Predicting multiple ligand binding modes using self-consistent pharmacophore hypotheses.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of ligand binding modes is crucial for accelerating drug discovery.
• Current virtual screening methods still require experimental validation for precise binding modes.
• Experimental structural biology is often used to guide lead optimization after initial compound identification.

Purpose of the Study:

• To develop an improved method for predicting protein-ligand binding modes for known binders.
• To enhance the accuracy of virtual docking predictions by focusing on known interacting ligands.
• To reduce the reliance on wet-lab experiments in the early stages of drug discovery.

Main Methods:

• The algorithm performs traditional protein-ligand docking for known binders.
• Candidate binding modes are evaluated for self-consistency with a generated pharmacophoric map.
• The approach identifies a set of poses that maximally exploit key interaction points within the active site.

Main Results:

• The algorithm achieved predictions with an average RMSD < 2.5 Å across four tested protein systems.
• This represents a significant improvement of 0.5-1.0 Å (up to 25%) over naive docking predictions.
• The method demonstrated effectiveness for thrombin, CDK2, DHFR, and HIV-1 protease.

Conclusions:

• The developed algorithm offers a substantial improvement in protein-ligand binding mode prediction accuracy.
• This method is independent of the specific docking technique used.
• It has the potential to significantly enhance the reliability of virtual docking in drug discovery research.