Divide and Cluster: The DIVINE Framework for Deterministic Top-Down Analysis of Molecular Dynamics Trajectories

Area of Science:

• Computational Chemistry
• Biophysics
• Data Science

Background:

• Molecular dynamics (MD) simulations generate large datasets requiring robust analysis.
• Existing clustering methods for MD trajectories often lack determinism or scalability.
• Efficiently partitioning complex conformational landscapes remains a challenge.

Purpose of the Study:

• Introduce DIVIsive N-ary Ensembles (DIVINE), a deterministic clustering framework for MD trajectories.
• Address the limitations of existing methods regarding scalability, reproducibility, and computational cost.
• Provide a robust and interpretable tool for analyzing complex molecular systems.

Main Methods:

• DIVINE employs a top-down, recursive splitting approach based on n-ary similarity principles.
• It avoids computationally expensive O(N^2) pairwise distance matrix calculations.
• Incorporates deterministic anchor initialization (NANI) and weighted variance for cluster selection.

Main Results:

• DIVINE achieved comparable or superior clustering quality to bisecting k-means on a 305 μs HP35 folding trajectory.
• Demonstrated significant reductions in runtime and elimination of stochastic variability.
• Enabled efficient exploration of clustering resolutions in a single pass.

Conclusions:

• DIVINE offers a scalable, interpretable, and deterministic alternative for MD trajectory clustering.
• Its single-pass design enhances efficiency for analyzing large-scale molecular simulations.
• The framework is available as part of the open-source MDANCE package.