Identification of Differentially Expressed Genes: Two Bayesian Approaches

• Activity Number: 348
• Type: Topic Contributed
• Date/Time: Wednesday, August 10, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Lisete de Sousa*+ and Maria A. Turkman and Luka Clarke and Margarida Amaral
• Companies: University of Lisbon and FCUL, DEIO and FCUL, DQB and FCUL, DQB
• Keywords: Microarrays ; Bayesian approach ; posterior odds ; Monte Carlo methods

Abstract

The level of DNA transcripted to RNA in a normal organism and in a mutation may be compared using global techniques for gene expression (genomics) as microarrays. This work aims the development of methods to detect differentially expressed genes in cell lines collected from patients with Cystic Fibrosis (CF), relative to control cell lines (nonCF). Experiments in large-scale microarrays, representing essentially the entire human genome (40K), were performed comparing the RNA expression patterns between three cell line pairs (CF vs nonCF). Due to complexity of the data, the Bayesian approach appears to have all the ingredients necessary to properly deal with microarray data or any other kind of genetic data. We present two empirical Bayesian approaches as alternatives to the nonparametric method. In spite of both being similar approaches, one assumes the normality of the log-expression levels, while the other is based on a Gamma distribution on the expression levels---BEstDEG.