Protein-Ligand Electrostatic Binding Free Energies from Explicit and Implicit Solvation.

Area of Science:

• Computational chemistry
• Molecular modeling
• Biophysics

Background:

• Accurate and efficient solvent models are crucial for atomistic calculations, particularly for predicting protein-ligand binding affinities.
• Generalized Born (GB) implicit solvent models offer a computationally faster alternative to explicit solvent models.

Purpose of the Study:

• To evaluate the accuracy of the GBNSR6 implicit solvent model for electrostatic solvation and binding free energies in small protein-ligand complexes.
• To compare GBNSR6 performance against three common explicit solvent models: TIP3P, TIP4PEw, and OPC.
• To assess the impact of atomic radii adjustments on GBNSR6 accuracy.

Main Methods:

• Evaluation of GBNSR6 for electrostatic solvation free energies (ΔG(pol)) and binding free energies (ΔΔG(pol)).
• Comparison with explicit solvent models (TIP3P, TIP4PEw, OPC).
• Analysis of deviations (RMSD) and systematic errors in binding affinity predictions.

Main Results:

• GBNSR6 binding affinity deviations from TIP3P are comparable to inter-model deviations among explicit solvents.
• A simple uniform scaling of atomic radii significantly improves GBNSR6 accuracy, reducing deviations from explicit models.
• Explicit solvent models show large, concerning discrepancies in binding energy estimates (up to ~9 kcal/mol), highlighting model sensitivity.

Conclusions:

• GBNSR6, with optimized radii, provides accurate binding free energy estimates comparable to explicit solvent models.
• The significant discrepancies between explicit solvent models underscore the need for reliable implicit solvation models.
• Implicit solvent models like GBNSR6 present a faster and potentially more consistent approach for binding energetics calculations.