PDBspheres: a method for finding 3D similarities in local regions in proteins

Area of Science:

• Structural bioinformatics
• Computational drug discovery
• Protein structure analysis

Background:

• Identifying structural similarities in protein ligand-binding sites is crucial for drug discovery and understanding protein function.
• Existing methods may not comprehensively capture the nuances of binding site structures and their relationships.

Purpose of the Study:

• To introduce PDBspheres, a novel structure-based computational method for finding and evaluating structural similarities in protein regions relevant to ligand binding.
• To demonstrate the efficacy of PDBspheres in detecting, characterizing, and clustering protein binding sites.

Main Methods:

• Development of PDBspheres, utilizing an extensive library of over 2 million protein structure regions ('spheres') derived from the Protein Data Bank (PDB).
• Application of the Local-Global Alignment (LGA) algorithm for structural alignment and similarity detection between a protein of interest and library spheres.
• Implementation of an all-atom-based similarity metric, incorporating side chain placement for enhanced confidence assessment of structural matches.

Main Results:

• PDBspheres successfully detects and characterizes binding sites in protein structures.
• The method demonstrated robust performance on a curated dataset of 2528 ligand-bound and ligand-free crystal structures.
• Application to the PDBbind 2019 dataset allowed for effective clustering of pockets and assessment of structural similarities across 4876 protein binding sites.

Conclusions:

• PDBspheres provides a powerful, strictly structure-based approach for analyzing protein binding site similarities.
• The method facilitates the discovery of structurally related binding sites, aiding in ligand design and functional annotation.
• PDBspheres contributes a valuable tool for large-scale structural analysis and comparison of protein binding pockets.