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

We introduce the Open Force Field (OpenFF) 2.0.0 small molecule force field for drug-like molecules, code-named Sage, which builds upon our previous iteration, Parsley. OpenFF force fields are based on direct chemical perception, which generalizes easily to highly diverse sets of chemistries based on substructure queries. Like the previous OpenFF iterations, the Sage generation of OpenFF force fields was validated in protein-ligand simulations to be compatible with AMBER biopolymer force fields. In this work, we detail the methodology used to develop this force field, as well as the innovations and improvements introduced since the release of Parsley 1.0.0. One particularly significant feature of Sage is a set of improved Lennard-Jones (LJ) parameters retrained against condensed phase mixture data, the first refit of LJ parameters in the OpenFF small molecule force field line. Sage also includes valence parameters refit to a larger database of quantum chemical calculations than previous versions, as well as improvements in how this fitting is performed. Force field benchmarks show improvements in general metrics of performance against quantum chemistry reference data such as root-mean-square deviations (RMSD) of optimized conformer geometries, torsion fingerprint deviations (TFD), and improved relative conformer energetics (ΔΔE). We present a variety of benchmarks for these metrics against our previous force fields as well as in some cases other small molecule force fields. Sage also demonstrates improved performance in estimating physical properties, including comparison against experimental data from various thermodynamic databases for small molecule properties such as ΔHmix, ρ(x), ΔGsolv, and ΔGtrans. Additionally, we benchmarked against protein-ligand binding free energies (ΔGbind), where Sage yields results statistically similar to previous force fields. All the data is made publicly available along with complete details on how to reproduce the training results at https://github.com/openforcefield/openff-sage.

Best practices for constructing, preparing, and evaluating protein-ligand binding affinity benchmarks [Article v0.1].

Free energy calculations are rapidly becoming indispensable in structure-enabled drug discovery programs. As new methods, force fields, and implementations are developed, assessing their expected accuracy on real-world systems (benchmarking) becomes critical to provide users with an assessment of the accuracy expected when these methods are applied within their domain of applicability, and developers with a way to assess the expected impact of new methodologies. These assessments require construction of a benchmark-a set of well-prepared, high quality systems with corresponding experimental measurements designed to ensure the resulting calculations provide a realistic assessment of expected performance when these methods are deployed within their domains of applicability. To date, the community has not yet adopted a common standardized benchmark, and existing benchmark reports suffer from a myriad of issues, including poor data quality, limited statistical power, and statistically deficient analyses, all of which can conspire to produce benchmarks that are poorly predictive of real-world performance. Here, we address these issues by presenting guidelines for (1) curating experimental data to develop meaningful benchmark sets, (2) preparing benchmark inputs according to best practices to facilitate widespread adoption, and (3) analysis of the resulting predictions to enable statistically meaningful comparisons among methods and force fields. We highlight challenges and open questions that remain to be solved in these areas, as well as recommendations for the collection of new datasets that might optimally serve to measure progress as methods become systematically more reliable. Finally, we provide a curated, versioned, open, standardized benchmark set adherent to these standards (PLBenchmarks) and an open source toolkit for implementing standardized best practices assessments (arsenic) for the community to use as a standardized assessment tool. While our main focus is free energy methods based on molecular simulations, these guidelines should prove useful for assessment of the rapidly growing field of machine learning methods for affinity prediction as well.

Development and Benchmarking of Open Force Field v1.0.0-the Parsley Small-Molecule Force Field.

We present a methodology for defining and optimizing a general force field for classical molecular simulations, and we describe its use to derive the Open Force Field 1.0.0 small-molecule force field, codenamed Parsley. Rather than using traditional atom typing, our approach is built on the SMIRKS-native Open Force Field (SMIRNOFF) parameter assignment formalism, which handles increases in the diversity and specificity of the force field definition without needlessly increasing the complexity of the specification. Parameters are optimized with the ForceBalance tool, based on reference quantum chemical data that include torsion potential energy profiles, optimized gas-phase structures, and vibrational frequencies. These quantum reference data are computed and are maintained with QCArchive, an open-source and freely available distributed computing and database software ecosystem. In this initial application of the method, we present essentially a full optimization of all valence parameters and report tests of the resulting force field against compounds and data types outside the training set. These tests show improvements in optimized geometries and conformational energetics and demonstrate that Parsley's accuracy for liquid properties is similar to that of other general force fields, as is accuracy on binding free energies. We find that this initial Parsley force field affords accuracy similar to that of other general force fields when used to calculate relative binding free energies spanning 199 protein-ligand systems. Additionally, the resulting infrastructure allows us to rapidly optimize an entirely new force field with minimal human intervention.

