Evaluating the performance of MM/PBSA for binding affinity prediction using class A GPCR crystal structures.

Area of Science:

• Computational Chemistry
• Structural Biology
• Pharmacology

Background:

• The growing number of G protein-coupled receptor (GPCR) crystal structures enables evaluation of structure-based drug design (SBDD) techniques.
• Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) methods offer a balance between speed and accuracy for predicting binding affinities.

Purpose of the Study:

• To systematically assess the predictive performance of MM/PBSA for binding free energies across diverse Class A GPCR targets.
• To compare MM/PBSA performance against docking scores and investigate factors influencing prediction accuracy.

Main Methods:

• Utilized twenty Class A GPCR crystal structures and 934 known ligands.
• Performed MM/PBSA calculations, comparing predicted binding free energies with experimental data.
• Investigated the impact of using single minimized structures versus molecular dynamics simulations and ligand structural similarity.

Main Results:

• Correlations between predicted and experimental binding free energies varied significantly across targets (r = -0.334 to 0.781), with a poor overall average (r = 0.183).
• MM/PBSA outperformed docking scores for eight targets and performed worse for four.
• Single-structure MM/PBSA calculations yielded comparable results to molecular dynamics simulations, offering potential computational efficiency.

Conclusions:

• MM/PBSA performance in GPCR drug design is highly system-specific.
• Prospective application of MM/PBSA requires prior validation for each specific GPCR target and ligand set.
• Restricting calculations to structurally similar ligands may enhance prediction accuracy in some cases.