The Use of Go Terms to Understand the Biological Significance of Microarray Differential Gene Expression Data

Abstract

We show one way of using Gene Ontology (GO) to understand the biological relevance of statistical differences in gene expression data from microarray experiments. To illustrate our methodology we use the data from Pritchard et al. [2001]. Our approach involves three sequential steps: 1) analyze the data to sort genes according to how much they differ between/among organs using a linear model; 2) divide the genes based on “how much or how strongly” they differ, separating those more expressed in one organ vs. those more expressed in the other organ; 3) examine the relative frequency of GO terms in the two groups, using Fisher’s exact test, with correction for multiple testing, to assess which of the GO terms differ significantly between the groups of genes. We repeat steps 2) and 3) using a sliding window that covers all the sorted genes, so that we successively compare each group of genes against all others.

By using the GO terms, we obtain biological information about the predominant biological processes or molecular functions of the genes that are differentially expressed between organs, making it easier to evaluate the biological relevance of inter-organ differences in the expression of sets of genes. Moreover, when applied to novel situations (e.g., comparing different cancer conditions), this method can provide important hints about the biologically relevant aspects and characteristics of the differences between conditions. Finally, the proposed method is easily applied.