Enhanced Conformational Sampling with an Adaptive Coarse-Grained Elastic Network Model Using Short-Time All-Atom...

Area of Science:

• Computational Biology
• Biophysics
• Molecular Dynamics

Background:

• All-atom molecular dynamics (AA-MD) simulations are computationally expensive.
• Existing coarse-grained (CG) methods struggle to sample conformations distant from the initial structure without biased forces.

Purpose of the Study:

• To develop a novel adaptive CG elastic network model (CG-ENM) for efficient and diverse molecular structure sampling.
• To reduce the computational cost associated with parameter searching in CG-ENM.

Main Methods:

• Developed an adaptive CG-ENM incorporating dynamic cross-correlation coefficients from short-time AA-MD.
• Utilized Bayesian optimization to efficiently search the adaptive CG-ENM parameter space, reducing search costs by approximately 90% compared to random or exhaustive methods.
• Applied the adaptive CG-ENM to adenylate kinase (ADK), glutamine binding protein (GBP), and integrin (αV).

Main Results:

• The adaptive CG-ENM generated more diverse structural ensembles than conventional ENMs and long-time AA-MD simulations.
• Sampled structures showed notable proximity to holo-type structures for ADK and GBP.
• Successfully sampled extended conformations for the larger biomolecule integrin (αV), demonstrating the method's scalability and effectiveness for significantly altered structures.

Conclusions:

• The developed adaptive CG-ENM offers a computationally efficient approach for exploring diverse protein conformational landscapes.
• This method significantly enhances the sampling capabilities of CG-MD simulations, enabling the study of large conformational changes.
• The adaptive CG-ENM holds promise for advancing our understanding of protein dynamics and function through cost-effective simulations.