Adaptive Sampling Methods for Molecular Dynamics in the Era of Machine Learning

Area of Science:

• Computational biophysics
• Molecular dynamics simulations
• Protein dynamics

Background:

• Molecular dynamics (MD) simulations are essential for studying protein free energy landscapes.
• Sampling protein conformational changes is difficult due to long time scales.
• Enhanced sampling methods are needed to overcome these limitations.

Purpose of the Study:

• To provide a comprehensive overview of adaptive sampling algorithms for enhanced molecular dynamics simulations.
• To discuss the principles, guidelines, and applications of adaptive sampling.
• To highlight recent advancements, particularly those involving deep learning.

Main Methods:

• Focus on adaptive sampling techniques that enhance sampling without biasing forces.
• Discuss theoretically transparent adaptive sampling methods.
• Review select adaptive sampling methods applied to realistic systems.
• Examine deep learning-powered adaptive sampling approaches.

Main Results:

• Adaptive sampling preserves the thermodynamic ensemble while improving sampling efficiency.
• Various adaptive sampling methods have been successfully applied to complex biological systems.
• Deep learning is emerging as a powerful tool for advancing adaptive sampling methodologies.

Conclusions:

• Adaptive sampling offers a robust strategy for efficient conformational sampling in MD simulations.
• Continued development, especially with AI integration, promises further breakthroughs in computational protein dynamics.
• Understanding and applying adaptive sampling is crucial for advancing molecular simulations.