Docking pose selection by interaction pattern graph similarity: application to the D3R grand challenge 2015.

Area of Science:

• Computational chemistry and structural biology
• Drug discovery and development
• Bioinformatics and cheminformatics

Background:

• High-affinity ligands often share conserved interaction patterns with target proteins.
• Accurate prediction of protein-ligand binding poses is crucial for drug design.
• Existing scoring functions may struggle with diverse chemical scaffolds and limited structural data.

Purpose of the Study:

• To develop and evaluate a novel topological knowledge-based scoring function (GRIM) for rescoring protein-ligand docking poses.
• To assess GRIM's performance against state-of-the-art methods using benchmark datasets from the D3R Grand Challenge 2015.

Main Methods:

• GRIM converts protein-ligand atomic coordinates into a 3D graph representing interaction patterns.
• Proposed interaction graphs are compared to those derived from Protein Data Bank (PDB) template structures.
• A GRIM score is calculated based on the overlap of maximum common subgraphs.

Main Results:

• GRIM achieved high-quality top-ranked poses for HSP90α inhibitors (mean RMSD 1.06 Å), outperforming a state-of-the-art scoring function.
• For MAP4K4 inhibitors with greater chemical diversity and less prior structural knowledge, GRIM's accuracy decreased (mean RMSD 3.18 Å) but still surpassed an energy-based scoring function.
• GRIM ranked 3rd and 2nd in the D3R Grand Challenge 2015 for HSP90α and MAP4K4 datasets, respectively, demonstrating robustness.

Conclusions:

• GRIM is an effective rescoring method that leverages conserved interaction patterns for accurate docking pose prediction.
• The method is simple to implement, independent of docking engines, and applicable to targets with available X-ray structures.
• GRIM shows promise as a valuable tool in computational drug discovery, particularly when structural information is available.