Prediction of SAMPL4 host-guest binding affinities using funnel metadynamics

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Drug discovery

Background:

• Accurate prediction of binding affinities between ligands and macromolecules is vital for pharmaceutical research and development.
• Molecular dynamics simulations are powerful tools, but adequate conformational space sampling remains a significant challenge.

Purpose of the Study:

• To evaluate the efficacy of the funnel metadynamics method for predicting host-guest binding affinities.
• To assess the accuracy of binding free energy predictions for cucurbit[7]uril host-guest systems within the SAMPL4 blind challenge.

Main Methods:

• Application of funnel metadynamics to enhance ligand sampling in binding sites and solvated states.
• Utilizing molecular dynamics simulations with total times of 300-400 ns per ligand.
• Comparison of results against experimental data and assessment of force field performance.

Main Results:

• Funnel metadynamics reduced sampling errors to below 1 kcal/mol.
• Despite improved sampling, predictions did not outperform random guessing compared to experimental values.
• Significant differences (up to 11 kcal/mol) were observed between commonly used force fields.

Conclusions:

• While funnel metadynamics enhances sampling, accurate binding free energy prediction requires more rigorous force field parametrization.
• Further development in force field development is essential for reliable computational predictions in drug discovery.