Cherry-picking functionally relevant substates from long md trajectories using a stratified sampling approach.

Area of Science:

• Computational biology
• Biophysics
• Molecular modeling

Background:

• Molecular Dynamics (MD) simulations offer nanoscopic insights into protein dynamics, with current capabilities reaching the ns-μs timescale for large systems.
• Clustering MD trajectories is crucial for identifying distinct protein conformations and comparing simulation data with experimental results.
• Existing clustering methods often lack rigorous validation and benchmarking, especially for large proteins, and are typically applied to refined or simplified trajectory data.

Purpose of the Study:

• To propose and validate a robust strategy for classifying molecular dynamics (MD) simulation data.
• To benchmark the performance of various clustering algorithms and internal validation criteria for MD trajectory analysis.
• To compare the efficacy of trajectory classification in both real and principal component analysis (PCA) space.

Main Methods:

• Development of a systematic strategy for classifying MD simulation data.
• Benchmarking of multiple clustering algorithms and internal validation criteria.
• Application and comparison of classification methods on two diverse protein systems.
• Analysis of trajectory classification in both real and PCA space.

Main Results:

• Demonstration of a validated strategy for classifying large MD trajectories.
• Comparative analysis of clustering algorithm performance and validation criteria.
• Successful classification of trajectories for two distinct protein systems.
• Insights into the effectiveness of classification in real versus PCA space.

Conclusions:

• The proposed strategy offers a reliable method for analyzing and classifying large-scale MD simulation data.
• Rigorous benchmarking and validation are essential for accurate interpretation of clustering results in protein dynamics.
• This approach is particularly beneficial for analyzing trajectories from enhanced sampling techniques and large biomolecular systems.