Efficient prediction of binding free energies, binding modes and protein conformations for multiple ligands from molecular dynamic simulations.

Computational prediction of ligand binding affinity and orientation and preferred conformations of protein-ligand complexes are of proven relevance for the understanding of protein-drug interactions and for drug design purposes. These issues can be addressed highly efficiently using the one-step perturbation (OSP) approach based on only a few molecular dynamics (MD) simulations. The method takes full account of the protein flexibility, including the entropic contribution to the free energy of binding (DG). Physically meaningful free energies are calculated from simulations of unphysical reference compounds, involving atoms that are allowed to overlap with each other from time to time. The OSP method, which we have recently developed for and applied to ligand-protein interactions, is a promising approach but needs to be further developed to allow for both diversity and flexibility in the ligands that are being studied. So far, including structurally diverse ligands is only possible for rigid molecules, while ligand flexibility restricts the choice of ligands to a common scaffold. In this proposal, structural diversity in ligands to which the OSP method can be applied is introduced by the design of novel reference compounds that resemble multiple real ligands simultaneously. Ligand flexibility can be enhanced by incorporating automated docking approaches in the OSP approach or by simulating the ligand degrees of freedom in a changing protein environment. The calculations will be performed on currently available (clusters of) computers or on massively parallel supercomputers. To explore this latter possibility, contact has been established with IBM Research in Almaden, USA, where the use of the recently developed Blue Gene supercomputer will be explored. The proposed OSP method will be applied to relevant biological systems for which binding affinities and modes, structural rationalizations of these and possibly induced fit effects will be predicted.