Abstract:A force-matching-based method for supervised machine learning (ML) of coarse-grained (CG) free energy (FE) potentials─known as multiscale coarse-graining via force-matching (MSCG/FM)─is an efficient method to develop microscopically informed CG models that are thermodynamically and statistically equivalent to the reference microscopic models. For low-resolution models, when the coarse-graining is at supramolecular scales, objective-oriented clustering of nonbonded particles is required and the reduced description becomes a function of the clustering algorithm. In the present work, we explore the dependence of the ML of the CG Helmholtz FE potential on the clustering algorithm. We consider coarse-graining based on partitional (k-means, leading to Voronoi diagram) and hierarchical agglomerative (bottom-up) clustering algorithms common in unsupervised ML and develop theory connecting the MSCG/FM learned CG Helmholtz potential and the clustering statistics. By combining the agglomerative clustering and the MSCG/FM learning in a recursive manner, we propose an efficient ML methodology to develop the fine-to-low resolution hierarchies of the CG models. The methodology does not suffer from degrading accuracy or increased computational cost to construct larger hierarchies and as such does not impose an upper size limitation of the CG particles resulting from the extended hierarchies. The utility of the methodology is demonstrated by obtaining the bottom-up agglomerative hierarchy for liquid nitromethane from all-atom molecular dynamics (MD) simulations. For agglomerative hierarchies, we prove the existence of renormalization group transformations that indicate self-similarity and allow for learning the low-resolution MSCG/FM potentials at low computational cost by rescaling and renormalizing the certain finer-resolution members of the hierarchy. The hierarchies of the CG models can be used to carry out simulations under constant-pressure conditions.

Statistically Optimal Force Aggregation for Coarse-Graining Molecular Dynamics

Machine Learning of coarse-grained Molecular Dynamics Force Fields

Learning Effective Molecular Models from Experimental Observables.

Two for One: Diffusion Models and Force Fields for Coarse-Grained Molecular Dynamics

Coarse Graining Molecular Dynamics with Graph Neural Networks

Systematic bottom-up molecular coarse-graining via force and torque matching using anisotropic particles

A Scalable Graph Neural Network Method For Developing An Accurate Force Field Of Large Flexible Organic Molecules

Learning coarse-grained force fields for fibrogenesis modeling

Ensemble learning of coarse-grained molecular dynamics force fields with a kernel approach.

Multi-body Effects in a Coarse-Grained Protein Force Field.

Machine learned coarse-grained protein force-fields: Are we there yet?

Synthetic Force-Field Database for Training Machine Learning Models to Predict Mobility-Preserving Coarse-Grained Molecular-Simulation Potentials

Learning data efficient coarse-grained molecular dynamics from forces and noise

Navigating protein landscapes with a machine-learned transferable coarse-grained model

Systematic coarse-graining of epoxy resins with machine learning-informed energy renormalization

Structural Coarse-Graining via Multi-Objective Optimization with Differentiable Simulation

Development of Coarse-Grained Models for Polymers by Trajectory Matching

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

Generative Coarse-Graining of Molecular Conformations

Hierarchical Machine Learning of Low-Resolution Coarse-Grained Free Energy Potentials

Machine learning assisted coarse-grained molecular dynamics modeling of meso-scale interfacial fluids