Scoring ligand similarity in structure-based virtual screening.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Accurate scoring of high-affinity compounds is crucial for effective virtual screening.
• Existing protein-ligand scoring methods focus on molecular interactions, while ligand-based methods assess shape and chemical similarity to a reference.

Purpose of the Study:

• To test the hypothesis that a hybrid scoring approach can improve virtual screening efficacy.
• To combine protein-ligand scoring with ligand-based scoring to identify high-ranking molecules that are complementary to the binding site and mimic a known ligand.

Main Methods:

• A hybrid scoring strategy was developed, using protein-ligand scoring to select candidates followed by ligand-based scoring (EON) for ranking.
• This approach was applied to screen approximately 70,000 National Cancer Institute (NCI) compounds for thrombin inhibitors.
• Analysis of top-scoring compounds was performed using five different protein-ligand scoring functions and EON scoring with three reference compounds.

Main Results:

• The hybrid approach successfully identified the majority (4/5) of newly discovered low to mid-micromolar thrombin inhibitors.
• Protein-ligand scoring alone identified only one new inhibitor.
• There was minimal overlap between compounds selected by protein-ligand versus ligand-based scoring, indicating complementary information.

Conclusions:

• A hybrid scoring approach, integrating protein-ligand and ligand-based methods, significantly enhances the identification of high-affinity compounds in virtual screening.
• The choice of scoring function profoundly impacts screening outcomes, even among functions measuring similar interactions.
• Matchprint analysis within the SLIDE toolset confirmed conserved interactions in top-scoring compounds, particularly in the S1 pocket.