Improving Protein-Small Molecule Structure Predictions with Ensemble Methods, or Using Computers to Guess How Tiny Things Fit Together

Abstract

Protein-small molecule structure prediction, or protein-ligand docking, is a computational method for modeling how binding partners will interaction on an atomic level. Accurate prediction of protein-small molecule interactions is an important step in the structure based drug discovery pipeline. Biological molecules are flexible and adopt different conformational shapes when binding with small molecules. Capturing this flexibility while maintaining computational efficiency is a critical challenge for docking software. This research developed novel methods within the Rosetta Macromolecular Modeling Suite to consider structural ensembles of proteins and small molecules during docking. The additional structural information is complemented with experimental structure-activity relationship data, which previously was only considered retroactively. The new ensemble docking methods was applied in collaboration to targets of pharmaceutical interest including metabotropic glutamate receptors, protease-activated receptors, and STAT proteins.