Scoring protein-protein interactions using the width of gene ontology terms and the information content of common ancestors

Abstract

Several methods have been proposed to measure the semantic similarity of proteins. In particular, the Gene Ontology (GO) is often used to estimate the semantic similarity of proteins annotated with GO terms since it provides the largest and reliable vocabulary of gene products and their characteristics. We developed a new measure for semantic similarity of proteins involved in protein-protein interactions using the width of GO terms and the information content of their common ancestors in the GO hierarchy. A comparative evaluation of our method with other GO-based similarity measures showed that our method outperformed the others in most GO domains. © Springer-Verlag Berlin Heidelberg 2013.