Development and large-scale benchmarks of a protein-ligand absolute binding free energy toolkit

Abstract

Absolute binding free energy (ABFE) calculations offer a theoretically rigorous approach for predicting protein--ligand binding affinities without the scaffold constraints of relative binding free energy (RBFE) perturbations. However, broad adoption of ABFE in high-throughput hit discovery campaigns has been hindered by high computational costs and a lack of large-scale validation. Here, we present Felis, an open-source, automated, and scalable toolkit designed for high-throughput ABFE calculations. Paired with ByteFF, a previously developed data-driven molecular mechanics force field for drug-like molecules, Felis achieves ranking performance comparable to state-of-the-art RBFE methods on a diverse dataset comprising 43 protein targets and 859 ligands. Furthermore, we demonstrate robust convergence and ranking performance of Felis on a more challenging KRAS(G12D) dataset, where some ligands and the cofactor are highly charged. Crucially, all Felis predictions in this study were generated in a strict zero-shot manner, eschewing custom force-field modifications and alchemical schedule fine-tuning. This demonstrates the viability of Felis as an effective, ready-to-use tool for computational structure-based drug design.

Figures (8)

• Figure 1: Overview of the ABFE thermodynamic cycle. The absolute binding free energy is given by $\Delta G_\mathrm{bind} = \Delta G_\mathrm{solvent} + \Delta G_\mathrm{restraint} + \Delta G_\mathrm{protein}$, where $\Delta G_\mathrm{solvent}$ is the free energy of decoupling the ligand from solvent, $\Delta G_\mathrm{restraint}$ is the analytical correction for the Boresch-style restraints, and $\Delta G_\mathrm{protein}$ is the free energy of coupling the ligand to the protein environment while releasing the Boresch-style restraints.
• Figure 2: Ranking performance of Felis-ABFE and FEP+ RBFE across the full benchmark set of 43 target proteins. Felis-ABFE simulations (3ns or 5ns with three independent replicas) achieved ranking performance comparable to FEP+ RBFE with 20ns sampling.
• Figure 3: Comparison of AM1-BCC vs ABCG2 charges across the full benchmark dataset of 43 proteins, using Felis-ABFE (5ns, three independent replicas). ByteFF25-AM1-BCC and ByteFF25-ABCG2 are trained on the same quantum chemistry data for different charge schemes, respectively.
• Figure 4: Case study on KRAS(G12D) MRTX1133 series. A. Convergence of $\Delta G$ predictions for the KRAS(G12D)-MRTX1133 complex as a function of MD simulation length. Box plots display the distribution of $\Delta G$ from 10 independent trials at each duration, with individual points overlaid. B. Felis-ABFE ranking metrics for the ligand series using 3ns $\times$ 3, 5ns $\times$ 3, and 10ns $\times$ 3 simulation protocols.
• Figure 5: Comparison of Felis-ABFE with ByteFF25-AM1-BCC and FEP+ RBFE rossMaximalCurrentFEPBenchmark2023 results on the target opls_stress/hc_bace_2.