Comparison of topological, shape, and docking methods in virtual screening.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Virtual screening is crucial for identifying drug candidates.
• Benchmarking different virtual screening methods is essential for optimizing drug discovery pipelines.
• Comparing ligand-based and structure-based approaches provides insights into their relative effectiveness.

Purpose of the Study:

• To benchmark the performance of 2D ligand similarity, 3D ligand similarity, and protein structure-based docking methods.
• To evaluate the impact of chemical typing on 3D ligand similarity.
• To identify the most effective virtual screening strategies for drug discovery.

Main Methods:

• Evaluated 2D ligand similarity (Daylight, TOPOSIM), 3D ligand similarity (SQW, ROCS), and docking (FLOG, FRED, Glide) across 11 targets.
• Assembled active and decoy compound sets from MDDR and Merck databases.
• Utilized mean enrichment factor to assess method performance.

Main Results:

• Ligand-based methods generally outperformed docking algorithms on average.
• 3D ligand similarity methods required chemical typing to surpass docking.
• Glide and FRED showed the best performance among docking methods.
• Performance varied significantly across different targets and databases.

Conclusions:

• Ligand-based virtual screening, particularly 3D similarity with chemical typing, is a highly effective strategy.
• Docking methods like Glide and FRED offer competitive performance.
• The choice of virtual screening method and database is critical for successful drug discovery.