Computational fragment-based approach at PDB scale by protein local similarity

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• The increasing availability of protein-ligand structures facilitates advanced computational Fragment-Based Drug Design (FBDD) protocols.
• Developing efficient methods to mine and utilize this structural data is crucial for accelerating drug discovery.

Purpose of the Study:

• To create a novel database of MED-Portions, which are structural objects representing protein-fragment binding sites.
• To develop and validate a computational protocol for identifying potential drug leads using MED-Portions, MED-SuMo, and MED-Hybridise.

Main Methods:

• A database of MED-Portions was constructed by mining available protein-ligand structures.
• The MED-SuMo software was used to superpose similar protein interaction surfaces, enabling retrieval of matching MED-Portions.
• The MED-Hybridise toolkit was employed to combine MED-Portions in 3D for generating hit-like molecules.

Main Results:

• The protocol was applied to 107 diverse protein binding sites, demonstrating its rapidity and applicability.
• A qualitative correlation between MED-Portion hydrophobicity and binding site hydrophobicity indicated protocol selectivity.
• Active molecules were retrieved for a protein kinase and a G-Protein Coupled Receptor (GPCR) using PubChem bioassays.

Conclusions:

• The developed MED-Portion/MED-SuMo/MED-Hybridise protocol shows significant potential for identifying relevant drug leads from any protein 3D structure.
• The high occurrence of interfamily MED-Portions (25% for protein kinase, ~100% for GPCR) highlights the broad applicability of the method across protein superfamilies.