A new similarity measure over Gene Ontology with application to protein subcellular localization

The Gene Ontology provides a controlled vocabulary to unify the presentation of gene and gene product attributes across species and genomes. It is widely used in biological data analysis and supported by popular biological databases. How to measure the relationship between GO terms has become a hot topic nowadays. In this paper, we propose a new method to measure the semantic similarity between Gene Ontology terms. This method is based on information content, and takes full advantage of structural information. We apply the method to protein subcellular location prediction. The results show that our algorithm outperforms the state-of-the-art methods.