A0828
Title: A Bayesian multiple testing procedure infused with historical data, with application in gene expression testing
Authors: Ya Su - Virginia Commonwealth University (United States) [presenting]
Abstract: The process of identification of expressed genes between experimental groups is a difficult task due to the heteroscedasticity across a massive number of genes. This problem is addressed using a Bayesian framework, which facilitates the incorporation of prior knowledge obtained from different platforms or organisms. A new test statistic and sign-adjusted FDR that emphasizes information regarding the direction of the differentially expressed genes are proposed. The statistic is proven to achieve the highest count of true positives compared to all legitimate sign-adjusted false positive controlled methods. Simulation results provide numerical evidence. The approach is demonstrated through the analysis of two gene expression datasets.
