Abstract
Misassigned Mg2+ ions are pervasive in RNA structural databases, obscuring mechanistic interpretation, undermining comparative analyses, and compromising machine-learning training sets. Here, we present Cat_Wiz, a Coot-integrated, stereochemistry-guided toolkit that facilitates the localization, diagnosis, correction, and annotation of Mg2+ binding sites. Cat_Wiz comprises three modules: MG_diagnosis,which validates and regularizes existing assignments; MG_detect, which identifies unmodeled ion binding sites; and MG_clamp, which classifies recurrent Mg2+ clamp motifs. Cat_Wiz also includes a complete binding site annotation system. The stereochemical principles implemented in Cat_Wiz were derived from an earlier analysis of the 1.55 Å resolution Escherichia coli ribosome and from surveys of the Cambridge Structural Database (CSD). These principles provide a robust experimental foundation for characterizing Mg2+ binding sites. Applications to ribosomes, hammerhead ribozymes, group I introns, and quaternary RNA assemblies demonstrate that Cat_Wiz rapidly locates overlooked ions, corrects misassignments, and improves stereochemical fidelity in hours rather than days. Beyond refinement, Cat_Wiz generates curated data that can seed diverse machine learning and artificial intelligence (AI) models. This transparent, cost-effective framework establishes reproducible standards for RNA-ion assignments and will drive progress in the design of RNA 3D architectures through the identification of unique Mg2+-dependent backbone folds. Cat_Wiz, that is based on universal stereochemical principles, applies also to Mg2+ binding sites in proteins and related biomolecular systems.