140 How can we maximise discovery potential from sparse clinical information?

Abstract

Collaborative studies exploit large data sets to discover new associations with diseases. However, analyses of clinical features struggle with big data. Beyond sheer volume, the challenge is in harmonisation of terminology and the level of detail of information provided for analysis by different sources; in epilepsy this is most pertinent to seizures.We applied two ontologies (representations of the relationship between different types of seizure) to the same sparse details from 82 individuals. We compared the amount of information imputed using the Human Phenotype Ontology (HPO, a reference of clinical phenotypes widely used by geneticists) and an ontology of seizures created by the Epilepsiome Task Force of the International League Against Epilepsy Genetics Commission (based upon ILAE classifications).We translated each individual’s seizure descriptions into 2 terms (median) from each ontology. Imputa- tion generated 2 additional terms per individual using the HPO and 4 using the Epilepsiome ontology. The Epilepsiome ontology was able to increase the amount of information available for analysis by two- thirds compared to the HPO.Well-designed ontologies can be used to impute descriptions of complex clinical features at multiple less specific levels of precision, addressing a major limiting step in harmonisation of data for collabora- tion or re-use.rhys.thomas@ncl.ac.uk