Terminology- and ontology-based phenotyping (TOP)

The junior research group ‘Terminology- and Ontology-based Phenotyping (TOP)’ is part of the SMITH consortium and is funded by the German Federal Ministry of Education and Research as part of the Medical Informatics Initiative (MII).

We focus mainly on algorithmic phenotyping, that is the software-supported identification of individuals or cohorts with certain characteristics (phenotypes). Having analysed the notion of a phenotype in an ontological fashion, we developed a technique for modelling phenotypic knowledge on this basis. The phenotype models developed in this manner are structured, explainable and executable. This means that queries can be defined on their basis and executed on various data sources. Established medical terminologies such as LOINC or SNOMED play an important role in identifying relevant data in the data source. Phenotypic models can be used, for example, to recognise certain diseases, to assess risk, to avoid adverse events or to identify study participants.

Another aspect of our research is the semantic search in text documents (for example clinical reports). Here we develop NLP methods (such as for document clustering or terminology extraction) that support an ontology-based search.

We implement both components, i.e., the phenotyping on structured data and the semantic search in text documents, in our so-called TOP Framework. Using this framework, phenotype models and Search Ontologies (classification of concepts of a text search query) can be developed by experts without IT knowledge and executed on the desired data sources. The results are data sets that fulfil all the criteria of the query – in the first case, persons who exhibit the defined phenotypes, and in the second case, text documents that contain or are represented by the concepts.