Non-bonded force field model with advanced restrained electrostatic potential charges (RESP2).

The restrained electrostatic potential (RESP) approach is a highly regarded and widely used method of assigning partial charges to molecules for simulations. RESP uses a quantum-mechanical method that yields fortuitous overpolarization and thereby accounts only approximately for self-polarization of molecules in the condensed phase. Here we present RESP2, a next generation of this approach, where the polarity of the charges is tuned by a parameter, δ, which scales the contributions from gas- and aqueous-phase calculations. When the complete non-bonded force field model, including Lennard-Jones parameters, is optimized to liquid properties, improved accuracy is achieved, even with this reduced set of five Lennard-Jones types. We argue that RESP2 with δ≈0.6 (60% aqueous, 40% gas-phase charges) is an accurate and robust method of generating partial charges, and that a small set of Lennard-Jones types is good starting point for a systematic re-optimization of this important non-bonded term.

Benchmark assessment of molecular geometries and energies from small molecule force fields.

Background: Force fields are used in a wide variety of contexts for classical molecular simulation, including studies on protein-ligand binding, membrane permeation, and thermophysical property prediction. The quality of these studies relies on the quality of the force fields used to represent the systems. Methods: Focusing on small molecules of fewer than 50 heavy atoms, our aim in this work is to compare nine force fields: GAFF, GAFF2, MMFF94, MMFF94S, OPLS3e, SMIRNOFF99Frosst, and the Open Force Field Parsley, versions 1.0, 1.1, and 1.2. On a dataset comprising 22,675 molecular structures of 3,271 molecules, we analyzed force field-optimized geometries and conformer energies compared to reference quantum mechanical (QM) data. Results: We show that while OPLS3e performs best, the latest Open Force Field Parsley release is approaching a comparable level of accuracy in reproducing QM geometries and energetics for this set of molecules. Meanwhile, the performance of established force fields such as MMFF94S and GAFF2 is generally somewhat worse. We also find that the series of recent Open Force Field versions provide significant increases in accuracy. Conclusions: This study provides an extensive test of the performance of different molecular mechanics force fields on a diverse molecule set, and highlights two (OPLS3e and OpenFF 1.2) that perform better than the others tested on the present comparison. Our molecule set and results are available for other researchers to use in testing.

Biomolecular Solvation Structure Revealed by Molecular Dynamics Simulations.

To compare ordered water positions from experiment with those from molecular dynamics (MD) simulations, a number of MD models of water structure in crystalline endoglucanase were calculated. The starting MD model was derived from a joint X-ray and neutron diffraction crystal structure, enabling the use of experimentally assigned protonation states. Simulations were performed in the crystalline state, using a periodic 2 × 2 × 2 supercell with explicit solvent. Water X-ray and neutron scattering density maps were computed from MD trajectories using standard macromolecular crystallography methods. In one set of simulations, harmonic restraints were applied to bias the protein structure toward the crystal structure. For these simulations, the recall of crystallographic waters using strong peaks in the MD water electron density was very good, and there also was substantial visual agreement between the boomerang-like wings of the neutron scattering density and the crystalline water hydrogen positions. An unrestrained simulation also was performed. For this simulation, the recall of crystallographic waters was much lower. For both restrained and unrestrained simulations, the strongest water density peaks were associated with crystallographic waters. The results demonstrate that it is now possible to recover crystallographic water structure using restrained MD simulations but that it is not yet reasonable to expect unrestrained MD simulations to do the same. Further development and generalization of MD water models for force-field development, macromolecular crystallography, and medicinal chemistry applications is now warranted. In particular, the combination of room-temperature crystallography, neutron diffraction, and crystalline MD simulations promises to substantially advance modeling of biomolecular solvation.

Assessing the Conformational Equilibrium of Carboxylic Acid via Quantum Mechanical and Molecular Dynamics Studies on Acetic Acid.

