PreMedOnto: A Computer Assisted Ontology for Precision Medicine

Abstract

This paper proposes an ontology learning framework that combines text mining, information extraction and retrieval. The proposed model takes advantage of existing structured knowledge by reusing terms and concepts from other ontologies. We further apply the methodology to create a detailed ontology for the emerging precision medicine (PM) domain by collecting a corpus of relevant articles and mapping its frequent terms to existing concepts. The resulting ontology consists of 543 annotated classes. The ontology was also tested for effectiveness by applying two evaluation frameworks to validate its design and quality. The results demonstrate that the ontology learning system is able to capture and represent the semantics of the PM domain with high precision and significance. Moreover, the computer-assisted construction process reduced dependency on expert knowledge. The developed PreMedOnto ontology could be further used to enhance the potentials of other NLP applications in the PM domain.

Text Mining for Precision Medicine

“…The weights configuration for the recommender scoring function was set to the default settings. The final ranking of ontologies to be reused was: National Cancer Institute Thesaurus (NCIT) 2 , Medical Subject Headings (MeSH) 3 and Interlinking Ontology for Biological Concepts (IOBC) 4 . From the selected ontologies, we import all candidate classes with their ancestors, and verify that all remaining concepts per cluster are included in the module as child nodes.…”