Improving virtual screening results with MM/GBSA and MM/PBSA rescoring

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Virtual screening (VS) is crucial for computer-aided drug design, accelerating the identification of potential therapeutic ligands.
• Current VS methods often suffer from low accuracy in binding affinity prediction due to limitations of scoring functions.
• More rigorous binding affinity calculations are time-consuming, necessitating efficient rescoring strategies.

Purpose of the Study:

• To improve the accuracy of virtual screening by rescoring docking results.
• To evaluate the effectiveness of MM/PB(GB)SA methods for binding affinity rescoring.
• To provide a user-friendly program for rescoring VS results.

Main Methods:

• Performed virtual screening on three benchmark datasets.
• Employed energy minimization with MM/PB(GB)SA (molecular mechanics energies combined with the Poisson-Boltzmann or generalized Born and surface area) for rescoring.
• Utilized metrics such as AUC, EF, and BEDROC for performance evaluation.

Main Results:

• Rescoring with MM/PB(GB)SA methods led to essential improvements in binding energy prediction.
• Demonstrated significant enhancements in virtual screening accuracy.
• Validated the effectiveness of the proposed rescoring approach across benchmark sets.

Conclusions:

• MM/PB(GB)SA rescoring substantially improves the accuracy of virtual screening.
• The developed program, based on AmberTools, offers an efficient tool for enhancing VS results.
• This approach aids in reducing false positives and expediting the drug development pipeline.