Eigenvalue Ratios Reveal Shared Binding Pocket Shapes in RNA and Protein Structures.

Molecular recognition in drug design relies on accurate characterization of ligand-binding pockets on macromolecular targets such as proteins and RNA. While protein binding sites have been extensively described, the geometric organization of RNA pockets remains comparatively underexplored. Here, we introduce a unified and size-independent geometric framework to describe and compare RNA and protein binding pocket shapes. Pocket geometry is captured using size-independent and residue-agnostic measures of global anisotropy, enabling direct comparison across RNA and protein binding pockets without introducing macromolecule-specific assumptions. This approach defines 4 interpretable pocket shape archetypes: sphere-like, rod-like, disk-like, and strongly anisotropic. Application to balanced datasets of 300 RNA and 300 protein binding pockets reveals a largely shared geometric landscape, with substantially overlapping shape descriptors within each archetype. However, archetype frequencies differ: Sphere-like pockets are more frequent in proteins, whereas disk-like and strongly anisotropic pockets are enriched in RNA, while rod-like pockets occur at comparable frequencies. Notably, strongly anisotropic pockets lacking a dominant symmetry axis represent a substantial fraction of pockets in both datasets. By organizing diverse binding sites into a small number of reproducible geometric regimes, this framework reduces structural heterogeneity and provides a transferable geometrical reference for comparative analysis of RNA and protein pocket architectures, thereby supporting the exploration of RNA pocket accessibility in structure-based studies.