Protein Language Models and Structure-Based Machine Learning for Prediction of Allosteric Binding Sites in Protein Kinases: An Explainable AI Framework Guided by Energy Landscape–Derived Frustration

This Zenodo archive provides the datasets, processed prediction outputs, evaluation results, figures, and scripts used in the study “Protein Language Models and Structure-Based Machine Learning for Prediction of Allosteric Binding Sites in Protein Kinases: An Explainable AI Framework Guided by Energy Landscape–Derived Frustration”. The archive mirrors the directory structure expected by the analysis pipeline and contains two dataset variants: KinCoRe, which includes all protein–ligand complexes used for the main analyses reported in the manuscript. KinCoRe-UNQ, a reduced variant retaining a single representative structure per protein (UniProt ID), which was used for Supporting Information analyses to control for over-representation of proteins with multiple solved structures. All results presented in the manuscript are derived from the KinCoRe dataset, while KinCoRe-UNQ is used for analyses reported in the Supporting Information. Protein structures are not included in this archive. All structures are publicly available from the RCSB Protein Data Bank and can be automatically downloaded using the provided scripts.