Lessons learned in empirical scoring with smina from the CSAR 2011 benchmarking exercise

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Molecular docking is crucial for identifying potential drug candidates.
• Accurate scoring functions are essential for ranking docked poses.
• Existing tools like AutoDock Vina have limitations in high-throughput screening and customizability.

Purpose of the Study:

• To present a general methodology for designing empirical scoring functions.
• To introduce smina, an optimized version of AutoDock Vina for high-throughput scoring and custom functions.
• To create and evaluate a custom scoring function using the CSAR 2010 dataset.

Main Methods:

• Developed a general methodology for empirical scoring function design.
• Utilized smina, a specialized AutoDock Vina variant.
• Employed the CSAR 2010 dataset for training and CSAR 2011 for benchmarking.
• Performed crossdocking experiments to assess pose prediction accuracy.

Main Results:

• The custom scoring function created using smina demonstrated superior performance in sampling low RMSD poses during crossdocking compared to the default AutoDock Vina function.
• The study identified key insights into improving scoring and ranking of potential ligands.

Conclusions:

• The developed methodology and smina offer a powerful approach for creating accurate, custom scoring functions.
• The findings highlight areas for future advancements in molecular docking and ligand ranking for drug discovery.