Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Tuesday Speaker Abstracts

VISUALIZING THE SARS-COV-2 REPLICATION TRANSCRIPTION COMPLEX WITH AI-DRIVEN ADAPTIVE MULTISCALE SIMULATIONS Arvind Ramanathan 1,2 ; Anda Trifan 3 ; Defne Gorgun 3 ; Zongyi Li 5 ; Alexander Brace 2 ; Maxim Zvyagin 1 ; Heng Ma 1 ; Austin Clyde 1,2 ; David Clark 4 ; Tom Burnley 6 ; Vishal Subbiah 7 ; Jessica Liu 7 ; Venkatesh Mysore 4 ; Tom Gibbs 4 ; John Stone 3 ; C. Srinivas Chennubhotla 9 ; Emad Tajkhorshid 3 ; Anima Anandkumar 4 ; Venkatram Vishwanath 1 ; Sarah A Harris 10 ; Geoffrey Wells 8 ; 1 Argonne National Laboratory, Data Science and Learning, Lemont, IL, USA 2 University of Chicago, CASE, Hyde Park, IL, USA 3 University of Illinois Urbana-Champaign, Department of Biochemistry , Urbana-Champaign, IL, USA 4 NVIDIA Inc., Santa Clara, CA, USA 5 California Institute of Technology, Pasadena, CA, USA 6 Science and Technology Facilities Council, Didcot, United Kingdom 7 Cerebras Inc., Los Gatos, CA, USA 8 University College of London, London, United Kingdom 9 University of Pittsburgh, Computational and Systems Biology, Pittsburgh, PA, USA 10 University of Leeds, Physics and Astronomy, Leeds, United Kingdom The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) replication transcription complex (RTC) is a multi-domain protein responsible for replicating and transcribing the viral mRNA inside a human cell. Attacking RTC function with pharmaceutical compounds is a pathway to treating COVID-19. Conventional tools, e.g., cryo-electron microscopy and all-atom molecular dynamics (AAMD), do not provide sufficiently high resolution or timescale to capture important dynamics of this molecular machine. Consequently, we develop an iterative workflow that bridges the gap between these resolutions, using mesoscale fluctuating finite element analysis (FFEA) continuum simulations and a hierarchy of AI-methods that continually learn and infer features from simulations while maintaining consistency between AAMD and FFEA resolutions. We further leverage a multi-site distributed workflow manager to orchestrate AI, FFEA, and AAMD jobs, providing optimal resource utilization across HPC centers. Our AI- enabled multiscale simulations provides mechanistic insights into how the SARS-CoV-2 RTC machinery operates, in terms of backtracking the bound RNA across two different enzyme complexes, including the viral RNA-dependent RNA polymerase (RDRP) and the non-structural protein-13 (nsp13). The intrinsic correlations between the two rather large subunits points to a cooperative mechanism that can be potentially exploited to devise novel small molecules that can target the RTC. Further, the insights from this study also points to potentially 'missing' links between other RTC experimental datasets -- complementing knowledge from across multiple studies. We posit that such AI-informed multiscale simulation techniques hold promise in gaining fundamental insights into the mechanism of how large molecular machines function while complementing experimental observations and potentially providing feedback in improving their overall quality and accessibility.

18