MM/GBSA is widely used in end-point binding free energy prediction in structure-based drug design. Here, combining with machine-learning optimization, we present a refinement version of MM/GBSA named Variable Atomic Dielectric MM/GBSA (VAD-MM/GBSA) that exhibits improved accuracy for predicting binding affinities of diverse protein-ligand systems and is promising to be used in post-processing of structure based virtual screening.
Simulated annealing-based optimization is used to search the optimal atomic dielectric constant distributions, which are correlative with 87 descriptors of the ligand and the pocket of protein by XGBoost regression.
Click here for more information.
If you have any questions, please contact us.