CLoNe

CLoNe performs clustering of large biomolecular structural ensembles to identify conformational states in protein datasets derived from molecular dynamics and integrative modeling.

Key Features:

• Algorithm basis: Builds on the Density Peaks algorithm of Rodriguez and Laio (2014) for cluster identification.
• Single parameter ('pdc'): Uses a probability density cutoff ('pdc') as the sole adjustable parameter to control cluster granularity, with integer values between 1 and 10 typically sufficient; increasing pdc reduces the number of clusters while decreasing it increases them.
• Local density estimation: Performs probabilistic analysis of local density distributions based on nearest neighbors.
• Shape- and size-invariant clustering: Identifies relevant clusters irrespective of their shape, size, distribution, or quantity.
• Ensemble input support: Operates on large structural ensembles produced by molecular dynamics and integrative modeling approaches.
• Conformational event detection: Extracts meaningful conformations such as membrane binding events and ligand-binding pocket openings.
• Structural motif recognition: Detects dominant dimerization motifs and inter-domain organizations within protein ensembles.
• Cluster export: Saves identified clusters as individual trajectories for downstream analysis.
• Visualization integration: Provides scripts for automated integration with molecular visualization software.

Scientific Applications:

• Conformational ensemble analysis: Identifying dominant and rare conformational states in protein structural ensembles from molecular dynamics and integrative modeling.
• Binding and activation studies: Detecting membrane association events and ligand-binding pocket openings relevant to function.
• Quaternary and domain organization: Characterizing dimerization motifs and inter-domain arrangements in protein complexes.
• Robust clustering of complex datasets: Clustering structural datasets where clusters vary in shape, size, distribution, or number.

Methodology:

Implements the Density Peaks framework (Rodriguez & Laio 2014) with a probabilistic local density estimation based on nearest neighbors and a single probability density cutoff ('pdc') parameter to control cluster number (typical integer values 1–10).