Binding Free Energy Estimator 3

Binding Free Energy Estimator 3 (BFEE3) is here! BFEE3 includes many upgrades:
(1) LDDM, a high-throughput alchemical route for absolute binding free-energy calculations;
(2) WTM-λABF, an efficient algorithm for alchemical transformations;
(3) a streamlined geometrical route for protein-protein binding free-energy calculations;
(4) quick setup options for common calculations;
(5) an AI assistant for generating input files and answering questions.

BFEE is a Python-based software package that automates absolute binding free-energy calculations through either the alchemical or geometric route using molecular dynamics simulations.

Theoretical background

The degrees of freedom of the protein-ligand or host-guest system are described by a series of geometric variables, or collective variables, as first described by the Karplus group. In BFEE, generalized geometric variables based on best-fit rotation are used, making the method, in principle, applicable to any protein-ligand complex. See this paper for an introduction to these variables.

In the geometric route, the degrees of freedom are investigated one by one through one-dimensional free-energy calculations. In BFEE, WTM-eABF is used, while other importance-sampling algorithms, such as plain eABF, are also supported.

The alchemical route is a variant of the double decoupling method (DDM). It uses a thermodynamic cycle in which the ligand and the geometric restraints are decoupled independently to ensure convergence of the simulations.

More recently, high-efficiency methods such as Lucid DDM and WTM-λABF have also been implemented in BFEE3. These methods are generally recommended as the first choice for absolute binding free-energy calculations.
这里是标准结合自由能计算方法的中文介绍。

Features

Generates all input files for absolute binding free-energy calculations;
Supports protein-protein and protein-ligand complexes;
Performs post-processing automatically;
Supports NAMD for both the alchemical and geometric routes, and GROMACS for the geometric route, as molecular dynamics engines;
Supports many file formats for the input complex structure, including PSF/PDB/PRM, PRM7/RST7, and TOP/PDB;
Supports both rigid ligands and protein-protein complexes, which exclude the RMSD CV, and flexible ligands and protein-protein complexes, which include the RMSD CV, through the streamlined geometrical route;
...

Requirements

Python 3.6+, PySide6, numpy, scipy, matplotlib, parmed, and MDAnalysis.
NAMD 3.0 or later / GROMACS 2024 or later.
Note: Since both NAMD and GROMACS have incorporated Colvars into their latest binaries, each release of BFEE3 corresponds to a specific version of NAMD/GROMACS. Please always use the corresponding or later versions of the MD engines for free-energy calculations!

Installation

We suggest installing BFEE through conda. It is safer to install it in a new conda environment.

conda create --name bfee   (optional)
conda activate bfee        (optional)
conda install -c conda-forge BFEE2

IMPORTANT: Please force numpy<2.3 when installing BFEE, because a recent update of numpy has broken ParmEd (https://github.com/ParmEd/ParmEd/issues/1406).

Usage

Simply run BFEE2Gui.py in a terminal or PowerShell. On Microsoft Windows, you may need to use the absolute path.
A step-by-step tutorial is provided here.
A tutorial on the new streamlined geometrical route is provided in the Supporting Information of this paper.

Test files

You can download the supplementary data here to test BFEE3.

Citations

When possible, please consider mentioning Fu et al. Nat. Protoc. 2022, 17, 1114–1141 when BFEE is used in your project.

Additional references:
WTM-λABF: Zhou et al. J. Phys. Chem. Lett. 2025, 16, 4419–4427 and Zhou et al. Acc. Chem. Res. 2026, 59, 90–102
The "LDDM" route Bian et al. Nat. Comput. Sci., 2025, 5, 621–626
BFEE2: Fu et al. J. Chem. Inf. Model. 2021, 61, 2116–2123
BFEE2 for protein-protein binding free-energy calculations: Fu et al. J. Chem. Inf. Model. 2023, 63, 2512–2519
Alchemical and geometric routes: Gumbart et al. J. Chem. Theory Comput. 2013, 9, 794–802
Lucid DDM method: Bian et al. Nat. Comput. Sci. 2025, 5, 621–626
WTM-λABF: Zhou et al. Acc. Chem. Res. 2026, 59, 90–102
WTM-eABF: Fu et al. Acc. Chem. Res. 2019, 52, 3254–3264 and Fu et al. J. Phys. Chem. Lett. 2018, 9, 4738–4745
NAMD3: Phillips et al. J. Chem. Phys. 2020, 153, 044130
Gromacs: Abraham et al. SoftwareX, 2015, 1, 19-25
Collective variables: Fu et al. J. Chem. Theory Comput. 2017, 13, 5173–5178
Colvars module: Fiorin et al. Mol. Phys. 2013, 111, 3345–3362 and Fiorin et al. J. Phys. Chem. B 2024, 128, 11108–11123
The "mother" of all restraint-based binding free-energy calculations: Hermans et al. Isr. J. Chem. 1986, 27, 225–227

Disclaimer

The methods, protocols, and computational workflows implemented in BFEE3 are provided as practical recommendations for conducting the relevant calculations. They are not intended to guarantee the correctness of any resulting data or conclusions, nor should they be regarded as the only acceptable procedures for performing such calculations.

Data Security

API keys are stored locally only. Molecular structure files and force field files are also accessed and processed locally. If a third-party LLM service is used, user conversation content will be sent to the provider. We therefore recommend providers that supports Zero Data Retention (ZDR).

Contact us

This software is distributed under the GPLv3 license. For more information about BFEE, please contact Haohao Fu (fhh2626@nankai.edu.cn) and Haochuan Chen (yjcoshc@mail.nankai.edu.cn).