Encoding protein-ligand interaction patterns in fingerprints and graphs

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Protein-ligand interactions are crucial for drug efficacy.
• Comparing these interactions efficiently is challenging.
• Existing methods lack a universal, coordinate-invariant descriptor.

Purpose of the Study:

• To introduce a novel, universal method for describing protein-ligand interaction patterns.
• To develop tools for aligning protein-ligand complexes based on these patterns.
• To demonstrate the utility of the method in drug discovery applications.

Main Methods:

• A novel 210-integer fingerprint (TIFP) was developed to capture molecular interaction patterns.
• Interactions (hydrophobic, H-bond, ionic, etc.) are detected and described by pseudoatoms.
• TIFP fingerprints were computed for ~10,000 protein-ligand complexes.
• Two alignment tools, Ishape and Grim, were created based on TIFP.

Main Results:

• The TIFP fingerprint is coordinate frame-invariant and applicable to any protein-ligand complex.
• Interaction pattern similarity was shown to strongly correlate with binding site similarity.
• Ishape and Grim enable complementary protein-ligand complex alignments.
• The TIFP method was successfully applied to interaction-biased alignment, docking pose refinement, and virtual screening.

Conclusions:

• The TIFP fingerprint provides a simple yet powerful descriptor for protein-ligand interactions.
• The companion alignment tools enhance the analysis of protein-ligand complex relationships.
• This approach offers significant potential for advancing drug discovery and molecular modeling.