Comparative effectiveness, clinical translational research, and personalized medicine depend on the accessibility of data and information. This dissertation introduces an ontology-driven architecture for data federation. Ontological systems such as SNOMED CT have been traditionally used for data exchange, to preserve meaning for data elements across systems. We present a novel method of using ontologies to (1) enable data integration from distributed sources (2) facilitate a federated query interface. The contributions of this dissertation are trifold.
First, we developed a unified model consisting of a sleep domain ontology and a units ontology into which data sources can be mapped using an intuitive Data Source to Ontology Mapper (DSOM). The DSOM allows for terminology standardization while accommodating local nomenclature. The domain ontology enriches the expressiveness of complex queries sent across data sources, based on graph traversal and ontological equivalence relationships. We provide further extensions to the ontology framework using formulae, for the purpose of inferring, discovering, and validating data across disparate sources.
Second, we developed a powerful query interface called the VISual Aggregator and Explorer (VISAGE) for building and executing federated queries across mapped data sources. The interface streamlines the process of identifying case-control patient populations for study design through the use of embedded statistical tools. A framework is provided for the researcher to choose matching criteria and allows the researcher to create control sets based on frequency matching.
Third, we implemented a fully operational and user-friendly system based on the ontology-driven architecture, as part of Physio-MIMI, an NCRR-funded multi-CTSA-site project. Using the Agile methodology to facilitate multi-site collaboration, our implementation leverages a suite of latest software development technologies: Ruby On Rails, Prototype JavaScript, AJAX, and Highcharts for a dynamic Web 2.0 interface, Mutual SSL for security and privacy, semantic web ontologies
using OWL, and ontology development environments such as Protege. Preliminary comparative analyses and user evaluations of aspects of the system are provided, demonstrating superior results in several usability categories.