FINDSITE: a threading-based approach to ligand homology modeling

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Virtual screening is crucial for drug discovery but faces limitations.
• Remotely related proteins often share binding sites and similar ligands.
• Conserved anchor functional groups in ligands dictate binding site interactions.

Purpose of the Study:

• To develop a novel computational method for drug discovery.
• To exploit conserved ligand features for improved protein-ligand binding prediction.
• To enhance the accuracy and efficiency of virtual screening.

Main Methods:

• Developed FINDSITE(LHM) using homology modeling and structural information from weakly related proteins.
• Employed conserved anchor functional groups and variable ligand regions to predict binding modes.
• Benchmarked FINDSITE(LHM) against classical docking approaches using large-scale datasets.

Main Results:

• FINDSITE(LHM) achieved an average ligand RMSD of approximately 2.5 Å, outperforming classical docking.
• Successfully recovered binding residues and specific contacts in weakly homologous receptor models.
• Demonstrated significantly improved enrichment factors in virtual screening for HIV-1 protease inhibitors.

Conclusions:

• FINDSITE(LHM) is an accurate and computationally inexpensive algorithm for drug discovery.
• The method effectively utilizes conserved ligand features for predicting binding sites.
• FINDSITE(LHM) shows promise for assisting in the discovery of novel biopharmaceuticals.