III: Small: Methods for Auditing and Enhancing Completeness of Ontologies

University Of Kentucky Research Foundation, Lexington KY

Investigators

Abstract

An ontology is a formal representation of the knowledge by a set of concepts (terms) and the relationships between those concepts within a domain of specialty. Ontologies have been widely used for orchestrating the coding, management, exchange, and sharing of the increasingly large amounts of digital data produced by the biomedical research enterprise. For example, SNOMED CT, one of the largest and most complex biomedical ontologies, supports the development of high-quality electronic health records and facilitates patient subgroup identification, clinical decision support, and healthcare delivery quality measurement. Given such important roles that biomedical ontologies play, quality issues (such as incompleteness in coverage of subclasses), if not addressed, can affect the quality of diagnoses and decisions. However, incompleteness issues such as missing hierarchical relations and missing concepts are infeasible to be addressed by manual work alone due to the size and complexity of biomedical ontologies. The goal of this project is to develop automated and scalable approaches for identifying potential incompleteness issues as well as suggesting solutions to fix them. This project will incorporate the computational aspects of the proposed work into curriculum development and educational offerings related to data science and promote women participation in data science. To audit and enhance completeness of ontologies, this project explores the following research tasks: (1) Development of a robust reasoning framework for detecting and repairing missing subclass or hierarchical relations. This will result in suggestions that directly enhances the subclass completeness of ontologies; (2) Development of novel methods for identifying missing concepts and creating appropriate name labels for the identified missing concepts. This will result in enhancement in the concept completeness of ontologies; (3) Generation of supporting evidence for suggested solutions by leveraging rich extrinsic knowledge. For further verification of the robustness of the proposed approaches, domain experts will be involved in validation of the discovered incompleteness issues. The proposed approaches are applied to three large ontologies in biomedicine: SNOMED CT, Gene Ontology, and NCI Thesaurus. Suggested ontology changes will be communicated to the respective ontology owners for incorporation in subsequent versions. The project website will include further information about this project, and provide access to publications, software, datasets and curriculum material. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.