Goal of research: development of new mathematical methods and algorithms for multiscale model in material science and molecular biology of biomembranes based on atomistic approaches, implementations of these methods using modern supercomputer technologies and application to selected scientific problems. Analysis of supercomputer algorithms and technology efficiency.
Methodology: methods of classical molecular dynamics and statistical physics; Einstein-Smoluchowski and Green-Kubo methods for calculating diffusion coefficients and viscosity; for biological systems: homology modeling of proteins, molecular dynamics with parameters optimized for proteins and membranes, de novo protein structure prediction using the molecular hydrophobic potential approach.
Empirical base of research: open source software (LAMMPS, Gromacs, VASP, etc.) was used for calculations on high-performance computing equipment, including the latest generation of graphics accelerators.
Results of research: Using the method of molecular dynamics and Green-Kubo calculations, dependences of the density and shear viscosity were obtained for 2,2,4-trimethylhexane in the range of 1-10 kilobar pressure. The results were presented on 10th competition on modeling the properties of industrial liquids. The resulting model can be recommended for reliable calculations of the viscosity of hydrocarbons at pressures up to 5-6 thousand atmospheres.
Molecular aspects of protein-protein recognition and protein-lipid interactions are described for a number of critical cell receptors: the ErbB family, glycophorin A, neuraminidase-1, and model peptides. It has been shown that membrane properties can significantly affect the behavior of proteins, in particular, modulate the stability of active and inactive states of receptors. This provides the basis for the future creation of membrane-active compounds with given functions.
The configuration of a “trap” is proposed as a prototype of promising antimicrobial compounds targeting the “Achilles heel” of bacterial cells - lipid II molecule. We underlined the key role of hydrogen bonding in the formation of a stable configuration of a “trap” with pyrophosphate.
We performed tests of modern molecular dynamics codes using graphics accelerators and possible ways for further optimization are proposed. Also, the efficiency of calculations on the Russian Elbrus-9C processors is evaluated, and the possibility of achieving high performance is demonstrated on test problems.
Level of implementation, recommendations on implementation or outcomes of the implementation of the results: computational performance estimates and energy efficiency of software code on different processor architectures are proposed to be used in development of new supercomputer systems in NRU HSE.
