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

Area of Science:

• Computational chemistry
• Molecular modeling
• Materials science

Background:

• Accurate and efficient coarse-grained (CG) force fields are essential for simulating large, complex molecular systems.
• Current methods for CG force field development often require significant manual intervention and can be time-consuming.

Purpose of the Study:

• To present an automated mapping and optimization framework for molecular simulation (AMOFMS).
• To streamline and enhance the process of developing high-precision CG force fields.
• To reduce manual effort and accelerate the optimization of CG force fields.

Main Methods:

• Developed a neural-network-based mapping function, DSGPM-TP (deep supervised graph partitioning model with type prediction), for atomistic-to-CG conversion.
• Integrated bottom-up and top-down methodologies for flexible optimization targets.
• Implemented a parallel optimizer to significantly accelerate the force field optimization process.

Main Results:

• DSGPM-TP accurately and efficiently converts atomistic structures to CG mappings.
• AMOFMS successfully optimized parameters for systems like POPC and PEO.
• The framework demonstrated robustness and effectiveness in developing high-precision CG force fields.

Conclusions:

• AMOFMS provides a general and flexible framework for automated CG force field development.
• The automated approach significantly reduces the time and manual effort required for force field optimization.
• This framework enables high-precision molecular simulations of large and complex systems.