Volume-scaled common nearest neighbor clustering algorithm with free-energy hierarchy.

Area of Science:

• Computational Chemistry
• Biophysics
• Statistical Mechanics

Background:

• Markov state modeling (MSM) combined with molecular dynamics (MD) simulations is powerful for studying slow molecular processes.
• Current MSM workflows often use conventional clustering, which may not be optimal for complex, Boltzmann-weighted MD data.
• Developing specialized algorithms for the discretization step in MSM is crucial for analyzing intricate molecular systems.

Purpose of the Study:

• To introduce a novel density-based clustering algorithm tailored for Boltzmann-weighted data from MD simulations.
• To adapt the common nearest neighbor (CNN) algorithm into a volume-scaled version (vs-CNN).
• To establish a direct link between clustering density and free energy for improved analysis of molecular landscapes.

Main Methods:

• Development of the volume-scaled common nearest neighbor (vs-CNN) clustering algorithm.
• Application of vs-CNN to discretize data from molecular dynamics simulations.
• Utilizing Boltzmann inversion to connect the density-based criterion to free energy.

Main Results:

• The proposed vs-CNN algorithm provides a density-based criterion directly related to free energy.
• This method facilitates a hierarchical approach to identify conformational clusters.
• The algorithm is suitable for analyzing rugged free-energy landscapes in complex molecular systems.

Conclusions:

• The vs-CNN algorithm offers an improved discretization strategy for MSM applied to MD data.
• It enables a more direct interpretation of molecular conformational states in terms of free energy.
• This advancement aids in the comprehensive study of complex molecular dynamics and their associated energy landscapes.