Clustering Protein Binding Pockets and Identifying Potential Drug Interactions: A Novel Ligand-Based Featurization...

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Protein-ligand interactions are crucial for drug discovery, influencing both therapeutic efficacy (on-target) and safety (off-target effects).
• Current methods often characterize proteins at a broad family, function, or pathway level, lacking granular detail for precise interaction prediction.

Purpose of the Study:

• To develop a novel ligand-based featurization and mapping approach for human protein pockets.
• To enable identification of closely related protein targets and prediction of drug interactions within a hybrid protein-ligand feature space.

Main Methods:

• Protein pockets are characterized using ligands that bind to their best co-complex template matches from the Protein Data Bank (PDB).
• This structure-based template matching allows for a granular, protein-pocket level characterization of the human proteome.
• The featurization method was applied to cluster a subset of the human proteome.

Main Results:

• The novel featurization provides a granular characterization of protein pockets, moving beyond traditional protein-level classifications.
• Clustering analysis demonstrated the method's ability to group related protein pockets.
• The approach was validated by evaluating the predicted cluster associations for over 7000 compounds.

Conclusions:

• This ligand-based featurization offers a simple, interpretable, and granular method for characterizing protein pockets.
• The approach enhances the prediction of protein-ligand interactions, crucial for both identifying effective therapeutics and assessing potential safety concerns.
• This method advances drug discovery by providing a more detailed understanding of the human proteome at the protein-pocket interaction level.