[Paper Review] Binding Free Energies without Alchemy

Michael Brocidiacono, Brandon Novy+3 more|arXiv (Cornell University)|Mar 12, 2026

Protein Structure and Dynamics0 citations

TL;DR

DBFE is an end-state ABFE method in implicit solvent that requires no alchemical intermediates, enabling a single complex simulation per ligand and offering competitive accuracy with substantial speedups for virtual screening.

ABSTRACT

Absolute Binding Free Energy (ABFE) methods are among the most accurate computational techniques for predicting protein-ligand binding affinities, but their utility is limited by the need for many simulations of alchemically modified intermediate states. We propose Direct Binding Free Energy (DBFE), an end-state ABFE method in implicit solvent that requires no alchemical intermediates. DBFE outperforms OBC2 double decoupling on a host-guest benchmark and performs comparably to OBC2 MM/GBSA on a protein-ligand benchmark. Since receptor and ligand simulations can be precomputed and amortized across compounds, DBFE requires only one complex simulation per ligand compared to the many lambda windows needed for double decoupling, making it a promising candidate for virtual screening workflows. We publicly release the code for this method at https://github.com/molecularmodelinglab/dbfe.

Motivation & Objective

• Motivate faster, more scalable absolute binding free energy calculations for virtual screening.
• Develop an end-state ABFE approach that eliminates alchemical intermediates.
• Show DBFE’s performance relative to DD and MM/GBSA on host-guest and protein-ligand benchmarks.
• Assess the impact of implicit solvent models and conformational entropy corrections on accuracy.

Proposed method

• Compute three end-state simulations: receptor-only, ligand-only, and receptor-ligand complex.
• Introduce a restraint on relative translation/rotation (
• Define U_{} as a combination of normal and Bingham distributions to model translational-rotational restraints.
• Use a KD-tree-based filtering to identify clash-free samples between decoupled and bound states.
• Compute  via BAR between the coupled and clash-free subsets.
• Assemble final 3 by summing 3 for each leg of a thermodynamic cycle and applying a standard-state correction.

Figure 1: The full thermodynamic cycle used by DBFE.

Experimental results

Research questions

• RQ1 How does an end-state implicit-solvent ABFE method compare to alchemical DD and MM/GBSA on host-guest systems?
• RQ2 Can end-state simulations plus efficient sampling deliver competitive binding free energies for protein-ligand complexes?
• RQ3 What is the impact of conformational entropy corrections in implicit-solvent ABFE methods on accuracy?
• RQ4 Is per-ligand computational cost reduced sufficiently to be practical for virtual screening workflows?],
• RQ5 key_findings）：[

Key findings

• On a host-guest benchmark (50 complexes), DBFE achieves RMSE 4.1 kcal/mol, Pearson r 0.58, and Spearman 0.56, outperforming OBC2 DD (r=0.48).
• On a protein-ligand benchmark (54 complexes), DBFE attains RMSE 5.3 kcal/mol, Pearson r 0.65, and Spearman 0.64, slightly worse than OBC2 MM/GBSA (r=0.71, 0.70).
• TIP3P DD outperforms all OBC2 methods with r=0.88, indicating explicit solvent benefits in DD benchmarks.
• DBFE provides a ~26x per-ligand cost reduction in the complex-iteration step compared to OBC2 DD, due to amortized receptor and ligand simulations.
• DBFE’s improved host-guest correlation suggests conformational entropy corrections are important for small, well-defined binding sites.
• DBFE’s performance relative to MM/GBSA implies implicit solvent limitations are a primary bottleneck, with entropy corrections adding noise in some protein-ligand cases.

Figure 2: Comparison of all methods versus experimental $\Delta G$ on the host-guest benchmark. Each subplot shows predicted versus experimental binding free energies, with bootstrap metrics (RMSE, Pearson $r$ , Spearman $\rho$ ) and 95% confidence intervals.