Over ten months have passed since the Partnership for Advanced Computing in Europe (PRACE) announced that it would be fast-tracking proposals to use supercomputing resources to fight the then-novel coronavirus beginning to sweep across the globe. Just a month later, PRACE – which aggregates and allocates supercomputing resources to enhance European competitiveness and serve the social good – had allocated hundreds of millions of core hours to dozens of projects aimed at ending the pandemic. Now, as COVID-19 cases begin to wane in many countries, PRACE is looking back on fourteen projects supported by its computing resources over the course of the pandemic.
PRACE’s roster boasts seven core supercomputers: Joliot-Curie at GENCI in France; JUWELS, Hawk and SuperMUC-NG at the Gauss Centre for Supercomputing in Germany; Marconi at CINECA in Italy; MareNostrum 4 at the Barcelona Supercomputing Center; and Piz Daint at ETH Zurich. Each of the below projects, highlighted in a series of articles on PRACE’s website, was supported by millions of core hours on one of those systems.
Understanding the virus at a molecular level
Many of the projects focused on understanding the SARS-CoV-2 virus on a molecular level, especially as it pertained to identifying drug targets. At the Max Planck Institute of Biophysics, for instance, researchers used PRACE resources to understand the structure and dynamics of the virus’ spike protein, which allows it to infect human cells. The team found that the spike protein was connected to the virus by way of a flexible, hinged stalk. “This is an aspect of the spike protein that only emerged through the simulations of our PRACE project,” said Gerhard Hummer, who led the project.
Hummer’s research also raised questions about the sugary glycan shields protecting the spike protein, which were the focus of another PRACE-supported project led by Elisa Fadda of Maynooth University. “Working alongside my colleague Rommie Amaro in San Diego, we have shown that SARS-CoV-2 is entirely unique from other viruses due to what is known as its glycan shield,” Fadda said. “What makes it unique in this coronavirus is that specific glycans within the shield are intrinsically involved in the mechanism of the spike protein that allows it to latch on to our cells. Without these glycans, the spike protein would be useless and the virus would not be contagious or dangerous in any way.” The multi-institutional work on modeling the glycan shield went on to win the 2020 Gordon Bell Special Prize for High Performance Computing-Based COVID-19 Research.
SARS-CoV-2’s RNA allows it to replicate, and has retained prominence over the last year as a potential target for therapeutics. Kresten Lindorff-Larsen and Sandro Bottaro of the University of Copenhagen took advantage of PRACE supercomputing-enabled molecular dynamics to better understand that RNA’s three-dimensional structure.
Jean-Philip Piquemal of Sorbonne University, meanwhile, took a deep dive into two of the main drug targets on the virus: the spike protein and the main protease. Using PRACE resources and his own “Tinker HP” molecular dynamics code, Piquemal ran what he called “the longest simulation ever done” (in terms of simulated time) in its class. “Using the hundreds of GPUs provided to us by our allocation from PRACE, we have been lucky enough to have access to a huge amount of computing power to throw at the problems we are trying to solve,” he said.
Hunting for effective drugs
Of course, with those drug targets modeled, simulations are necessary for the drugs themselves. Vangelis Daskalakis of the Cyprus University of Technology used PRACE-hosted molecular dynamics simulations to screen a large number of molecules for their ability to bind to targets on the virus..
Other researchers honed in more narrowly. Francesco Luigi Gervasio of University College London, for instance, is leading two PRACE-supported projects to identify therapeutics for COVID-19: one aiming to design peptides to block the virus from entering human cells, another aiming at a less notorious protein on the virus known as nonstructural protein 1.
Rebecca Wade of Heidelberg University is running another PRACE project examining the possible antiviral properties of classes of drugs that might be able to inhibit the spike protein by decreasing its flexibility. “Thanks to the resources provided by PRACE, we have been able to investigate [heparan sulphate proteoglycans] and heparin and how they both interact with the spike protein and with each other,” Paiardi said. “Although heparin itself would probably not be used as an antiviral due to its anticoagulant properties, we hope that by understanding its antiviral mechanisms we might be able to suggest some modifications to it that would make it more suitable for therapeutic use.”
Heparin is an old drug, and like Wade, many other researchers looked to our pharmaceutical past for possible weapons against COVID-19. At the University of Lugano USI in Switzerland, a team led by Vittorio Limongelli used free energy calculations to identify how existing pharmaceuticals could bind to SARS-CoV-2’s molecular targets, all supported by PRACE supercomputer time.
Companies also got in on the COVID-19 research. Dublin-based drug company Nuritas received a PRACE allocation to investigate peptides that could be used to slow the progression of COVID-19 in infected patients. “The main goal of this project is to find peptides that show activity against the virus responsible for the COVID-19 pandemic,” said Hansel Gómez Martínez, a research scientist at Nuritas. “More specifically, we are ideally looking for peptides that are already in the market or at the stage of advanced clinical trials so that they can quickly be repurposed for antiviral therapies.”
Analyzing how COVID-19 spreads
Some of the most important research conducted over the course of the pandemic was macroscopic, not microscopic. Catalan-based company Mitiga Solutions used PRACE’s supercomputing power to supercharge their early epidemic warning system, Epi-EWS, for a global rollout. Rafael Villanueva of the Polytechnic University of Valencia, meanwhile, used PRACE resources to conduct network modeling to help elucidate how COVID-19 spreads within populations.
Yet other researchers focused on how viral droplets behaved once they emerged from an infected person. Led by Gaetano Sardina from Chalmers University of Technology, the COVID-DROPLETS project leveraged PRACE supercomputers to investigate how far those droplets spread. “Our numerical simulations provide a more accurate estimation of what happens when someone sneezes than the classical model,” Sardinia explained. “We can see that the lifetime of droplets in a sneeze are much longer compared to isolated droplets in the same environment.”
Similarly, the CFDforCOVID project led by Florent Duchaine of CERFACS used a fluid dynamics model (run on PRACE computers) to simulate how those droplets spread in enclosed spaces like buildings and vehicles. “Overall, since we began this PRACE project around six months ago, we have been able to carry out five large simulations, so it has been very productive for us in that sense,” Duchaine said.
Looking to the future
Of course, COVID-19 isn’t standing still, either. The virus has mutated into some alarming variants over the last year, raising concerns over whether it could eventually “escape” antibodies – or even vaccines. Modesto Orozco of the Spanish Institute for Research in Biomedicine used PRACE resources to investigate how the virus evolved, and how it could mutate in the future.
“When a viral mutation negatively affects the binding, the virus does not progress,” Orozco said. “We see one sequence of it and then it disappears. However, when mutations improve the binding, it becomes imprinted in the next generation of viruses and becomes dominant very quickly. For instance, you can see that there is one mutation on the spike protein, originally observed in Spain, that somehow helps to expose the RBD and thus helps it bind to ACE2. This mutation has now colonised the whole of Europe. Of course, this particular area of research does not have a definitive end, as the virus is still spreading, and mutating and data continues to be gathered through genome sequencing.”
To learn more about any of these PRACE-supported projects, visit PRACE’s website here.