Use of the Half-Normal Probability Plot to Identify Significant Effects for Microarray Data

• Activity Number: 111
• Type: Contributed
• Date/Time: Monday, August 12, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section*
• Author(s): Greg Dyson*+ and Jeff Wu
• Affiliation(s): University of Michigan and University of Michigan
• Keywords: Microarray ; Multiple Comparisons ; Gene Expression ; Robust ; Change-Point

Abstract

We propose a novel technique, motivated by methodologies for analysis of designed experiments, to identify significant effects in replicated microarray experiments. Given a robust measure of location and scale for each gene, a half-normal probability plot of a comparative test statistic is constructed. We then fit robust regression lines to sections of the half-normal plot to determine the location of the "change-points." This procedure only requires a small number of comparisons in order to declare a larger subset of genes to be significant. Additionally, it is flexible in that it can give the scientist input on a reasonable number of genes to be declared significant. Illustration of this methodology, as well as extensions to other types of gene expression data, will be discussed.