A Relative Binding Free Energy Framework for Structurally Dissimilar Molecules

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Relative binding free energy (RBFE) calculations are vital for drug discovery but limited to molecules with similar structures.
• Traditional RBFE methods struggle with diverse small molecules lacking a common core or binding mode.
• Absolute binding free energy (ABFE) methods can handle diverse molecules but are computationally expensive and converge slowly.

Purpose of the Study:

• To introduce a computationally efficient framework, core-hopping binding free energy (CBFE) calculations, for determining relative binding free energies.
• To enable accurate predictions for small molecules with different cores and binding modes, overcoming limitations of traditional RBFE.
• To provide a practical tool for challenging drug discovery campaigns involving diverse chemical structures.

Main Methods:

• Developed a novel CBFE framework leveraging Alchemical Enhanced Sampling (ACES) with optimized transformation pathways and flexible λ-spacing.
• Incorporated λ-dependent Boresch restraints to enhance sampling and accuracy.
• Implemented and benchmarked CBFE across 4 protein systems with 56 small molecules using GPU-accelerated AMBER (pmemd.cuda).

Main Results:

• CBFE results showed consistency with RBFE for congeneric ligands.
• CBFE offered significant improvements in computational cost and precision compared to ABFE for diverse small molecules.
• The method demonstrated accuracy and efficiency in determining relative binding free energies across different molecular scaffolds and binding modes.

Conclusions:

• CBFE calculations provide an accurate and computationally efficient method for determining relative binding free energies between small molecules with diverse cores.
• This approach expands the applicability of binding free energy calculations to a wider range of molecules in drug discovery.
• The CBFE framework is fully implemented in AMBER, making it readily available for researchers.