Collaborative Assessment of Molecular Geometries and Energies from the Open Force Field

Abstract

Force fields form the basis for classical molecular simulations, and their accuracy is crucial for the quality of, for instance, protein–ligand binding simulations in drug discovery. The huge diversity of small-molecule chemistry makes it a challenge to build and parameterize a suitable force field. The Open Force Field Initiative is a combined industry and academic consortium developing a state-of-the-art small-molecule force field. In this report, industry members of the consortium worked together to objectively evaluate the performance of the force fields (referred to here as OpenFF) produced by the initiative on a combined public and proprietary dataset of 19,653 relevant molecules selected from their internal research and compound collections. This evaluation was important because it was completely blind; at most partners, none of the molecules or data were used in force field development or testing prior to this work. We compare the Open Force Field “Sage” version 2.0.0 and “Parsley” version 1.3.0 with GAFF-2.11-AM1BCC, OPLS4, and SMIRNOFF99Frosst. We analyzed force-field-optimized geometries and conformer energies compared to reference quantum mechanical data. We show that OPLS4 performs best, and the latest Open Force Field release shows a clear improvement compared to its predecessors. The performance of established force fields such as GAFF-2.11 was generally worse. While OpenFF researchers were involved in building the benchmarking infrastructure used in this work, benchmarking was done entirely in-house within industrial organizations and the resulting assessment is reported here. This work assesses the force field performance using separate benchmarking steps, external datasets, and involving external research groups. This effort may also be unique in terms of the number of different industrial partners involved, with 10 different companies participating in the benchmark efforts.

Perspectives on Ligand/Protein Binding Kinetics Simulations: Force Fields, Machine Learning, Sampling, and User-Friendliness

“…Zhu showed similar results in its benchmark calculations of ligands FFs when compared to experimental osmotic coefficient data . Amore et al demonstrated that different FFs provide varying conformer energies and geometries during the optimization of multiple molecules and fragments . We documented that the parametrization of benzamidine using default GAFF parameters led to unsatisfactory outcomes, altering the ligand binding conformation and the free-energy landscape of the benzamidine/trypsin binding process .…”

“…An accurate estimate of kinetic rates requires a proper identification of TSs . However, subpar parameters might overly stabilize specific ligand (and protein) conformations, thus altering the identification and energetic evaluation of metastable and transient states. , It has been shown that inadequacies in the van der Waals and electrostatic terms of FFs may influence partition functions, , and also impact the osmotic coefficients of some chemical entities . Using the paradigmatic system benzamidine/trypsin, we have recently shown that proper parametrization of the ligand torsional potentials is essential to reconstruct the correct ligand binding conformation and free-energy surface .…”

Development and Benchmarking of Open Force Field 2.0.0: The Sage Small Molecule Force Field

“…Specifically, we use a set of QM optimized conformer geometries and energies to assess how well we can reproduce conformer energetics and geometries with MM parameters. A much larger data set of QM-optimized gas-phase geometries, named “OpenFF Industry Benchmark Season 1 v1.1” on QCArchive, is now used for benchmarking MM optimized geometries . We built this data set in collaboration with our industry partners to benchmark force fields more generally, and it consists of nearly 9847 unique molecules, and a total of 76713 conformers.…”

A transferable double exponential potential for condensed phase simulations of small molecules

“…For example, the Sage library contains 167 rotatable torsion parameters, compared to the 146 K of OPLS3e. 33,35 While this minimal parameter set does come with reduced accuracy compared to highly parameterised, proprietary force fields, 36 users can make use of OpenFF-BespokeFit 33 to fit bespoke torsion parameters to molecules of interest and improve agreement with QM data. The OpenFF software stack additionally includes robust tools, such as ForceBalance 19 and Evaluator, 20 which we have taken advantage of in the current study to automatically curate and train against thousands of experimental and quantum mechanical reference data points.…”