CommonNNClustering─A Python Package for Generic Common-Nearest-Neighbor Clustering

Area of Science:

• Data Science
• Computational Chemistry
• Biophysics

Background:

• Density-based clustering is vital for data science, adept at finding complex cluster shapes and handling outliers.
• These methods are particularly useful in molecular dynamics simulations for identifying low-energy conformations.
• Such conformations are key for building kinetic Markov-state models.

Purpose of the Study:

• To present CommonNNClustering, an open-source Python package.
• To offer an efficient and user-friendly reimplementation of the common-nearest-neighbor (CommonNN) clustering method.
• To facilitate hierarchical clustering and result evaluation in data analysis.

Main Methods:

• Implementation of the common-nearest-neighbor (CommonNN) clustering algorithm.
• Development of functionalities for hierarchical clustering.
• Design of a generic API for flexibility and customization.

Main Results:

• An open-source Python package, CommonNNClustering, is now available.
• The package provides efficient and easy-to-use tools for density-based clustering.
• Functionalities for hierarchical clustering and result evaluation are included.

Conclusions:

• CommonNNClustering offers a flexible and customizable solution for density-based clustering.
• The package enhances the analysis of molecular dynamics simulations.
• It provides a foundation for constructing kinetic Markov-state models.