OpenFF 2.3.0rc2 Relative Binding Free Energy Benchmarks

Overview

Relative binding free energy benchmarks of the openff-2.3.0rc2 release using the JACS set from (https://github.com/OpenFreeEnergy/IndustryBenchmarks2024).

Contents

Benchmark variants

The following benchmark variants were run:  * `openff-2.3.0rc2_elf10`: simulations using OpenFF 2.3.0rc2 and AM1BCCELF10 charges using the OpenEye Toolkit.  * `openff-2.3.0rc2_nagl`: simulations using OpenFF 2.3.0rc2 and AshGC v0.1rc3 charges

For additional validation, an extra 3 replica were run for both variants across all sets. Results have been collated in separatefiles which are labelled with the tag `6replica` in their name.

RBFE data

The `data` folder contains all the ddG data for each transformation in each set.

Per system data files

These are stored as TSV files within subdirectories for each system in the JACS set.Reference CSV data (labelled as `{system}_ref.csv`), taken from the Ross et al. 2023 dataset (https://github.com/schrodinger/public_binding_free_energy_benchmark)can also be found.

Aggregate dG and ddG files

CSV files with all the dG and ddG values for all datasets can be found in `data`.

Simulation details

* All simulations were carried out using the RelativeHybridTopology protocol in OpenFE v1.4.* Mappings were created using Kartograf, these are the same as those used in the OpenFE industry benchmarks.* Transformation edges were created using Lomap, these are the same as those used in the OpenFE industry benchmarks.* Simulations used openfe default settings, except the following:  * Dodecahedron solvent box with 1 nm padding in complex and 1.5 nm padding in solvent  * 2.5 ps HREX exchange rate  * 0.9 nm short range interaction cutoff  * 3 replica were originally run, and then a further 3 were added to the `6replica` files.