SOM

SOM implements the Self-Organizing Map algorithm in Python to cluster and visualize conformational ensembles of biological macromolecules, enabling analysis of conformational space sampling and large-scale dataset mining.

Key Features:

• Scalability: Supports large-scale SOMs with computation time that scales linearly with dataset size.
• Descriptor computation: Calculates descriptors for each conformation in a trajectory for downstream mapping and analysis.
• U-matrix visualization: Maps descriptors onto a 3D U-matrix that represents distances between neighboring neurons to reveal data structure.
• Flooding algorithm for cluster identification: Uses a hierarchical flooding algorithm to delineate clusters by identifying local basins from global minima to maxima.
• Comprehensive workflow implementation: Implements an end-to-end analysis workflow from data input to result interpretation within the library.

Scientific Applications:

• Conformational space analysis: Extraction of meaningful patterns and clusters from conformational ensembles of macromolecules.
• Molecular dynamics and interaction studies: Aid in elucidating molecular dynamics and interaction-related structural diversity.

Methodology:

Descriptors are calculated for each molecular conformation and mapped onto a U-matrix using the SOM algorithm; the flooding algorithm then processes the U-matrix to identify and delineate clusters hierarchically.