Abstract:
Supercomputer docking is becoming one of the main tools at the initial stage of drug development - at the stage of developing inhibitors of specified therapeutic target proteins. This was convincingly shown by publications during the COVID-19 pandemic devoted to the development of inhibitors of the SARS-CoV-2 coronavirus, which caused the pandemic. Inhibitors block the work of the target protein, which plays a key role in the development of the disease, and thus slow down or stop the development of the disease. Rational drug development takes place in the form of a continuous pipeline, including the search for inhibitors, experimental verification of their inhibitory activity, preclinical animal trials and clinical trials in humans. Inhibitors are the basis of medicines. Docking is a molecular modeling method used to find the position of an inhibitor candidate molecule (ligand) in the active center of a target protein and evaluate the binding energy of the ligand to the protein. The higher the binding energy of the ligand to the protein, the more likely it is that such a compound will show inhibitory activity in experiments, the more effective a drug based on such an inhibitor will be. The efficiency of the entire development pipeline, duration, and development costs depend on the accuracy of the docking. The docking paradigm assumes that the ligand binds in the active center of the target protein near the global minimum energy of the protein-ligand system. Therefore, docking programs rely on algorithms for searching for a global minimum on a complex multidimensional energy surface. The multidimensionality of this energy surface is determined by the number of degrees of freedom of the protein-ligand system, which can be 10-20 even for a rigid protein. How to calculate energy? Force field and quantum chemistry (mechanics). Three docking programs created at Moscow State University are considered: SOL, FLM, and SOL-P, using the MMFF94 force field. SOL is a classic docking program using a global optimization genetic algorithm and a grid of pre-calculated interaction potentials of ligand atoms with a protein; SOL is adapted for virtual screening of many thousands (up to 1 million) molecules on the Lomonosov-2 supercomputer. FLM and SOL-P are a new generation of supercomputer multiprocessor docking programs, which differ from all existing docking programs in the absence of many simplifications and approximations, including the absence of pre-calculated grids of potentials for interactions of ligand atoms with a protein. The FLM program is based on the massive parallel execution of local energy optimizations of the protein-ligand system from random ligand positions while varying the positions of all ligand atoms. The SOL-P generalized docking program is based on a global optimization method based on the decomposition of multidimensional tensors into tensor trains. Quantum chemical postprocessing. Quantum quasi-docking is the way to create quantum docking. Some examples of success in the search for new inhibitors.
