ProteinDBS

ProteinDBS compares protein tertiary structures by analyzing two-dimensional distance matrices derived from three-dimensional coordinates to identify structurally similar proteins.

Key Features:

• Real-Time Search Capability: Performs rapid retrieval of structurally similar proteins with ranked results returned within seconds against a database of over 152,959 Protein Data Bank (PDB) protein chains.
• Advanced Feature Extraction and Indexing: Converts two-dimensional distance matrices into multi-dimensional feature spaces and indexes them using an Entropy Balanced Statistical (EBS) k-d tree.
• Dual Query Methods: Accepts queries by PDB protein chain ID or by uploading new structures for comparison.
• Global-to-Local Comparison Techniques: Implements global-to-global and local-to-local comparison methods to compare entire structures and specific substructures.
• High Precision and Speed: Achieves high precision and fast computation times, with local comparison performance reported to exceed other methods on a non-redundant database of 3,276 proteins.

Scientific Applications:

• Functional and Evolutionary Analysis: Identifies structural similarities and common substructures to infer functional and evolutionary relationships among proteins.
• Domain and Fold Discovery: Detects shared substructures to aid discovery of new protein domains or folds.
• Database Curation Support: Highlights structural similarities to assist human curation and expansion of protein structure classifications.
• Comprehensive Structural Comparison: Enables simultaneous global and local structural comparisons for detailed structural analysis.

Methodology:

Extracts two-dimensional distance matrices from three-dimensional protein coordinates, applies computer visualization techniques to extract visual patterns, transforms matrices into multi-dimensional feature vectors via feature extraction algorithms, indexes features with an Entropy Balanced Statistical (EBS) k-d tree, and performs comparisons using global-to-global and local-to-local methods.