Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Wednesday Speaker Abstracts

LARGE SCALE PROTEIN-LIGAND BINDING FREE ENERGY CALCULATIONS IN THE CLOUD AND HPC CENTERS Vytautas Gapsys 1 ; David F Hahn 2 ; Carsten Kutzner 1 ; Gary Tresadern 2 ; David L Mobley 3 ; Christian Kniep 5 ; Austin Cherian 6 ; Ludvig Nordstrom 7 ; Markus Rampp 4 ; Helmut Grubmüller 1 ; Bert L de Groot 1 ; 1 Max Planck Institute for Multidisciplinary Sciences, Goettingen, Germany 2 Janssen Research & Development, Janssen Pharmaceutica N. V., Beerse, Belgium 3 University of California, Irvine, Irvine, CA, USA 4 Max-Planck Computing and Data Facility, Garching, Germany 5 Amazon Web Services, Amazon Development Center, Berlin, Germany Rational drug design benefits from molecular dynamics based protein-ligand binding free energy calculations which offer a remarkable accuracy in guiding ligand optimization process. While such calculations are computationally expensive, the advances in simulation engines, compute infrastructures and specialized software for the free energy calculation setup have now substantially reduced the barriers for using these methods in everyday projects. Here, I will present several large scale ligand screening studies illustrating how these calculations can be efficiently performed in HPC facilities and in the AWS cloud setting. Relying on the open source software (GROMACS and PMX) we can set up, perform and analyze large scale free energy calculations for hundreds of compounds in a matter of days. The use case of the simulations in an HPC center illustrates the ease with which such type of calculations can be scaled up: the throughput is limited only by the available resources. The cloud based services offer a possibility to scale up in a virtually unlimited manner, yet it is important to carefully select the compute architectures for the cost efficient calculations. The results obtained from the performed scans provide valuable information on the current state of the field in predicting free energy differences for pharmaceutically relevant targets. 6 Amazon Web Services, Singapore, Singapore 7 Amazon Web Services, London, United Kingdom

21