Argonne Taps Supercomputing Network to Study How Coronavirus Spreads 

Argonne National Laboratory is leading and coordinating a variety of computational efforts to better understand both the COVID-19 disease and the SARS-CoV-2 virus that causes it—and to help accelerate the development of treatment options like antiviral drugs and vaccines. 

The Energy Department’s Illinois-based lab officials are pooling assets with others from a range of labs, universities and private research centers to jointly maximize the use of some top supercomputing resources to address the global health emergency caused by the novel coronavirus. 

“You're trying to do years worth of work in months or days,” Argonne’s Associate Laboratory Director for Computing, Environment and Life Sciences Rick Stevens recently told Nextgov about the work. “So, you know, we're kind of inventing how fast we can push these things—and of course, we're using methods that are really new.”

Stevens has seen a lot at Argonne, where he’s worked since 1982—but the recent collaboration sprung up by the pandemic seems almost unprecedented. He offered a glimpse into some of the lab’s high-performance computing endeavors supporting the governmentwide fight against COVID-19. 

The overall efforts include team members and assets from Argonne and Brookhaven national labs, J. Craig Venter Institute, and a range of universities including the Universities of Chicago, Illinois, Virginia, Texas, and California San Diego, as well as Rutgers, Stony Brook, and George Mason universities, the University College London—and more. Stevens said the network first started to come together several weeks ago when Argonne researchers were closely watching the spread of COVID-19 and realized “we have to get serious.”

Officials aimed to run cycles of a range of simulations, programs and models on the Argonne’s supercomputers to ideally arrive at new breakthroughs against the nascent virus. Stevens said cycles are essentially the “computer geek slang name for running on a processor.” They involve running an application continuously on a machine for durations that can last an hour, 100 hours or 1,000 hours, he said. And so upon that initial realization, he and his team started putting out calls to “friends,” and the broader supercomputing community to get efforts up and running.

“I sent out a note to the supercomputing centers saying ‘we need more cycles,’ because we were just flattening machines that we have access to,” Stevens explained. “So we tried sorting out where the different resources are that people were offering up and what programs needed to run on which machines—so we’re running on like eight or nine machines.”

It’s been several weeks since the Argonne team put out the first calls, and now the cadre of experts involved are working collaboratively from around the country, coordinating remotely over email and messaging services, and sharing data over the relevant infrastructures and Globus, a University of Chicago-run non-profit, secure research data management service.

Less than a month in, that network is “kind of running nonstop,” Steven said. He went on to highlight what some of the efforts entail—and aim to accomplish. 

Epidemiological Modeling, ‘Like SimCity’

Stevens said Argonne officials are producing and assessing epidemiological models to simulate and improve understanding around how the virus spreads across the population. These agent-based computational models can help represent what impacts the actions and interactions of agents—in other words, individuals or groups—may have on a system as a whole.

“This is modeling people going about their normal business,” Stevens explained. “Think of it as, you know, like SimCity right?”

According to Argonne’s own release on the work, the agent-based model that researchers constructed “includes almost 3 million separate agents, each of whom can travel to any of 1.2 million different locations.” The models can be programmed to have various transmission and interaction coefficients, as well as age distributions and other parameters. “That'll give you a baseline estimate of how fast the infection spreads and what the outcome is,” Stevens said. And from there, researchers can then use that model to test out a range of intervention strategies to see how actions like closing schools and restaurants or restricting flights might slow COVID-19’s spread. Stevens said the epidemiological efforts are aimed at addressing questions around the virus’ potential impact on hospitals and critical services, as well as elements of the infection curve and peak.

“That's running now on a number of the big machines,” he said. “And we are going to continue doing that for the next couple of months, I think.” 

One of the ultimate goals in this light, he said, is to help inform the possible policies the government might put in place to help slow the novel coronavirus’ spread down. 

Host Analysis

Viruses reproduce themselves in the cells of living hosts, and those hosts can be humans, animals and beyond. Stevens and his team are therefore conducting host analysis to better understand how hosts respond to COVID-19. As he puts it, when a virus infects a cell, that cell reacts, it changes the gene expression pattern, the immune system is activated and more.

