On Meta-Analytic Procedures for Pooling the Results of Several Microarray Studies
by
E. Olusegun George
The University of Memphis
Coauthors: Gaolin Zheng and Giri Narasimhan

It is becoming increasingly common to have several centers perform scientific investigations of the same scientific problem, using different, often incompatible microarray platforms. For example, some of the studies might use cDNA spotted arrays, while others use Affymetrix probe sets. The typical methods employed for pooling is to combine the P-values of each gene from each center, and use the P-value of the combination statistics as the gene-specific consensus P-value for that gene. These consensus P-values are then adjusted for false discovery or family wise error using one or more of the existing procedures. Among the procedures for combining the P-values, an omnibus procedure due to Fisher(1932) is most popular, although the Logit method of George and Mudholkhar (1983) and Mudholkar and George (1979) has also recently been gaining some popularity (see for example Hwang et al. 2005 ). These two procedures have excellent power properties and are Bahadur efficient. However, in the context of microarray studies, it is not very clear whether, for each gene, the adjusted P-values should be combined rather than unadjusted P-values. Further, because of possible correlations between genes the combination statistics may be correlated. We will discuss the cons and pros of using the adjusted P-values in the combination procedures, as well as the advantages of incorporating consideration for correlations into the gene selection process. We will illustrate the our results by combining the P-values of microarray studies that test for periodicity in Saccharomyces cerevisiae –Budding yeast (Cho, et al. 1998) and Schizosaccharomyces pombe.(Olivia, et al. 2005), based on tests of periodicity by Wichert et al. (2004) and De Lichtenberg et al. (2005), and compare our results with those of Zheng et al. (2006), which are based on the unadjusted P-values.

Date received: March 18, 2008

Copyright © 2008 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # caxa-10.