Improving Force Field Accuracy by Training against Condensed-Phase Mixture Properties

Boothroyd, Simon; Madin, Owen; Mobley, David L.; Wang, Lee‐Ping; Chodera, John D.; Shirts, Michael R.

doi:10.1021/acs.jctc.1c01268

Cited by 20 publications

(75 citation statements)

References 100 publications

(143 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This leads to challenges with curating appropriate sets of experimental physical property data from the literature, as well as the computational cost of simulating sets of physical property data with molecular dynamics. Most physical properties used in training, which include densities [19], enthalpies of vaporization [19], enthalpies of mixing [26], solvation free energies [20] and dielectric constants [27,28], require equilibrium simulations in one or more phases, and in some cases may require alchemical simulation techniques [29]. In conjunction with the need to train against larger datasets to ensure accuracy and transferability, this makes optimization of LJ parameters a difficult problem.…”

Section: Non-bonded Training Is Expensive and Difficultmentioning

confidence: 99%

“…Using regularized least squares optimization with the L-BFGS-B algorithm [32], we minimized an objective function that captures the ability of a parameter set to reproduce physical property observables. Using this framework, we studied the benefits of including physical property data of mixtures in training LJ parameters [26], then applied that training method to a production force field, OpenFF 2.0.0 "Sage" [33].…”

Section: Non-bonded Training Is Expensive and Difficultmentioning

confidence: 99%

“…We test this approach by performing multi-fidelity LJ optimization for 12 LJ parameters from the OpenFF 1.0.0 "Parsley" force field. In training, we use a set of 56 pure compound physical properties curated in a previous paper [26]. We also benchmark the results against test sets curated in the same paper; while newer versions of OpenFF exist, using OpenFF 1.0.0 allows a direct comparison to our previous optimization.…”

Section: Surrogate Modeling Can Accelerate Non-bonded Trainingmentioning

confidence: 99%

“…All simulations use a 2 fs timestep and a Langevin integrator with BAOAB splitting [55]. More complete simulation details are available in our previous work [26], which uses the same simulation workflows.…”

Section: Physical Property Simulationsmentioning

confidence: 99%

“…We focused on two separate optimization tasks, both described in our previous study [26]. Both tasks optimize the same set of 12 LJ parameters (𝑅 𝑚𝑖𝑛∕2 and 𝜖 for 6 LJ SMIRKS types), and both use the same small molecules (alkanes, alcohols, ethers, esters and ketones) in the their training sets.…”

Section: Optimization Tasksmentioning

confidence: 99%

See 4 more Smart Citations

Using physical property surrogate models to perform multi-fidelity global optimization of force field parameters

Madin

Shirts

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Dispersion-repulsion interactions, commonly represented in atomistic force fields by the Lennard-Jones (LJ) potential, play an important role in the accuracy of molecular simulations. Training the force field parameters used in the LJ potential is challenging, generally requiring adjustment based on simulations of macroscopic physical properties. The computational expense of these simulations limits the size of training data set and number of optimization steps that can be taken, often requiring modelers to perform optimizations within a local parameter region. To allow for global LJ parameter optimization against large training sets, we introduce a multi-fidelity optimization technique which uses Gaussian process surrogate modeling to build inexpensive models of physical properties as a function of LJ parameters. This allows for fast evaluation of objective functions, greatly accelerating searches over parameter space. We use an iterative framework which performs global optimization at the surrogate level, followed by validation at the simulation level and surrogate refinement. Using this technique on two previously studied training sets, containing up to 195 physical property targets, we refit a subset of the LJ parameters for the OpenFF 1.0.0 ``Parsley'' force field. We demonstrate that this multi-fidelity technique can find improved parameter sets compared to a purely simulation-based optimization by searching more broadly and escaping local minima. In most cases, these parameter sets are transferable to other similar molecules in a test set. This multi-fidelity technique provides a platform for fast optimization against physical properties that can be refined and applied in multiple ways to the development of molecular models.

