SAMPL9 blind predictions using nonequilibrium alchemical approaches

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Supramolecular Chemistry

Background:

• The Statistical Assessment of the Modeling of Proteins and Ligands (SAMPL) challenge benchmarks computational methods for molecular interactions.
• Host-guest chemistry, particularly the binding of cationic guests to macrocyclic hosts like pillararenes, is crucial for molecular recognition and drug design.

Purpose of the Study:

• To provide blind predictions for the SAMPL9 host-guest challenge.
• To evaluate the performance of a novel virtual double system single box approach for calculating absolute binding free energies.
• To assess the binding affinities of WP6 (carboxy-pillar[6]arene) with ammonium and diammonium guests.

Main Methods:

• Utilized the virtual double system single box approach with enhanced sampling of bound and unbound states.
• Employed fast switching nonequilibrium alchemical simulations for free energy calculations.
• Applied Pearson and Kendall correlation coefficients to assess prediction accuracy.

Main Results:

• Achieved acceptable performance in predicting host-guest binding free energies.
• Demonstrated improved accuracy compared to previous submissions involving calixarenes and cyclodextrins.
• Observed a large mean signed error attributed to finite charge correction methods.

Conclusions:

• The nonequilibrium approaches are reliable for absolute binding free energy calculations.
• The virtual double system single box method shows promise for host-guest binding predictions.
• Further refinement of charge correction methods may improve accuracy in future calculations.