iNOA

Abstract

Gene Ontology (GO) consortium provides the largest functional annotations of genes, and GO enrichment analysis is almost regarded as a standard-like analytical method in computationally biomedical researches. Network biology reveals that molecular interactions rather than individual genes perform their functions in cells, which makes network ontology analysis (NOA) of gene functions come into being. In NOA, the edge of a biomolecular network is annotated by the shared GO terms of interacting genes that proved to be better in function enrichments. However, semantic similarity is ubiquitous between GO terms. The annotations that simply endue edges without considering semantic relationship will deprive lots of crucial information of GO terms. Here, we explore an improvement of NOA with GO semantic similarity (iNOA). iNOA first retrieves the functions of all involved genes from the GO database, and then calculates the semantic similarity of the annotated terms in two genes as that in NOA. We further implement the semantic similarity based on GO graph in the measurement. The semantic similarity value calculated for two GO terms replaces the original counting number of the same GO terms annotated to the two genes of an edge. Then we apply a hypergeometric test to enrich important functions in a network. To prove the effectiveness of iNOA, a specific endometrial cancer gene network is constructed. The enriched results indicate iNOA can obtain more effective and more specific GO functions than NOA and the other gene list methods.