Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Poster Abstracts

7-POS Board 7 THE MULTISCALE PROBLEM OF MICROTUBULE CATASTROPHE AND ITS EXASCALE CHALLENGES Daniel Beckett 1 ; Gregory A. Voth 1 ; 1 University of Chicago, Chicago, IL, USA Microtubules (MTs) constitute the largest components of the eukaryotic cytoskeleton and facilitate a plethora of cellular functions. During mitosis, microtubules form the mitotic spindle, making them a potent drug target for many successful chemotherapeutic agents that interfere with dynamic instability (DI): the ability of microtubules to rapidly switch from polymerizing to depolymerizing states (known as catastrophe) and vice-versa. MT catastrophe is heralded by the hydrolysis of guanosine triphosphate (GTP) bound by the tubulin heterodimer which induces conformational and dynamical changes in MT structure that are still not fully understood. Unravelling the mystery of MT catastrophe is a multiscale problem with exascale challenges at each level and the objective of these studies is a complete molecular picture of the processes underlying catastrophe. Results will be presented for a suite of MT catastrophe-related simulations. First, the catalytic mechanism of GTP hydrolysis in the MT lattice is derived from quantum mechanics/molecular mechanics (QM/MM) hybrid simulations paired with metadynamics free energy sampling. These are long timescale (>100 ps) simulations for the computationally intensive QM portion with relatively large system sizes that require advances in processing speed to be practically tractable, with the presented results requiring millions of CPU hours to acquire. Second, all-atom (AA) molecular dynamics (MD) studies describing the release pathway of inorganic phosphate post-GTP hydrolysis will be presented. These require multiple replicas of >100 ns MD trajectories across multiple systems to converge. Lastly, AA MD studies of a full MT segment will be presented as well as progress on the classification of MT dynamics using a coarse-grained/machine learning approach. These full MT simulations require both very large system size and long timescales. Future studies required to understand MT dynamics will be discussed, as well as their exascale demands.

48