An overview of the SAMPL8 host-guest binding challenge

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Drug Discovery

Background:

• The SAMPL (Statistical Assessment of the Modeling of Proteins and Ligands) series focuses on computational modeling challenges relevant to pharmaceutical drug discovery.
• SAMPL8 specifically addressed the prediction of absolute binding affinities in host-guest systems.

Purpose of the Study:

• To assess the performance of various computational methods in predicting binding affinities for host-guest systems.
• To identify successful strategies and areas for improvement in molecular simulation techniques for drug discovery.

Main Methods:

• Blind challenge format evaluating participant predictions against experimental data.
• Utilized molecular simulations, including molecular dynamics (MD) and alchemical methods.
• Employed empirical corrections and hybrid approaches combining MD with quantum mechanics (QM) potentials.

Main Results:

• Predictive accuracy for binding affinities was mixed across participating methods.
• Empirical corrections, leveraging prior binding data, improved accuracy.
• A hybrid MD approach with reweighting to a force-matched QM potential performed well.
• Alchemical methods using the AMOEBA-polarizable force field achieved the best results for cavitand systems (RMSE < 1 kcal/mol).

Conclusions:

• Computational methods for predicting binding affinities show room for improvement.
• Empirical corrections and advanced simulation techniques like hybrid MD and polarizable force fields enhance prediction accuracy.
• Sensitivity to factors like ethyl group sampling and guest starting pose highlights critical areas for future methodological development.