The first OpenBind release: An open experimental structure-affinity dataset and benchmark for structure-based AI
- 1University of Oxford
Abstract
High-quality experimental datasets that link protein-ligand structures with binding affinity data are essential for developing and evaluating structure-based machine learning methods. To help address this need, we established OpenBind as an open-science initiative to generate large-scale experimental datasets for structure-based AI and molecular discovery. Here, we describe the first public OpenBind release, which, to the best of our knowledge, is the largest public single-target experimental structure-affinity dataset. The dataset focuses on enteroviral 2A protease, comprising 925 crystallographic binding events from 699 compounds and associated affinity measurements for 601 compounds. It combines structures from an initial fragment screen and follow-on molecules, together with affinity data, linking experimentally determined protein-ligand binding modes to biophysical measurements within a coherent antiviral discovery campaign. We used this dataset to evaluate protein-ligand structure prediction, binding-affinity prediction, and virtual screening using representative structure-based methods, including docking and cofolding. This exposed several challenges that are central to practical structure-based modelling: docking performance depends strongly on binding-pocket conformation, poses are difficult to rank, and structure-based affinity prediction remains challenging. Fine-tuning OpenFold3-p2 on the fragment-screen structures substantially improved pose prediction and virtual screening for related follow-on compounds, demonstrating how early-stage experimental structures can support target-specific model adaptation.
Related
- Structure-Guided Discovery of High-Affinity Methotrexate Aptamers for Electrochemical Drug Monitoring
- Boltz2-Notebook: An Interactive Google Colab Platform for Diffusion-Based Biomolecular Structure and Binding Affinity Prediction using the Boltz2 model.
- Physical priors improve performance of structure-based binding affinity models
- RNA3D-Decoy: A Large-Scale Benchmark Dataset of AI-Generated RNA Structure Models for Quality Assessment
- Separating benchmark discrimination from experimental retrieval in synthetic-lethality prediction
BibTeX
@misc{nelen2026first,
title = {{The first OpenBind release: An open experimental structure-affinity dataset and benchmark for structure-based AI}},
author = {J. Nelen and O. Khan and E. Adams and J. C. Aschenbrenner and W. Thompson and A. Ebrahim and E. Capkin and C. Vallee and OpenBind and E. J. Shotton and E. J. Griffen and J. D. Chodera and C. M. Deane and F. von Delft and M. AlQuraishi and F. Imrie},
howpublished = {bioRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.27.747600},
url = {https://www.biorxiv.org/content/10.64898/2026.08.27.747600v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - The first OpenBind release: An open experimental structure-affinity dataset and benchmark for structure-based AI
AU - J. Nelen
AU - O. Khan
AU - E. Adams
AU - J. C. Aschenbrenner
AU - W. Thompson
AU - A. Ebrahim
AU - E. Capkin
AU - C. Vallee
AU - OpenBind
AU - E. J. Shotton
AU - E. J. Griffen
AU - J. D. Chodera
AU - C. M. Deane
AU - F. von Delft
AU - M. AlQuraishi
AU - F. Imrie
PY - 2026
DA - 2026/08/27
PB - bioRxiv
DO - 10.64898/2026.08.27.747600
UR - https://www.biorxiv.org/content/10.64898/2026.08.27.747600v1
AB - High-quality experimental datasets that link protein-ligand structures with binding affinity data are essential for developing and evaluating structure-based machine learning methods. To help address this need, we established OpenBind as an open-science initiative to generate large-scale experimental datasets for structure-based AI and molecular discovery. Here, we describe the first public OpenBind release, which, to the best of our knowledge, is the largest public single-target experimental structure-affinity dataset. The dataset focuses on enteroviral 2A protease, comprising 925 crystallographic binding events from 699 compounds and associated affinity measurements for 601 compounds. It combines structures from an initial fragment screen and follow-on molecules, together with affinity data, linking experimentally determined protein-ligand binding modes to biophysical measurements within a coherent antiviral discovery campaign. We used this dataset to evaluate protein-ligand structure prediction, binding-affinity prediction, and virtual screening using representative structure-based methods, including docking and cofolding. This exposed several challenges that are central to practical structure-based modelling: docking performance depends strongly on binding-pocket conformation, poses are difficult to rank, and structure-based affinity prediction remains challenging. Fine-tuning OpenFold3-p2 on the fragment-screen structures substantially improved pose prediction and virtual screening for related follow-on compounds, demonstrating how early-stage experimental structures can support target-specific model adaptation.
ER -