show abstract

Section: Non-bonded Training Is Expensive and Difficultmentioning

confidence: 99%

Section: Non-bonded Training Is Expensive and Difficultmentioning

confidence: 99%

Section: Surrogate Modeling Can Accelerate Non-bonded Trainingmentioning

confidence: 99%

Section: Physical Property Simulationsmentioning

confidence: 99%

Section: Optimization Tasksmentioning

confidence: 99%

See 3 more Smart Citations

Using physical property surrogate models to perform multi-fidelity global optimization of force field parameters

Madin

Shirts

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Nonbonded Force Field Parameters from MBIS Partitioning of the Molecular Electron Density Improve Thermophysical Properties Prediction of Organic Liquids

Pulido,

Macaya,

Vöhringer-Martinez

2024

J. Chem. Eng. Data

View full text Add to dashboard Cite

The accuracy of predicting thermophysical properties through molecular dynamics simulations is constrained by the precision of the models used to describe molecular interactions. The Open Force Field Initiative has established a computational structure to develop new models and introduced two nonpolarizable force fields, Parsley and Sage. Sage version 2.0.0 focused on refining Lennard-Jones parameters to accurately reflect thermophysical properties. In this context, we evaluate the ability of our introduced D-MBIS nonbonded force field parameters to replicate liquid densities and enthalpies of evaporation of 49 neutral compounds from the ThermoML database using the openff-evaluator package. Our findings confirm that our ab initio derived nonbonded force field parameters with an implicit description of the polarization accurately mirror both thermophysical properties with a high degree of precision.

show abstract

Open Force Field Evaluator: An Automated, Efficient, and Scalable Framework for the Estimation of Physical Properties from Molecular Simulation

Boothroyd

Wang

Mobley

et al. 2022

J. Chem. Theory Comput.

Self Cite

View full text Add to dashboard Cite

Developing accurate classical force field representations of molecules is key to realizing the full potential of molecular simulations, both as a powerful route to gaining fundamental insights into a broad spectrum of chemical and biological phenomena and for predicting physicochemical and mechanical properties of substances. The Open Force Field Consortium is an industry-funded open science effort to this end, developing open-source tools for rapidly generating new high-quality small-molecule force fields. An integral aspect of this is the parameterization and assessment of force fields against high-quality, condensed-phase physical property data, curated from open data sources such as the NIST ThermoML Archive, alongside quantum chemical data. The quantity of such experimental data in open data archives alone would require an onerous amount of human and computational resources to both curate and estimate manually, especially when estimations must be obtained for numerous sets of force field parameters. Here, we present an entirely automated, highly scalable framework for evaluating physical properties and their gradients in terms of force field parameters. It is written as a modular and extensible Python framework, which employs an intelligent multiscale estimation approach that allows for the automated estimation of properties from simulation and cached simulation data, and a pluggable API for estimation of new properties. In this study, we demonstrate the utility of the framework by benchmarking the OpenFF 1.0.0 small-molecule force field and GAFF 1.8 and GAFF 2.1 force fields against a test set of binary density and enthalpy of mixing measurements curated using the framework utilities. Further, we demonstrate the framework's utility as part of force field optimization by using it alongside ForceBalance, a framework for systematic force field optimization, to retrain a set of nonbonded van der Waals parameters against a training set of density and enthalpy of vaporization measurements.

show abstract

Improving Force Field Accuracy by Training against Condensed-Phase Mixture Properties

Cited by 20 publications

References 100 publications

Using physical property surrogate models to perform multi-fidelity global optimization of force field parameters

Using physical property surrogate models to perform multi-fidelity global optimization of force field parameters

Nonbonded Force Field Parameters from MBIS Partitioning of the Molecular Electron Density Improve Thermophysical Properties Prediction of Organic Liquids

Open Force Field Evaluator: An Automated, Efficient, and Scalable Framework for the Estimation of Physical Properties from Molecular Simulation

Contact Info

Product

Resources

About