Columbia University Health Sciences and Maastricht University
TReK will provide to the Translator project several new or enhanced knowledge sources as well as workflows for data transformation and integration. Leveraging Columbia University Irving Medical Center's mature clinical data warehouse, we will provide several novel clinical knowledge resources, each one bringing different capabilities to Translator. We will enhance Columbia Open Health Data (COHD) [1] to broaden the scope of clinical knowledge and strengthen the confidence of discovered associations by adding, for example, temporal analyses, patient lab values and genetic data, and clinical knowledge extracted from clinical notes. We will develop analytical methods using representation learning to gain new insights into diseases. We will create a Knowledge Provider of PICO (Population, Intervention, Comparison, Outcome) elements from randomized clinical trial publications and compare real-world evidence to evidence in the literature in a scalable fashion. We will use the DSRI OpenShift cluster at Maastricht University to sustainably transform clinical data sources administered by Columbia University. The data will be exposed as a Reasoner API and a SPARQL endpoint that complies with the BioLink model. These services will be shared in the Translator ecosystem and will be generalizable to other data sources.