clusttraj: A Solvent-Informed Clustering Tool for Molecular Modeling.

Area of Science:

• Computational chemistry and molecular dynamics simulations.
• Development of novel algorithms for data analysis in cheminformatics.

Background:

• Molecular modeling generates vast datasets requiring robust analysis techniques.
• Standard clustering methods often struggle with atomic permutation in identical molecules, inflating root mean square deviation (RMSD).

Purpose of the Study:

• To introduce clusttraj, a new Python package for improved molecular configuration clustering.
• To address the challenge of atomic permutation in RMSD calculations for enhanced clustering accuracy.

Main Methods:

• The clusttraj program utilizes reordering schemes combined with the Kabsch algorithm to minimize RMSD.
• It employs a hierarchical clustering protocol with automated threshold determination using evaluation metrics.
• The package is designed as a Python library for flexible integration.

Main Results:

• Clusttraj effectively corrects inflated RMSD values caused by atomic permutations.
• The program's performance is validated on diverse systems, including pure water, solvated proteins, and small solutes.
• Analysis reveals the impact of system size and reordering methods on clustering outcomes.

Conclusions:

• Clusttraj offers a significant advancement in clustering molecular dynamics trajectories, particularly for systems with identical molecules.
• The solvent-informed approach and optimal pairing enhance the reliability of cluster analysis.
• The program is versatile and applicable to various molecular configuration datasets beyond solute-solvent systems.