An information theoretic approach to assessing Gene-Ontology-driven similarity and its application

Using information-theoretic approaches, this paper presents a cross-platform system to support the integration of Gene Ontology (GO)-driven similarity knowledge into functional genomics. Three GO-driven similarity measures (Resnik's, Lin's and Jiang's metrics) have been implemented to measure between-term similarity within each of the GO hierarchies. Two approaches (simple and highest average similarity) which are based on the aggregation of between-term similarities, are used to estimate the similarity between gene products. The system has been successfully applied to a number of applications including assessing gene expression correlation patterns and the relationships between GO-driven similarity and other functional properties.