Utilizing Machine Learning for Efficient Parameterization of Coarse Grained Molecular Force Fields

  title={Utilizing Machine Learning for Efficient Parameterization of Coarse Grained Molecular Force Fields},
  author={James L. McDonagh and Ardita Shkurti and David J Bray and Richard L. Anderson and Edward O. Pyzer-Knapp},
  journal={Journal of chemical information and modeling},
We present a machine learning approach to automated force field development in Dissipative Particle Dynamics (DPD). The approach employs Bayesian optimization to parameterize a DPD force field against experimentally determined partition coefficients. The optimization process covers a discrete space of over 40,000,000 points, where each point represents the set of potentials that jointly form a force field. We find that Bayesian optimization is capable of reaching a force field of comparable… 

Figures and Tables from this paper

Coarse-Grained Force Field Calibration Based on Multi-Objective Bayesian Optimization to Simulate Water Diffusion in Poly-ɛ-caprolactone.

A new calibration method based on multi-objective Bayesian optimization is developed to speed up the development of molecular dynamics force fields that are capable of predicting multiple properties accurately.

Machine Learning Directed Optimization of Classical Molecular Modeling Force Fields

A machine learning directed, multiobjective optimization workflow for force field parametrization that evaluates millions of prospective force field parameter sets while requiring only a small fraction of them to be tested with molecular simulations is presented.

A review of advancements in coarse-grained molecular dynamics simulations

ABSTRACT Over the last few years, coarse-grained molecular dynamics has emerged as a way to model large and complex systems in an efficient and inexpensive manner due to its lowered resolution,

A machine learning enabled hybrid optimization framework for efficient coarse-graining of a model polymer

The efficiency and transferability of the CG model is demonstrated through accurate predictions of chain statistics, the limiting behavior of the glass transition temperature, diffusion, and stress relaxation, where none were included in the parametrization process.

Distributed Multi-Objective Bayesian Optimization for the Intelligent Navigation of Energy Structure Function Maps For Efficient Property Discovery

This paper proposes the next evolution of the ESF map, which uses parallel Bayesian optimization to selectively acquire energy and property data, generating the same levels of insight at a fraction of the computational cost by limiting the expensive property calculations to a small fraction ofThe predicted crystal structures associated with a molecule.

A novel machine learning enabled hybrid optimization framework for efficient and transferable coarse-graining of a model polymer

The proposed framework combines the two fundamentally different classical optimization approaches for the development of coarse-grained model parameters; namely bottom-up and top-down approaches through integrating the optimization algorithms into a machine learning model, trained using molecular dynamics simulation data.

Accelerating computational discovery of porous solids through improved navigation of energy-structure-function maps

This work uses parallel Bayesian optimization to selectively acquire energy and property data and achieves a two orders of magnitude speedup on an ESF study that focused on the discovery of molecular crystals for methane capture, saving more than 500,000 central processing unit hours from the original protocol.

Performance efficient macromolecular mechanics via sub-nanometer shape based coarse graining

Overall, SBCG provides a simple yet robust approach to coarse graining that requires minimal user input and lacks any ad hoc interactions between protein domains, and takes full advantage of the latest GPU-accelerated NAMD3 yielding molecular sampling of over a microsecond per day for systems that span micrometers.

Coarse-grained molecular dynamics study based on TorchMD

The workflow in this work provides another option to study the protein folding and other relative processes with the deep learning CG model and shows that the main phenomenon of protein folding with TorchMD CG model is the same as the all-atom simulations, but with a less simulating time scale.

Gaussian Process Regression for Materials and Molecules

The focus of the present review is on the regression of atomistic properties: in particular, on the construction of interatomic potentials in the Gaussian Approximation Potential (GAP) framework; beyond this, the fitting of arbitrary scalar, vectorial, and tensorial quantities is discussed.



Bayesian parametrization of coarse-grain dissipative dynamics models.

A new bottom-up method based on Bayesian optimization of the likelihood to reproduce a coarse-grained reference trajectory obtained from analysis of a higher resolution molecular dynamics trajectory is introduced, related to force matching techniques, but using the total force on each grain averaged on a coarse time step instead of instantaneous forces.


  • Lei HuangB. Roux
  • Chemistry, Physics
    Journal of chemical theory and computation
  • 2013
This work proposes a method, General Automated Atomic Model Parameterization (GAAMP), for generating automatically the parameters of atomic models of small molecules using the results from ab initio quantum mechanical (QM) calculations as target data.

A Bayesian statistics approach to multiscale coarse graining.

Bayes' theorem, an advanced statistical tool widely used in signal processing and pattern recognition, is adopted to further improve the MS-CG force field obtained from the CG modeling, and can regularize the linear equation resulting from the underlying force-matching methodology.

Perspective: Machine learning potentials for atomistic simulations.

  • J. Behler
  • Materials Science
    The Journal of chemical physics
  • 2016
Recent advances in machine learning (ML) now offer an alternative approach for the representation of potential-energy surfaces by fitting large data sets from electronic structure calculations, which are reviewed along with a discussion of their current applicability and limitations.

Development of DPD coarse-grained models: From bulk to interfacial properties.

The method is extended to improve transferability across thermodynamic conditions by developing a CG model of n-pentane from constant-NPT atomistic simulations of bulk liquid phases and applying the CG-DPD model to the calculation of the surface tension of the liquid-vapor interface over a large range of temperatures.

CAMELOT: A machine learning approach for coarse-grained simulations of aggregation of block-copolymeric protein sequences.

Simulations based on the CAMELOT approach are used to show that the adsorption and unfolding of the wild type N17 and its sequence variants on the surface of polyQ tracts engender a patchy colloid like architecture that promotes the formation of linear aggregates.

Machine Learning Predictions of Molecular Properties: Accurate Many-Body Potentials and Nonlocality in Chemical Space

A systematic hierarchy of efficient empirical methods to estimate atomization and total energies of molecules and is achieved by a vectorized representation of molecules (so-called Bag of Bonds model) that exhibits strong nonlocality in chemical space.

Deriving effective mesoscale potentials from atomistic simulations

It is shown how an iterative method for potential inversion from distribution functions developed for simple liquid systems can be generalized to polymer systems and it is proved that it is not possible to use a single force field for different concentration regimes.

Building a More Predictive Protein Force Field: A Systematic and Reproducible Route to AMBER-FB15.

The AMBER-FB15 protein force field was developed by building a high-quality quantum chemical data set consisting of comprehensive potential energy scans and employing the ForceBalance software package for parameter optimization, which allows for more significant thermodynamic fluctuations away from local minima.

Geometry Optimization with Machine Trained Topological Atoms

The geometry optimization of a water molecule with a novel type of energy function called FFLUX is presented, which bypasses the traditional bonded potentials, and kriging models are robust enough to optimize the molecular geometry to sub-noise accuracy.