Fast And Easy Mapping Of Relational Data To Rdf For Rapid Learning Health Care

This abstract describes a system that supports the creation and validation of relational data to ontologies. The proposed system is developed for rapid learning health care (RLHC), i.e., an evidence based approach to train cancer prediction models on clinical care data that is stored in multiple networked hospitals. Since clinical care data cannot leave the hospital due to privacy issues, distributed (machine) learning can be used, where the models are transferred, rather than the actual patient data. This requires clinical care data to be represented in a findable, accessible, interoperable and reusable (FAIR) manner. The proposed system uses an approach called Ontology Based Data Access (OBDA) to query the hospital data. To this end the relational data is mapped to a conceptual layer in the form of ontologies, i.e., shared vocabularies. These ontologies represent the meaning and values of the data while hiding the complicated structure of the original data sources. Currently, the creation of these mappings forms an obstacle in RLHC, as it requires considerable effort and knowledge from users, and validation of the mappings is difficult.