Gene Regulatory Networks from Gene Ontology

Gene Ontology (GO) provides a controlled vocabulary and hierarchy of terms to facilitate the annotation of gene functions and molecular attributes. Given a set of genes, a Gene Ontology Network (GON) can be constructed from the corresponding GO annotations and semantic relations among GO terms. Transitive rules can be applied to GO semantic relations to infer transitive regulations among genes. Using information content as a measure of functional specificity, a shortest regulatory path detection algorithm is developed to identify transitive regulations in GON. Since direct regulations may be overlooked during the detection of gene regulations, gene functional similarities deduced from GO terms are used to detect direct gene regulations. Both direct and transitive GO regulations are then used to construct a Gene Regulatory Network (GRN). The proposed approach is evaluated on seven E.coli sub-networks extracted from an existing known GRN. Our approach was able to detect the GRN with 85.77% precision, 55.7% recall, and 66.26% F1-score averaged across all seven networks.