Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Tuesday Speaker Abstracts

TRACKING PHOTOSYNTHETIC REACTANT AND PRODUCT DIFFUSION ACROSS CYANOBACTERIAL CARBOXYSOMES ON EXASCALE COMPUTING PLATFORMS Daipayan Sarkar 1 ; Josh V Vermaas 1 ; 1 Michigan State University, Plant Research Laboratory, East Lansing, MI, USA Molecular simulation algorithms depend on rapidly evaluating Newton’s equations of motion across a moderate number of particles for a large number of timesteps. Advances in modern high performance computing architectures have driven algorithmic changes to develop GPU-resident molecular simulation engines. These advances have had a profound impact on the types of questions that can be addressed by molecular simulation at low cost. One example from our group are explicit solvent simulations for a model cyanobacterial carboxysome. The carboxysome is an organelle found in photosynthetic bacteria that locally concentrates carbon dioxide to improve the efficiency for RuBisCO, the key enzyme in photosynthetic carbon fixation. The carboxysome encapsulates RuBisCO and carbonic anhydrase, which is an enzyme that converts soluble bicarbonate into lipophilic carbon dioxide, increasing local carbon dioxide concentration for RuBisCO. Leveraging these new GPU-resident molecular simulation engines, we determine the permeability for the carboxysome to RuBisCO reactants and products through unbiased simulation. We find that the carboxysome itself is not selectively permeable to bicarbonate over carbon dioxide, as originally hypothesized. Instead, the carboxysome shell proteins form a general barrier to maintain the carbon dioxide gradient generated by carbonic anhydrase activity within the carboxysome. We highlight that the multimillion atom scale for this system would have required substantial computational resources as recently as a few years ago. However, utilizing new GPU architectures, systems at this scale achieve excellent performance that offer new opportunities for molecular simulation as it moves into the exascale regime.

17