OpenBind Enteroviral 2A Protease Benchmark Data

This record contains the prepared datasets and computational benchmarking outputs associated with the OpenBind enteroviral 2A protease structure–affinity dataset. The accompanying preprint describing the dataset, benchmark construction, and results is available at https://doi.org/10.64898/2026.08.27.747600.

The underlying experimental dataset comprises crystallographic protein–ligand complexes and associated binding-affinity measurements for EV-A71 2A protease, generated using Coxsackievirus A16 (CVA16) 2A protease as a closely related experimental surrogate. The original OpenBind data release, including the experimentally determined structures and affinity measurements, is available at Zenodo record 20026660 (DOI: 10.5281/zenodo.20026660).

This record focuses on the processed and computational data used for structure-based modelling and benchmarking. It includes:

• Prepared dataset: standardized and model-ready protein–ligand structures derived from the experimental OpenBind release.
• Docking data: docking inputs and/or generated docking poses used to evaluate structure-based prediction workflows.
• Cofolding data: inputs and predictions from protein–ligand cofolding workflows used to assess the ability of current structure-prediction methods to recover experimentally observed binding modes.

Together, these files provide the computational counterpart to the experimental OpenBind release and are intended to support benchmarking, method development, error analysis, and comparison of docking and cofolding approaches on a dense, experimentally characterized protein–ligand dataset.