At What Scale Should Microarray Data Be Analyzed?

• Activity Number: 274
• Type: Contributed
• Date/Time: Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shuguang Huang*+ and Kerry G. Bemis
• Companies: Eli Lilly and Company and Eli Lilly and Company
• Keywords: microarray ; t-test ; robustness ; power ; transformation ; normality

Abstract

The motivation for transforming microarray data in one way or another is usually to satisfy the model assumptions such as normality and/or homoscedasticity. Generally, two types of strategies are often applied to microarray data depending on the analysis need. One strategy, such as correlation analysis, considers all the gene intensities on the array simultaneously; the other, such as gene-by-gene ANOVA, analyzes each gene individually. We investigate the distributional properties of the Affymetrix genechip signal data under the two scenarios, with the focus on the impact of analyzing the data at an inappropriate scale. When all the genes on the array are considered together (pooled), the dependent relationship between the expression and its variation level can be satisfactorily removed by Box-Cox-type of transformation. When genes are analyzed individually, the distributional properties of the intensities are shown to be gene-dependent. Derivation and simulation show that some loss of power is incurred when a wrong scale is used, but due to the robustness of the t-test, the loss is rather acceptable when the fold-change is not very big.