Accurate hydrogen placement in molecular modeling is crucial for studying the interactions and dynamics of biomolecular systems. The carboxyl functional group is a prototypical example of a functional group that requires protonation during structure preparation. To our knowledge, when in their neutral form, carboxylic acids are typically protonated in the syn conformation by default in classical molecular modeling packages, with no consideration of alternative conformations, though we are not aware of any careful examination of this topic. Here, we investigate the general belief that carboxylic acids should always be protonated in the syn conformation. We calculate and compare the relative energetic stabilities of syn and anti acetic acid using ab initio quantum mechanical calculations and atomistic molecular dynamics simulations. We focus on the carboxyl torsional potential and configurations of microhydrated acetic acid from molecular dynamics simulations, probing the effects of solvent, force field (GAFF vs GAFF2), and partial charge assignment of acetic acid. We show that while the syn conformation is the preferred state, the anti state may in some cases also be present under normal NPT conditions in solution.

Toward Learned Chemical Perception of Force Field Typing Rules.

Molecular mechanics force fields define how the energy and forces in a molecular system are computed from its atomic positions, thus enabling the study of such systems through computational methods like molecular dynamics and Monte Carlo simulations. Despite progress toward automated force field parametrization, considerable human expertise is required to develop or extend force fields. In particular, human input has long been required to define atom types, which encode chemically unique environments that determine which parameters will be assigned. However, relying on humans to establish atom types is suboptimal. Human-created atom types are often developed without statistical justification, leading to over- or under-fitting of data. Human-created types are also difficult to extend in a systematic and consistent manner when new chemistries must be modeled or new data becomes available. Finally, human effort is not scalable when force fields must be generated for new (bio)polymers, compound classes, or materials. To remedy these deficiencies, our long-term goal is to replace human specification of atom types with an automated approach, based on rigorous statistics and driven by experimental and/or quantum chemical reference data. In this work, we describe novel methods that automate the discovery of appropriate chemical perception: SMARTY allows for the creation of atom types, while SMIRKY goes further by automating the creation of fragment (nonbonded, bonds, angles, and torsions) types. These approaches enable the creation of move sets in atom or fragment type space, which are used within a Monte Carlo optimization approach. We demonstrate the power of these new methods by automating the rediscovery of human defined atom types (SMARTY) or fragment types (SMIRKY) in existing small molecule force fields. We assess these approaches using several molecular data sets, including one which covers a diverse subset of the DrugBank database.

Escaping Atom Types in Force Fields Using Direct Chemical Perception.

Traditional approaches to specifying a molecular mechanics force field encode all the information needed to assign force field parameters to a given molecule into a discrete set of atom types. This is equivalent to a representation consisting of a molecular graph comprising a set of vertices, which represent atoms labeled by atom type, and unlabeled edges, which represent chemical bonds. Bond stretch, angle bend, and dihedral parameters are then assigned by looking up bonded pairs, triplets, and quartets of atom types in parameter tables to assign valence terms and using the atom types themselves to assign nonbonded parameters. This approach, which we call indirect chemical perception because it operates on the intermediate graph of atom-typed nodes, creates a number of technical problems. For example, atom types must be sufficiently complex to encode all necessary information about the molecular environment, making it difficult to extend force fields encoded this way. Atom typing also results in a proliferation of redundant parameters applied to chemically equivalent classes of valence terms, needlessly increasing force field complexity. Here, we describe a new approach to assigning force field parameters via direct chemical perception. Rather than working through the intermediary of the atom-typed graph, direct chemical perception operates directly on the unmodified chemical graph of the molecule to assign parameters. In particular, parameters are assigned to each type of force field term (e.g., bond stretch, angle bend, torsion, and Lennard-Jones) based on standard chemical substructure queries implemented via the industry-standard SMARTS chemical perception language, using SMIRKS extensions that permit labeling of specific atoms within a chemical pattern. We use this to implement a new force field format, called the SMIRKS Native Open Force Field (SMIRNOFF) format. We demonstrate the power and generality of this approach using examples of specific molecules that pose problems for indirect chemical perception and construct and validate a minimalist yet very general force field, SMIRNOFF99Frosst. We find that a parameter definition file only ∼300 lines long provides coverage of all but <0.02% of a 5 million molecule drug-like test set. Despite its simplicity, the accuracy of SMIRNOFF99Frosst for small molecule hydration free energies and selected properties of pure organic liquids is similar to that of the General Amber Force Field, whose specification requires thousands of parameters. This force field provides a starting point for further optimization and refitting work to follow.

Toward Automated Benchmarking of Atomistic Force Fields: Neat Liquid Densities and Static Dielectric Constants from the ThermoML Data Archive.

