RCDPeaks

RCDPeaks: Memory-Efficient Density Peaks Clustering for Long Molecular Dynamics Simulations

RCDPeaks performs clustering of long molecular dynamics (MD) simulation trajectories using the DP+ implementation of the Density Peaks algorithm, enabling large-scale trajectory analysis with substantially reduced memory consumption.

Key Features:

• DP+ Density Peaks Implementation: Implements the DP+ variant of the Density Peaks algorithm to reduce RAM usage; clusters trajectories of 1,000,000 frames using <4.5 GB RAM compared to >2 TB required by standard implementations.
• Computational Performance: Achieves approximately threefold faster clustering than the fastest alternative while maintaining lower memory demands.
• Automatic Parameter Selection: Automatically determines clustering parameters, eliminating manual tuning.
• Center Candidate Screening: Identifies cluster centers based on high local density and large relative distance from other high-density points.
• Geometrical Refinement: Applies post-clustering geometrical refinement to improve cluster accuracy in MD trajectory analysis.

Scientific Applications:

• Long-Timescale MD Analysis: Clusters large molecular dynamics trajectories for studies of protein dynamics, ligand interactions, and complex molecular systems over extended simulation timescales.

Methodology:

RCDPeaks applies the DP+ implementation of the Density Peaks algorithm to compute local point densities and relative distances, select cluster centers, assign remaining points to nearest higher-density neighbors, and refine resulting clusters using geometrical criteria.