FRED and HYBRID docking performance on standardized datasets.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Molecular docking is crucial for identifying potential drug candidates.
• Accurate prediction of ligand-protein interactions is essential for drug design.
• Evaluating docking software performance on standardized datasets is vital for progress.

Purpose of the Study:

• To assess the docking and virtual screening performance of FRED and HYBRID programs.
• To compare the effectiveness of ligand- and structure-based approaches in molecular docking.
• To investigate the impact of accounting for protein flexibility on docking accuracy.

Main Methods:

• Evaluation of FRED and HYBRID programs using two standardized datasets from the Docking and Scoring Symposium.
• Assessment of cognate docking accuracy and virtual screening performance (measured by AUC).
• Analysis of HYBRID's performance with and without consideration of protein flexibility using multiple crystal structures.

Main Results:

• FRED achieved 70% success in cognate docking within 2 Å and a mean AUC of 0.75 in virtual screening.
• HYBRID improved the mean AUC to 0.78 by integrating ligand- and structure-based information.
• HYBRID's mean AUC increased to 0.80 when accounting for protein flexibility, with minimal increase in docking time.

Conclusions:

• HYBRID demonstrates superior virtual screening performance compared to FRED.
• Incorporating protein flexibility significantly enhances HYBRID's docking accuracy.
• Both FRED and HYBRID are valuable tools for molecular docking, with HYBRID offering advanced capabilities for drug discovery.