Abstract
When an ontology-based knowledge graph (a KG, containing information about genes, pathways, phenotypes, and molecular interactions) is analyzed using a biomedical computational reasoning system, the system can return unwanted generic concepts. Such generic concepts, in the context of a query about potential drug treatments for a disease, might include "pharmaceutical preparations" or "antibiotics". Filtering answers based on hand-curated block-lists or ontology Information Content (IC) is a remedy, but to our knowledge, (1) neither method has been validated against expert judgments of generality, and (2) IC is not available for the majority of nodes in the representative KG used in this work. In this study, three blinded domain experts rated representative lists of KG concepts for generality. We built a semantic precision score (SPS) from a concept's name and description alone, using a biomedical language model (BioLORD). For the purpose of filtering out generic answers from the results returned from biomedical queries, we compared, using queries of 200 diseases, a hand-curated concept block-list and two IC thresholds. IC agreed moderately with the experts (Spearman ρ = 0.59) with little or negative resolution within IC bands. The SPS reached ρ = 0.76, kept its order within every band, and remained informative on the 50 concepts without IC (ρ = 0.59). At its decision boundary, the semantic filter removed 1,502 of 33,161 query answers, nearly three times the block-list's 513, while removing fewer of the query-specific DrugCentral indications. IC thresholds removed either few answers or many approved drugs. Because these lists also informed model selection, we repeated the comparison on 200 additional diseases with the model frozen. The SPS filter removed 5,135 of 91,203 answers against the block-list's 1,419, lost none of 1,296 indications against 17, and removed approved drugs no more often from the half of DrugCentral it had never seen than from the half it was trained on. Semantic meaning captures expert judgments of specificity more faithfully than ontology position, and semantic scoring can replace IC and manually maintained block-lists for filtering drug-repurposing results.