Abstracting Genes to Gene Ontology Terms Allows Comparison across Multiple Species

Comparing lists of genes generated from microarray analyses is tedious and subjective. The numbers of genes that change can be in the thousands and each gene typically has multiple functions. Clustering approaches simply produce smaller groups of genes. Furthermore, comparing experiments across species interjects an additional complication to data analysis — the genomes are not the same, making direct gene comparisons difficult or impossible. We were able to partially address this issue by abstracting the genes to gene ontology (GO) terms and testing for significance using a hypergeometric distribution function. To test the algorithm, mouse, rat, and dog were treated with LPS, a bacterial cell wall component that causes an acute phase response, and gene expression changes were compared. The GO terms are displayed using a hyperbolic viewer, allowing the biologist to navigate the data in a known framework. A GO term can be readily identified as significantly enriched in all 3 species, although there are relatively few identical genes.