Diffusion maps, clustering and fuzzy Markov modeling in peptide folding transitions

Area of Science:

• Computational chemistry
• Molecular dynamics
• Data science

Background:

• Molecular dynamics simulations generate vast datasets.
• Analyzing conformational space is crucial for understanding molecular behavior.
• Traditional methods may oversimplify complex dynamics.

Purpose of the Study:

• To demonstrate diffusion maps for analyzing molecular dynamics trajectories.
• To visualize and partition conformational space effectively.
• To construct robust Markov state models for conformational dynamics.

Main Methods:

• Diffusion maps for dimensionality reduction of molecular dynamics data.
• Fuzzy C-means clustering for partitioning conformation space.
• Transition-based state assignment for Markov state model construction.

Main Results:

• Diffusion maps provide effective low-dimensional representations of conformational dynamics.
• Identified clusters show good agreement with previous methods, capturing additional degrees of freedom.
• Robust Markov state models were constructed, suppressing non-Markovian effects.

Conclusions:

• Diffusion maps are powerful tools for analyzing and visualizing molecular dynamics.
• Manifold learning techniques complement traditional approaches in molecular conformation space analysis.
• This approach enhances the construction of reduced models for molecular dynamics.