Intermolecular Non-Bonded Interactions from Machine Learning Datasets

Jia-An Chen,Sheng D. Chao

DOI: https://doi.org/10.3390/molecules28237900

IF: 4.6

2023-12-01

Molecules

Abstract:Accurate determination of intermolecular non-covalent-bonded or non-bonded interactions is the key to potentially useful molecular dynamics simulations of polymer systems. However, it is challenging to balance both the accuracy and computational cost in force field modelling. One of the main difficulties is properly representing the calculated energy data as a continuous force function. In this paper, we employ well-developed machine learning techniques to construct a general purpose intermolecular non-bonded interaction force field for organic polymers. The original ab initio dataset SOFG-31 was calculated by us and has been well documented, and here we use it as our training set. The CLIFF kernel type machine learning scheme is used for predicting the interaction energies of heterodimers selected from the SOFG-31 dataset. Our test results show that the overall errors are well below the chemical accuracy of about 1 kcal/mol, thus demonstrating the promising feasibility of machine learning techniques in force field modelling.

chemistry, multidisciplinary,biochemistry & molecular biology

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? The main goal of this paper is to utilize machine learning techniques to construct intermolecular non-bonded interaction force fields for organic polymers. Specifically, the authors attempt to address the following issues: 1. **Balancing accuracy and computational efficiency**: - Accurately determining intermolecular non-covalent interactions is crucial for molecular dynamics simulations of polymer systems. However, in force field modeling, maintaining both high accuracy and low computational cost is challenging. 2. **Energy data representation problem**: - Representing computed energy data as continuous force functions is a major challenge. Traditional methods often rely on selecting specific functional forms to fit discrete data, which can lead to weaker predictive capabilities. 3. **Effective utilization of training data**: - The authors use the SOGF-31 dataset, previously computed through quantum chemistry methods, as the training set and employ the CLIFF kernel-based machine learning scheme to predict the interaction energies of heterodimers in the SOGF-31 dataset. 4. **Validating the feasibility of machine learning in force field modeling**: - The test results show that the overall error is far below chemical accuracy (approximately 1 kcal/mol), demonstrating the potential feasibility of machine learning techniques in force field modeling. Through this research, the paper aims to showcase the potential of machine learning techniques in accurately describing intermolecular non-bonded interactions and to provide new methods for future research.

Intermolecular Non-Bonded Interactions from Machine Learning Datasets

A Machine Learning Force Field for Bio-Macromolecular Modeling Based on Quantum Chemistry-Calculated Interaction Energy Datasets

Efficient Machine Learning Force Field for Large-Scale Molecular Simulations of Organic Systems

PhyNEO: A Neural-Network-Enhanced Physics-Driven Force Field Development Workflow for Bulk Organic Molecule and Polymer Simulations.

Efficient interatomic descriptors for accurate machine learning force fields of extended molecules

Machine learning force fields for molecular liquids: Ethylene Carbonate/Ethyl Methyl Carbonate binary solvent

Accurate interatomic force field for molecular dynamics simulation by hybridizing classical and machine learning potentials

Self-consistent Coulomb interactions for machine learning interatomic potentials

Machine Learning Interatomic Potentials and Long-Range Physics

Accurate machine learning force fields via experimental and simulation data fusion

MolE8: Finding DFT Potential Energy Surface Minima Values from Force-Field Optimised Organic Molecules with New Machine Learning Representations

Machine Learning of Accurate Energy-Conserving Molecular Force Fields

Hybrid Classical/Machine-Learning Force Fields for the Accurate Description of Molecular Condensed-Phase Systems

Systematic improvement of empirical energy functions in the era of machine learning

Parametrization of Non-Bonded Force Field Terms for Metal-Organic Frameworks Using Machine Learning Approach

Machine Learning of coarse-grained Molecular Dynamics Force Fields

Machine Learning Directed Optimization of Classical Molecular Modeling Force Fields

Estimation of the Flory-Huggins interaction parameter of polymer-solvent mixtures using machine learning

Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules

Molecular Dynamics with On-the-Fly Machine Learning of Quantum-Mechanical Forces