A novel method for expanding current annotations in gene ontology

Since the gap between the amount of protein sequence data and the reliable function annotations in public databases is growing, characterizing protein functions becomes a major task in the post genomic era. Some current ways to predict functions of a protein are based on the relationships between the protein and other proteins in databases. As a large fraction of annotated proteins are not fully characterized, annotating novel proteins is limited. Therefore, it is of high demand to develop efficient computation methods to push the current broad function annotations of the partially known proteins toward more detailed and specific knowledge. In this study, we explore the capability of a rule-based method for expanding the current annotations per some function categorization system such as Gene Ontology. Applications of the proposed method to predict human and yeast protein functions demonstrate its efficiency in expanding the knowledge space of the partially known proteins.