Exhaustive docking and solvated interaction energy scoring: lessons learned from the SAMPL4 challenge

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• The Solvated Interaction Energy (SIE) platform is evaluated for its prospective performance in predicting molecular interactions.
• The SAMPL4 datasets, including HIV-integrase inhibitors and host-guest systems, present a rigorous test for computational methods.

Purpose of the Study:

• To prospectively assess the Wilma-solvated interaction energy (SIE) platform for pose prediction, binding affinity prediction, and virtual screening.
• To evaluate new features of the docking algorithm and scoring function in the Wilma-SIE platform.

Main Methods:

• Prospective assessment of the Wilma-solvated interaction energy (SIE) platform on SAMPL4 datasets.
• Testing new docking algorithm and scoring function features.
• Analysis of correlations between predicted and experimental binding affinities.

Main Results:

• Wilma-SIE demonstrated good correlations with experimental binding affinities for host-guest systems.
• Absolute binding affinities were reproduced with appropriate scoring function training or comparative entropy estimation.
• For HIV-integrase ligands, SIE predictions showed limited correlation with affinities below 2 kcal/mol but correctly identified narrow dynamic ranges.
• Virtual screening for HIV-integrase yielded better-than-random results, with over a third of ligands docked within 2 Å of their actual poses.

Conclusions:

• The Wilma-SIE platform shows potential for binding affinity prediction and virtual screening, especially for systems with well-defined binding sites.
• Improvements in solvation treatment and using common protein structures can enhance prediction accuracy.
• The platform's applicability domain needs careful consideration, particularly for flexible and promiscuous binders.