A Bayesian Nonparametric Method for Differential Expression Analysis of RNA-Seq Data

• Activity Number: 347
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yiyi Wang*+
• Keywords: Bayesian discovery procedure ; Gene Ontology

Abstract

We developed a method for RNA-seq data to identifying differentially expressed genes based on the Bayesian Discovery Procedure of Guindani, et. al. (2009). Our model provides two novelties: 1) we use a negative binomial sampling model, and 2) we replace the usual random partition prior from the Dirichlet process with a random partition prior indexed by distances from Gene Ontology (GO) annotations. We show that the use of GO annotations in the clustering prior improves statistical power over the original Bayesian Discussion Procedure. For any set of genes having high probability of differential expression, the estimated false discovery rate is computed. Thresholds can be adjusted to achieve a desired estimated false discovery rate. We demonstrated with an actual dataset.