Abstract
In this study we constructed rules to systematically identify the differences as well as similarity in partonomy between two large biomedical ontologies. For a group of one-to-one mapped concepts, we distinguish among different types of parts defined for these mapped concepts, and the structural imbalance between two systems. As a result, 1.4% of the concept mappings have exactly the same part-whole relations, 68.4% do not have any parts, and 30.2% mappings are modeled differently, including additional parts, more detailed part paths, and different intermediate concepts in the part paths. This is a parallel study to our previous work where granularity mismatches in is-a classification among ontologies were identified. Using rules and the rule inference engine enables an automatic and scalable investigation of the structural incompatibility among biomedical ontologies.