Automated, Accurate, and Scalable Relative Protein-Ligand Binding Free-Energy Calculations Using Lambda Dynamics

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Predicting protein-ligand binding affinity is crucial for lead optimization in drug discovery.
• Traditional methods like free-energy perturbation (FEP) are computationally expensive, limiting their use.
• Lambda dynamics offers a more efficient alternative based on statistical mechanics.

Purpose of the Study:

• To develop and validate a computational workflow for multisite lambda dynamics (MSLD) calculations.
• To enable efficient screening of combinatorial libraries for lead optimization.
• To assess the accuracy and efficiency of MSLD compared to FEP.

Main Methods:

• Development of a workflow using CHARMM in BIOVIA Discovery Studio and Pipeline Pilot on GPUs.
• Implementation of a protocol for iterative tailoring of biasing potentials to flatten free-energy landscapes.
• Validation using a diverse dataset of congeneric ligands across seven proteins.
• Testing scalability for screening over 100 ligands in a single system.

Main Results:

• The MSLD workflow accurately predicts relative binding affinities for small and large combinatorial libraries.
• Average unsigned errors are below 1 kcal/mol with cumulative sampling times of 150 ns or less.
• The method demonstrates over an order of magnitude greater efficiency than contemporary FEP applications.
• Accurate and reliable performance was observed in validating ligand subsets and large-scale screening.

Conclusions:

• The developed MSLD workflow provides an efficient and accurate approach for screening combinatorial libraries.
• This method facilitates exploration of chemical space around lead compounds in drug optimization.
• MSLD offers a significant computational advantage over FEP for binding affinity predictions.