Presentation given at the Ontologies4Chem Workshop 2025. Abstract: Most of the ontologies used in the biosciences are large, lightly axiomatized hierarchical terminologies, often consisting of tens of thousands of terms. The developers of these ontologies have learned the hard way that ensuring consistency of classification and connectedness is a huge challenge, unless consistent design patterns are applied. This is as true of ChEBI as it is with many other similar ontologies, and with the rollout of the ChEBI 2.0 infrastructure, now is the time to be thinking about applying higher level metaclass organization and design patterns. The CHEMROF data model was designed as both a semantic schema for chemical entities and as a metaclass-level schema for ontologies such as CHEBI, providing organization units for entities such as atoms, polyatomic entities, macromolecules, salts, racemic mixtures, both at the structural level and at the grouping class level. CHEMROF also includes classes for modeling reactions, including biochemical reactions of the kind found in RHEA. In this presentation I will describe the overarching structure of CHEMROF, how it can be used as an exchange format for chemical knowledge, and how it can help with the future evolution of ChEBI. I will also describe how we are applying the ai4curation agents to assist with the development of CHEMROF and related LinkML schemas.
Christopher Mungall (Thu,) studied this question.