Performance of Different Filtering Methods When Testing for Differential Expression in Microarray Data

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Soledad Fernandez*+ and Parul Gulati and David Jarjoura and Lianbo Yu and Michael Pennell
• Companies: The Ohio State University and The Ohio State University and The Ohio State University and The Ohio State University and The Ohio State University
• Keywords: microarray ; multiple testing ; differential expression ; filtering

Abstract

Microarray data analysis involves a large number of hypothesis tests and hence requires multiple testing adjustments. This leads to low power when sample sizes are not large. Filtering reduces the number of tests and can lead to higher power. A large proportion of genes are unexpressed in tissue samples for microarray experiments, resulting in observed values that are the result of the random variation associated with processing chips rather than gene expression. Researchers commonly filter out genes with low mean signal or low variance. These approaches could be problematic with small sample sizes where gene-specific variance estimates are unreliable and mean signal is more sensitive to outliers. We propose a simple filtering procedure which is less sensitive to small sample sizes. The superior performance of the proposed procedure is demonstrated in a simulation study.