High-Performance Computing of Microtubule Protofilament Dynamics by Means of All-Atom Molecular Modeling

Abstract

Molecular dynamics (MD) simulation is a useful tool for understanding biological systems at the level of individual molecules and atoms. However, studying such massive biological systems as microtubules and even their constituent components (tubulin protofilaments) takes an enormous amount of processing power. In this paper, using MD calculations of individual microtubule protofilaments, we demonstrate how computational architecture and calculation options affect computing performance. When using the “GPU-resident” option in the GROMACS MD package, you may gain a fantastic computation acceleration by using the newest high-end graphics processing unit (GPU), even in conjunction with a rather outdated central processing unit (CPU). For instance, MD of the biomolecular system containing a tubulin protofilament in an explicitly specified solvent consisting of more than 300 thousand atoms can be investigated with performance of 171 ns/day at time step 2 fs when using a single-node computer with the latest CPU and GPU generation architecture (Intel Core i9-13900K and Nvidia RTX4090 respectively). Nevertheless, high performance computing platforms (e.g., the volta2 partition of "Lomonosov-2" supercomputer) can be very suitable for simulation experiments with a large number of independent calculations, such as the umbrella sampling technique. Obtained results allow one to choose the best price-performance solution to study molecular dynamics of biological systems.