Validation of a Field-Based Ligand Screener Using a Novel Benchmarking Data Set for Assessing 3D-Based Virtual...

Area of Science:

• Computational Chemistry
• Drug Discovery
• Cheminformatics

Background:

• Ligand-based methods are vital for virtual screening when target 3D structures are unavailable.
• Existing validation datasets may not optimally suit ligand-based screening method assessments.

Purpose of the Study:

• To validate the CSD field-based ligand screener using a novel, comprehensive benchmarking dataset.
• To assess the impact of query models and parameter settings on screening performance.
• To compare the screener's performance against established methods using external benchmark datasets.

Main Methods:

• Development of a new benchmark dataset with 56 targets, derived from UniProt IDs, ChEMBL actives, and DUD-E decoys.
• Evaluation of the CSD field-based ligand screener's performance using various query models, including multi-ligand overlays.
• Calculation of enrichment scores using standard DUD-E, DUD-E+, and the specialized DUD_Lib_VS_1.0 datasets.

Main Results:

• The CSD field-based ligand screener demonstrated effective virtual screening performance on the novel dataset.
• Utilizing overlays of multiple flexible ligands as queries enhances efficiency without parallel computations.
• Performance varied based on parameter settings and query model choices, highlighting the need for optimization.

Conclusions:

• The CSD field-based ligand screener is a valuable tool for drug discovery, particularly when target structures are unknown.
• The novel benchmark dataset provides a robust platform for validating ligand-based virtual screening methods.
• The method's ability to use multi-ligand overlays offers a significant advantage in computational efficiency and screening power.