Overcome the Small Sample Size Challenge in Differential Gene Expression Analysis Studies

• Activity Number: 131
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yan Huang*+
• Companies: University of California at Berkeley
• Keywords: differential gene expression analysis ; aggregated ranks ; gene rankings ; small sample size ; filtering

Abstract

In differential gene expression analysis studies (DGEAS) we often aim at obtaining a good estimate of a gene rankings based on differential expressions (DE). Genes that are up- or down-regulated under a set of biological or experimental conditions are particularly interesting as they serve as bio-markers for early-detection of diseases and are valuable for further studies. One of the common challenges in DGEAS is that due to the small number of replicates, the unstable DE variance estimates often lead to unreliable values for the test statistics (e.g., in the case of t-test). Therefore, without any modifications on the variance estimates, rankings obtained through test statistics are often inaccurate. There are 2 types of resources that could be useful to improving this problem: 1) the large number of genes in a study. 2) the high volume of data from studies of the same topic emerging in recent years. We proposed a method that simultaneously utilizes the two types of resources aforementioned and compared our method to several recent well-known gene-ranking methods.