao*+ and Song Zhang
• Companies: Southern Methodist University and The University of Texas Southwestern Medical Center
• Keywords: Functional Enrichment Analysis ; non-central hypergeometric distribution ; Gene Ontology

Abstract

Functional enrichment analysis (FEA) is used in high-throughput experiments to provide functional interpretation for lists of differentially expressed (DE)genes. The hypergeometric P-value is often used in FEA to investigate whether genes from functional groups, eg, as represented by Gene Ontology (GO) annotations, are over-represented in the DE gene list. However, it has 3 limitations: 1) it is computed independently for each GO term and neglects the interrelationship among GO terms; 2) it has a size constraint, ie, a lower limit determined by the size of GO term, making it biased towards selecting larger GO terms; and 3) overlapping genes in GO terms are repeatedly used. We propose a Bayesian model based on the non-central hypergeometric distribution to overcome the limitations. The GO dependence structure is incorporated through a prior on the non-centrality parameter. The resulting enrichment measure is a posterior probability which does not have the size constraint. Also, the overlapping information is removed from the likelihood function. We show that this method is more efficient in identifying sets of biologically-meaningful GO terms rather than individual isolated ones.