Researchers in the Genomes to Life program, as well as in other biomedical research programs, need a more effective way to search across the multiple heterogeneous databases that are compiling information approaching petabyte levels. Current approaches are limited because they are based on heuristics and keyword indexing and rely on manual annotation. This project addresses the problem by leveraging the expressive power of ontologies, augmented with inter-ontological links to derive similarities between gene products. These similarities are then used to empower search and cross-scale discovery. The essence of the innovation consists in deriving a semantic similarity between ontology terms, taking into account the semantic, syntactic, and structural information available in protein databases. Phase I will demonstrate the feasibility of automatic cross-referencing of gene ontologies and develop a retrieval algorithm that leverages weighted cross-referenced gene ontologies for the discovery of proteins among select groupings of microbial genomes. A prototype tool will be developed and compared with existing methods for querying microbial databases, in order to determine the effectiveness of the proposed method to mine biological data.
Commercial Applications and Other Benefits as described by the awardee: A tool that would better enable knowledge transfer by peering across ontological domains should have application to the exploitation of microbes to accelerate environmental cleanup and to improved disease prevention and control. Beyond biological research, users may range from legal practitioners, the military, educational institutions, economists, law enforcement and homeland security agencies, e-commerce, and business-to-business transactions over the Internet
