| Protein-protein and protein-ligand interactions are ubiquitous in biology. For many proteins, these interactions can be well simulated by assuming rigid body association, resulting in powerful predictions from protein-protein docking. These methods have also been applied to sample ensembles of interactions on binding pathways, showing agreement with experimental measures of encounter complexes. A novel algorithm for the prediction of antibody-antigen complexes has been developed, accounting for the asymmetry of these interactions to significantly improve the prediction of these complexes over the current state of the art. Although the overall shape of the free energy surface is not affected by conformational changes, accounting for side chain flexibility generally increases the accuracy of predictions, but the required calculations are very expensive. The complexity of side chain search was substantially reduced by restricting considerations to key side chains with multiple low energy conformers and generating ensembles of their potential conformational states. The same algorithm was used to obtain low energy protein conformers to be studied by computational solvent mapping, a method developed for the identification of binding hot spots. The resulting set of conformers has been shown to account for most conformational changes between bound and unbound structures in protein-protein complexes and also enabled the opening of pockets capable of binding drug sized molecules. Mapping combined with side chain analysis was used for predicting the druggability of protein-protein interaction targets and developing initial fragment hits into lead-like ligand molecules. |
