ИСТИНА

Molecular modeling should increase effectiveness of new inhibitors development. The reliable prediction is defined by the accuracy of docking and MD programs. The lecture will be devoted to main problems limiting accuracy of docking programs and how to solve them. Docking programs perform positioning of a compound (a ligand) in the active site of the target protein. Computed poses of the ligand are used for the calculation of the protein-ligand binding free energy which is directly connected with the inhibition constant. Docking programs are based on the paradigm which assumes that the ligand binding pose in the active site of the target protein corresponds to the global minimum of the protein-ligand energy or is near it and the docking problem is reduced to the global optimization problem on the multi-dimensional protein-ligand energy surface. Docking accuracy is defined by the completeness of the found low energy minima spectrum and by the accuracy of the configuration integral calculation in each of these minima. In this study we describe several docking programs including “classical” docking programs on the base of grid approximation for virtual screening of large ligand databases, e.g. the SOL program, and novel direct generalized docking programs FLM , SOL-T and SOL-P. The latter makes it possible to reject the rigid protein as well as the grid approximations, to take into account many proteins’ degrees of freedom and to increase the docking accuracy. Genetic and tensor train global optimization algorithms are brifly described as a base for docking programs. The latter make it possible to perform successful docking in the conformation space of 157 degrees of freedom: a flexible ligand and several dozen of moveable protein atoms. Mobility of protein atoms increases docking positioning accuracy. The important role of multi-processors supercomputer calculations in docking is demonstrated. Future perspectives the docking development are described.