"In order to correctly predict protein-ligand binding poses and free energies, it is essential to accurately take into account receptor flexibility. However, incorporating it into even the smallest region, for example, the binding site of a protein is computationally demanding. Even if this task can be accomplished there is a risk of running into false positives due to the enormous conformational space involved. So, it is the interplay between sampling and scoring. Nonetheless, to better mimic experiments, structure-based drug design methods need to identify and then incorporate the most populated receptor states in any docking and scoring campaign. This work addresses the development of a novel tool that has been implemented to incorporate receptor flexibility into the ligand-binding domain of a protein. This method enumerates conformational states on an energy landscape in a computationally tractable manner. The algorithm treats molecules at an atom pair level and uses a distance-based coordinate system, where each selected distance is associated with a pair-potential value selected from a look-up table. With a collection of conformations in hand, we then perform on-the-fly local partition function estimations on each of the 'seed structures' using the Movable Type (MT) method to estimate the associated free energy changes. This strategy helps to simultaneously generate relevant structures with the most favorable free energies. We initially applied our side chain flexibility method to a set of 159 protein-ligand systems and the docking score of Glide and our in-house Movable Type based scoring improved over the crystal docking score. Later we applied it to study the active site loop transitions seen in the Streptavidin-biotin system."--Pages ii-iii.
Read