Atomistic molecular simulations are a powerful way to make quantitative predictions, but the accuracy of these predictions depends entirely on the quality of the force field employed. Although experimental measurements of fundamental physical properties offer a straightforward approach for evaluating force field quality, the bulk of this information has been tied up in formats that are not machine-readable. Compiling benchmark data sets of physical properties from non-machine-readable sources requires substantial human effort and is prone to the accumulation of human errors, hindering the development of reproducible benchmarks of force-field accuracy. Here, we examine the feasibility of benchmarking atomistic force fields against the NIST ThermoML data archive of physicochemical measurements, which aggregates thousands of experimental measurements in a portable, machine-readable, self-annotating IUPAC-standard format. As a proof of concept, we present a detailed benchmark of the generalized Amber small-molecule force field (GAFF) using the AM1-BCC charge model against experimental measurements (specifically, bulk liquid densities and static dielectric constants at ambient pressure) automatically extracted from the archive and discuss the extent of data available for use in larger scale (or continuously performed) benchmarks. The results of even this limited initial benchmark highlight a general problem with fixed-charge force fields in the representation low-dielectric environments, such as those seen in binding cavities or biological membranes.

Efficient calculation of SAMPL4 hydration free energies using OMEGA, SZYBKI, QUACPAC, and Zap TK.

Several submissions for the SAMPL4 hydration free energy set were calculated using OpenEye tools, including many that were among the top performing submissions. All of our best submissions used AM1BCC charges and Poisson-Boltzmann solvation. Three submissions used a single conformer for calculating the hydration free energy and all performed very well with mean unsigned errors ranging from 0.94 to 1.08 kcal/mol. These calculations were very fast, only requiring 0.5-2.0 s per molecule. We observed that our two single-conformer methodologies have different types of failure cases and that these differences could be exploited for determining when the methods are likely to have substantial errors.

Shaping suvorexant: application of experimental and theoretical methods for driving synthetic designs.

Dual Orexin Receptor Antagonists (DORA) bind to both the Orexin 1 and 2 receptors. High resolution crystal structures of the Orexin 1 and 2 receptors, both class A GPCRs, were not available at the time of this study, and thus, ligand-based analyses were invoked and successfully applied to the design of DORAs. Computational analysis, ligand based superposition, unbound small-molecule X-ray crystal structures and NMR analysis were utilized to understand the conformational preferences of key DORAs and excellent agreement between these orthogonal approaches was seen in the majority of compounds examined. The predominantly face-to-face (F2F) interaction observed between the distal aromatic rings was the core 3D shape motif in our design principle and was used in the development of compounds. A notable exception, however, was seen between computation and experiment for suvorexant where the molecule exhibits an extended conformation in the unbound small-molecule X-ray structure. Even taking into account solvation effects explicitly in our calculations, we nevertheless find support that the F2F conformation is the bioactive conformation. Using a dominant states approximation for the partition function, we made a comprehensive assessment of the free energies required to adopt both an extended and a F2F conformation of a number of DORAs. Interestingly, we find that only a F2F conformation is consistent with the activities reported.

Using electronic polarization from the internal continuum (EPIC) for intermolecular interactions.

Recently, the vacuum-phase molecular polarizability tensor of various molecules has been accurately modeled (Truchon et al., J Chem Theory Comput 2008, 4, 1480) with an intramolecular continuum dielectric model. This preliminary study showed that electronic polarization can be accurately modeled when combined with appropriate dielectric constants and atomic radii. In this article, using the parameters developed to reproduce ab initio quantum mechanical (QM) molecular polarizability tensors, we extend the application of the "electronic polarization from internal continuu" (EPIC) approach to intermolecular interactions. We first derive a dielectric-adapted least-square-fit procedure similar to RESP, called DRESP, to generate atomic partial charges based on a fit to a QM abinitio electrostatic potential (ESP). We also outline a procedure to adapt any existing charge model to EPIC. The ability of this to reproduce local polarization, as opposed to uniform polarization, is also examined leading to an induced ESP relative root mean square deviation of 1%, relative to ab initio, when averaged over 37 molecules including aromatics and alkanes. The advantage of using a continuum model as opposed to an atom-centered polarizable potential is illustrated with a symmetrically perturbed atom and benzene. We apply EPIC to a cation-pi binding system formed by an atomic cation and benzene and show that the EPIC approach can accurately account for the induction energy. Finally, this article shows that the ab initio electrostatic component in the difficult case of the H-bonded 4-pyridone dimer, a highly polar and polarized interaction, is well reproduced without adjusting the vacuum-phase parameters.

