The Quasi-Bound State as a Predictor of Relative Binding Free Energy

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Relative binding free energy (ΔΔGbind) predictions are crucial for evaluating compound potency in drug design.
• Current alchemical methods are computationally expensive and limited in scope.
• Faster methods are needed for large-scale virtual screening.

Purpose of the Study:

• To introduce Dynamic Undocking (DUck) as a faster alternative for predicting binding free energy.
• To demonstrate DUck's ability to detect activity cliffs and predict relative binding free energy (ΔΔGbind).
• To compare DUck's performance against traditional alchemical methods.

Main Methods:

• Utilized Dynamic Undocking (DUck) to calculate the free energy to reach a quasi-bound state (ΔGQB).
• Applied DUck to congeneric series of HSP90α, CDK2, and BACE1 inhibitors.
• Analyzed systems following a one-step dissociation model.

Main Results:

• DUck accurately predicts ΔGQB, serving as a robust measure of structural robustness in protein-ligand complexes.
• ΔGQB effectively identifies outliers in the structure-activity relationship, known as activity cliffs.
• In specific cases, ΔGQB predictions were comparable to computationally demanding alchemical transformation methods.
• ΔGQB can inform predictions of relative binding kinetics and ΔΔGbind under certain conditions.

Conclusions:

• Dynamic Undocking (DUck) offers a computationally efficient high-throughput method for predicting relative binding free energy (ΔΔGbind).
• DUck excels at identifying activity cliffs, a significant challenge for knowledge-based drug discovery approaches.
• This method provides a cost-effective alternative to traditional alchemical transformations within its applicability domain, accelerating drug discovery pipelines.