Automatic multi-objective optimization of coarse-grained lipid force fields using SwarmCG.

Area of Science:

• Computational chemistry
• Molecular modeling
• Materials science

Background:

• Developing accurate coarse-grained (CG) molecular models requires extensive parameter tuning.
• Existing automatic methods often yield poorly transferable parameters.
• Transferability is crucial for applying CG models to diverse systems and conditions.

Purpose of the Study:

• To develop an automated approach for optimizing transferable CG molecular models.
• To create accurate and broadly applicable CG force fields (FFs).
• To investigate the precision-resolution balance in CG models.

Main Methods:

• Utilized a multi-objective optimization approach with the SwarmCG engine.
• Optimized CG lipid models against both high-resolution simulations and experimental data.
• Employed a 'bottom-up' (simulation data) and 'top-down' (experimental data) reference strategy.
• Ensured FF transferability by including diverse lipid bilayers in the training set.

Main Results:

• SwarmCG successfully optimized CG force fields (FFs) with good transferability.
• Achieved satisfactory agreement between simulated and experimental data (area per lipid, bilayer thickness).
• Provided insights into the trade-offs between CG model precision and resolution.
• Demonstrated the general applicability of the SwarmCG approach.

Conclusions:

• SwarmCG enables the efficient development of transferable CG molecular models.
• The method balances accuracy with broad applicability for molecular dynamics.
• This approach can be used to create novel FFs and refine existing ones.