“We want to try to understand what genetic signatures, whether it's gene expression patterns or maybe the underlying genome sequence, that determine or is related to how severe the disease is for a given individual,” Stevens said. “So can we predict, for example, before somebody is very symptomatic.”

This would allow medical professionals to potentially predict which patients might need a lot of attention early in the infection, and which might have an easier go at handling it. 

“We're collecting data. There's not a lot of data on this yet from COVID-19. But there is some data for related diseases,” Stevens said. “And so we're building some models that would allow us to do that.” 

Understanding COVID-19’s Evolution and How to Fight It

Sequences of virus’ genetic materials often prove to be vital in the development of relevant treatments. Stevens said there are hundreds of sequences now available for the novel coronavirus and thousands of sequences of related viruses. Honing in on them using advanced computing and simulation capabilities, Argonne and its collaborators are working to better understand crucial questions around where the virus came from, whether and how its adapting to its hosts, if it’s growing more or less virulent, and more. 

He added that scientists are mapping the mutational data they have on proteins to observe how much COVID-19 changes the active sites of the proteins. “Any drug is going to be targeting the active site and if the virus is mutating the active sites, then it means that the [to be developed] drugs could be less effective—or at least the simulations that we're doing would have to be updated continually as the virus evolves,” Stevens said.

The research teams are also putting a great deal of effort into drug screening to identify the antiviral medications that can adequately fight the virus, Stevens said. 

By tapping into advanced light sources to observe the tiniest of structures and modeling, computing and simulation capabilities to better understand drug targets, they’re ultimately aiming to help rapidly identify new drugs to fight COVID-19 or others that can be quickly repurposed to stop it. It’s a long, iterative process that uses machine learning and artificial intelligence techniques to screen and identify drugs with components that might be able to combat COVID-19. 

And the Supercomputing Consortium

As these supercomputing efforts and many more were ongoing, the White House in late March also launched the COVID-19 High Performance Computing Consortium, which includes Argonne as one of several labs among its participants. Spearheaded by the White House Office of Science and Technology Policy, Energy Department, and tech-giant IBM, the consortium offers all scientists an access point to tap into 16 of America’s top computing systems.

Stevens said the new consortium is in no way trying to manage the labs’ or other participants' already ongoing work, which he noted has not been disrupted by its founding. Instead, he said the COVID-19 consortium is truly just facilitating access to supercomputers for groups that otherwise wouldn't have access to them, in a collective effort against the novel coronavirus.

“The consortium is really about providing an entry mechanism for other groups [in academia or industry] that might have a need for large scale computing and don't really have any other way to get it. It provides a way for projects to be centrally submitted, and then the labs essentially can decide which ones they want to take on,” Stevens said. “But it may or may not involve actual collaboration—it may be more about providing access to the systems.”

The Coming Upgrade

In 2021, Argonne National Laboratory is set to gain America’s first-ever exascale supercomputer, named Aurora. Initially anticipated to be the world’s fastest supercomputer at the time of its launch, the system will be capable of performing at least one exaflop, which is a quintillion calculations per second. Stevens is leading the lab’s exascale computing initiative.

While the work Argonne is currently doing in response to the worldwide health crisis wouldn’t be possible without the supercomputers researchers are tapping in to, Stevens said the efforts are making them “wish they already had” the exascale machines—“because we just need so much computing and we just don't have enough computing.”

“If we didn't already have an exascale initiative, one would have gotten created from this crisis because we just need orders of magnitude more computing to do this,” Stevens said.

Still, he’s hopeful about the solutions that are potentially being catalyzed by the supercomputing capabilities Argonne and its collaborators are already collectively harnessing against COVID-19. 

“I'm pretty optimistic actually that we'll get through it,” Stevens said. “And maybe in the process, we'll learn some new things about how to use these machines and how to work together. I think that's a really positive side effect.”