Prediction of large conformational changes of a protein binding pocket associated with ligand binding

Docking simulation, a key technique in virtual screening, typically treats proteins as rigid bodies. However, proteins are inherently flexible, and ligand binding can induce significant conformational changes, affecting prediction accuracy. This study proposes a new approach to identify protein binding pockets that exhibit substantial conformational changes upon lig-and binding, potentially impacting docking simulation accuracy. In this research, we developed a prediction model using graph neural network to identify protein pockets with large conformational changes. To train the model, we constructed a dataset by calculating conformational changes in ligand-binding sites between multiple holo structures corresponding to the apo structure. We evaluated the performance of the prediction model and the results demonstrated that our model could iden-tify proteins with significant conformational changes, although the prediction accuracy remains low, with an AUC of 0.58 on the test data. This study highlights the potential of deep learning approaches in addressing the challenges of protein flexi-bility in virtual screening and docking simulations.