Gaussian-mixture umbrella sampling.

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Biophysics

Background:

• Exploring free energy landscapes is crucial for understanding molecular processes.
• Traditional methods struggle with high-dimensional free energy surfaces.
• Efficiently escaping local minima is a key challenge in molecular simulations.

Purpose of the Study:

• Introduce a novel biased molecular dynamics technique, Gaussian-mixture umbrella sampling (GAMUS).
• Develop a method to efficiently escape free energy minima in multidimensional problems.
• Enable the identification of free energy minima in complex reaction coordinates.

Main Methods:

• GAMUS employs adaptive umbrella sampling for efficient exploration.
• Prior simulation data are reweighted using a maximum likelihood formulation.
• The approximate probability density is modeled using a Gaussian-mixture model.

Main Results:

• GAMUS successfully escapes free energy minima in multidimensional systems.
• The method was applied to alanine dipeptide (2D) and tripeptide (4D) systems.
• Demonstrated ability to identify free energy minima in complex reaction coordinates.

Conclusions:

• GAMUS is an effective technique for exploring complex free energy landscapes.
• The method enhances the efficiency of molecular dynamics simulations.
• GAMUS provides a robust approach for identifying critical minima in multidimensional problems.