BioRGroup dataset: R-group expansion of ChEBI molecules referenced in the Rhea database

The application of artificial intelligence in cheminformatics highlights the necessity of comprehensive datasets that fully utilize all available chemical information. While generalist databases such as PubChem provide extensive compound coverage, specialised resources such as Rhea, which relies on the ChEBI ontology, are critical for the study of enzyme-catalysed reactions. A notable challenge arises from the presence of generic structures in ChEBI molecules, which incorporate R-groups as placeholders for various molecular fragments. This creates difficulties for their use in computational pipelines, such as those applied in retro-biosynthesis and biocatalysis. To address this issue, a curated dataset is presented that resolves R-group-containing ChEBI entries into fully defined molecular instances. The pipeline extracts generic molecules from Rhea, identifies compatible substitutions using PubChem and RDKit, and applies tailored filters to generate chemically valid enumerations. The dataset, BioRGroup, is delivered in a standard file format, thereby enabling the systematic integration of previously under-utilised generic structures into computational workflows, enhancing the scope and granularity of chemical data analysis.