Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Poster Abstracts

11-POS Board 11 EXASCALE READY MACHINE LEARNING DRIVEN MULTISCALE INFRASTRUCTURE TO EXPLORE RAS-RAF BIOLOGY Peer-Timo Bremer 1,2 ; Helgi I Ingo´lfsson 3 ; JDAC4C Pilot 2 Consortium 1 ; 1 Lawrence Livermore National Laboratory, CASC, Livermore, CA, USA 2 University of Utah, SCI, Salt Lake City, Utah (UT), USA

3 Lawrence Livermore National Laboratory, Physical and Life Sciences, Livermore, CA, USA Biophysical simulations at various scales can provide important insights into biological mechanisms. These can range from continuous macro-scale simulations that approach common experimentally observable length and timescales and resolve phenomena, such as protein clustering, to atomistic molecular dynamics simulations resolving pheonomena, such as bonding or signaling. However, understanding a complex process such as RAS-RAF signaling requires information at all scales and more importantly an understanding how scales interact. We will present recent results from the MuMMI multiresolution framework that uses advanced machine learning to directly couple massively parallel simulation ensembles at three computing scales – continuum, coarse-grain MD, and atomistic MD – to provide unprecendented insights into RAS- RAF interactions. In particular, we will report results from a large-scale simulation campaign of RAS-RAF proteins on a complex plasma membrane performed on and scalled across all of Summit, the worlds 2 nd largest supercomputer.

70