Design and synthesis of dipeptidyl nitriles as potent, selective, and reversible inhibitors of cathepsin C.

A series of dipeptide nitriles with a thienyl alanine in P2 were identified as potent and selective cathepsin C inhibitors. Incorporation of a substituted cyclopropyl moiety in P1 effectively protects these derivatives against hydrolase activity in whole blood.

Predictions of hydration free energies from all-atom molecular dynamics simulations.

Here, we computed the aqueous solvation (hydration) free energies of 52 small drug-like molecules using an all-atom force field in explicit water. This differs from previous studies in that (1) this was a blind test (in an event called SAMPL sponsored by OpenEye Software) and (2) the test compounds were considerably more challenging than have been used in the past in typical solvation tests of all-atom models. Overall, we found good correlations with experimental values which were subsequently made available, but the variances are large compared to those in previous tests. We tested several different charge models and found that several standard charge models performed relatively well. We found that hypervalent sulfur and phosphorus compounds are not well handled using current force field parameters and suggest several other possible systematic errors. Overall, blind tests like these appear to provide significant opportunities for improving force fields and solvent models.

Small molecule hydration free energies in explicit solvent: An extensive test of fixed-charge atomistic simulations.

Using molecular dynamics free energy simulations with TIP3P explicit solvent, we compute the hydration free energies of 504 neutral small organic molecules and compare them to experiments. We find, first, good general agreement between the simulations and the experiments, with an RMS error of 1.24 kcal/mol over the whole set (i.e., about 2 kT) and a correlation coefficient of 0.89. Second, we use an automated procedure to identify systematic errors for some classes of compounds, and suggest some improvements to the force field. We find that alkyne hydration free energies are particularly poorly predicted due to problems with a Lennard-Jones well depth, and find that an alternate choice for this well depth largely rectifies the situation. Third, we study the non-polar component of hydration free energies - that is, the part that is not due to electrostatics. While we find that repulsive and attractive components of the non-polar part both scale roughly with surface area (or volume) of the solute, the total non-polar free energy does not scale with the solute surface area or volume, because it is a small difference between large components and is dominated by the deviations from the trend. While the methods used here are not new, this is a more extensive test than previous explicit solvent studies, and the size of the test set allows identification of systematic problems with force field parameters for particular classes of compounds. We believe that the computed free energies and components will be valuable to others in future development of force fields and solvation models.

Identification of a nonbasic, nitrile-containing cathepsin K inhibitor (MK-1256) that is efficacious in a monkey model of osteoporosis.

Herein, we report on the identification of nonbasic, potent, and highly selective, nitrile-containing cathepsin K (Cat K) inhibitors that are built on our previously identified cyclohexanecarboxamide core structure. Subsequent to our initial investigations, we have found that incorporation of five-membered heterocycles as P2-P3 linkers allowed for the introduction of a methyl sulfone P3-substitutent that was not tolerated in inhibitors containing a six-membered aromatic P2-P3 linker. The combination of a five-membered N-methylpyrazole linker and a methyl sulfone in P3 yielded subnanomolar Cat K inhibitors that were minimally shifted (<10-fold) in our functional bone resorption assay. Issues that arose because of metabolic demethylation of the N-methylpyrazole were addressed through introduction of a 2,2,2-trifluoroethyl substituent. This culminated in the identification of 31 (MK-1256), a potent (Cat K IC 50 = 0.62 nM) and selective (>1100-fold selectivity vs Cat B, L, S, C, H, Z, and V, 110-fold vs Cat F) inhibitor of cathepsin K that is efficacious in a monkey model of osteoporosis.

Discovery of [(3-bromo-7-cyano-2-naphthyl)(difluoro)methyl]phosphonic acid, a potent and orally active small molecule PTP1B inhibitor.

A series of quinoline/naphthalene-difluoromethylphosphonates were prepared and were found to be potent PTP1B inhibitors. Most of these compounds bearing polar functionalities or large lipophilic residues did not show appreciable oral bioavailability in rodents while small and less polar analogs displayed moderate to good oral bioavailability. The title compound was found to have the best overall potency and pharmacokinetic profile and was found to be efficacious in animal models of diabetes and cancer.