Supercomputing Software and Hardware

Article # 10_2020

18 с.

PDF

Nikolay A. Anikin, Aleksandr Y. Muskatin, Mikhail B. Kuz’minskiy, Sergey N. Lednev, Aleksandr V. Smirnov, Aleksandr I. Rusakov
Cardinal acceleration of calculations of giant biomolecules by quantum chemistry methods, requiring the use of supercomputers and / or GRID systems

Calculations of the electronic structure of molecules by quantum chemical methods long ago are performed using supercomputers. Today they are conducted on the leader of the TOP500 supercomputer list and will be realized on the first exaflops supercomputer in the USA.
A brief review of modern quantum chemistry methods and their supercomputer application for calculations of primarily large molecules shows the need for accelerated approximation techniques to realize the possibilities of such computations. The need is especially urgent for massive calculations of such giant biomolecules as protein-ligand docking complexes. For this, we have proposed and implemented software that gives excellent acceleration at an acceptable accuracy of approximation calculations for calculating the molecular integrals of non-empirical methods of quantum chemistry. For massive calculations of docking complexes using semiempirical methods, we propose and implement a new technique in software. It uses some localizations of ligand-protein interactions due to the formation of groups from a full set of ligands.
This method allows us to achieve acceleration by orders of magnitude and also intended for use in future non-empirical calculations. The proposed methods and programs for the necessary massive calculations of docking complexes naturally fit into a batch job processing system and can be used in a GRID environment. Such a GRID system arises on the computing resources of the Yaroslavl State University and the Institute of Organic Chemistry of the Russian Academy of Sciences on the base of EGI standardized UMD 4 software. (In Russian)

Key words: high speed quantum chemical methods, docking complexes, GRID.

article citation

 DOI

https://doi.org/10.25209/2079-3316-2020-11-2-75-92