A Neural-Network-Based Mapping and Optimization Framework for High-Precision Coarse-Grained Simulation

Zhixuan Zhong,Lifeng Xu,Jian Jiang

2024-08-13

Abstract:The accuracy and efficiency of a coarse-grained (CG) force field are pivotal for high-precision molecular simulations of large systems with complex molecules. We present an automated mapping and optimization framework for molecular simulation (AMOFMS), which is designed to streamline and improve the force field optimization process. It features a neural-network-based mapping function, DSGPM-TP (Deep Supervised Graph Partitioning Model with Type Prediction). This model can accurately and efficiently convert atomistic structures to CG mappings, reducing the need for manual intervention. By integrating bottom-up and top-down methodologies, AMOFMS allows users to freely combine these approaches or use them independently as optimization targets. Moreover, users can select and combine different optimizers to meet their specific mission. With its parallel optimizer, AMOFMS significantly accelerates the optimization process, reducing the time required to achieve optimal results. Successful applications of AMOFMS include parameter optimizations for systems such as POPC and PEO, demonstrating its robustness and effectiveness. Overall, AMOFMS provides a general and flexible framework for the automated development of high-precision CG force fields.

Computational Physics

What problem does this paper attempt to address?

The paper aims to address the issue of force field parameter optimization in high-precision Coarse-Grained (CG) simulations. Specifically, the paper proposes an Automated Mapping and Optimization Framework for Molecular Simulations (AMOFMS) for force field development in molecular simulations. The main objectives include: 1. **Automated Mapping**: Using a deep learning-based method (DSGPM-TP model) to automatically map atomic-level structures to coarse-grained representations, reducing the need for manual intervention. 2. **Parameter Optimization Framework**: Designing a comprehensive framework that allows users to freely combine top-down macroscopic property optimization methods and bottom-up microscopic statistical optimization methods, and supports the selection of various optimizers. 3. **Accelerating the Optimization Process**: Significantly speeding up the optimization process through parallel optimizers, reducing the time required to achieve optimal results. 4. **Generality and Flexibility**: Providing a general and flexible framework applicable to different types of molecular systems, capable of handling systems ranging from simple small organic molecules to complex biological macromolecules. The key contributions of the framework include: - **Automated Mapping Functionality**: Simplifying the complex and labor-intensive CG mapping process, improving the reproducibility of molecular simulations. - **Flexible Optimization Framework**: Allowing users to define and customize optimization parameters based on specific needs. - **Multi-Objective Optimization**: Integrating top-down and bottom-up methods to ensure the force field reflects both microscopic properties and matches macroscopic characteristics. - **Parallel Optimizers**: Greatly accelerating the optimization process, making it possible to handle large and complex systems. The paper validates the effectiveness and accuracy of AMOFMS through application cases in different molecular systems (such as the POPC lipid bilayer system and the polyethylene oxide (PEO) system). These cases demonstrate AMOFMS's ability to improve the consistency of simulation results with experimental data, particularly in simulating structural and dynamic behaviors.

A Neural-Network-Based Mapping and Optimization Framework for High-Precision Coarse-Grained Simulation

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

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

A 3D Structure Mapping-Based Efficient Topology Optimization Framework

Optimization of Reactive Force Field Simulation: Refactor, Parallelization, and Vectorization for Interactions

Machine Learning Directed Optimization of Classical Molecular Modeling Force Fields

Simulate Time-integrated Coarse-grained Molecular Dynamics with Multi-Scale Graph Networks

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

Efficient and Precise Force Field Optimization for Biomolecules Using DPA-2

General Multiobjective Force Field Optimization Framework, with Application to Reactive Force Fields for Silicon Carbide

Structured Mesh-Oriented Framework Design and Optimization for a Coarse-Grained Parallel CFD Solver Based on Hybrid MPI/OpenMP Programming

A review of advancements in coarse-grained molecular dynamics simulations

The automated optimisation of a coarse-grained force field using free energy data

Machine Learning of coarse-grained Molecular Dynamics Force Fields

An Initial Design-enhanced Deep Learning-based Optimization Framework to Parameterize Multicomponent ReaxFF Force Fields

Development and Evaluation of Geometry Optimization Algorithms in Conjunction with ANI Potentials.

Enhanced GROMACS: toward a better numerical simulation framework

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

Optimizing Force Fields with Experimental Data Using Ensemble Reweighting and Potential Contrasting

Adaptive coupling of a deep neural network potential to a classical force field

An Extendible, Graph-Neural-Network-Based Approach for Accurate Force Field Development of Large Flexible Organic Molecules