What tools can we deploy in the fight against Covid-19? Taken together, the world has enough sheer compute power to attack this complex problem aggressively and quickly.
As I write this, the U.K. and the rest of Europe are experiencing more or less strict lockdown conditions as part of the attempt to slow the spread of the novel coronavirus, Covid-19. This extraordinarily dangerous and contagious virus is still spreading fast throughout communities all over Europe.
What tools can we deploy in the fight against this deadly enemy? We are all practicing social distancing to slow the spread. A community of makers has mobilized to 3D-print personal protection equipment for health-care workers, and British factories have been modified to produce emergency ventilators. Advanced drug and vaccine research is under way in almost every applicable research institution around the world.
And we have the sheer, brute force of a world of powerful computers to help us tackle the search for a solution — which is effectively a computing problem — head on.
In the U.S., the White House has launched the Covid-19 High Performance Computing Consortium, whose members are volunteering compute time for researchers working on problems related to the pandemic. This includes research on epidemiology, antiviral drug and vaccine development, modeling how the virus spreads, and even studying which social distancing measures work best. Researchers submit their proposals online and are assigned compute time on any one of the participating U.S. supercomputers.
That machine might be the world’s biggest supercomputer, Summit. Housed at Oak Ridge National Laboratory, Summit is capable of 200 petaFLOPS (that’s 200 quadrillion, or 200,000 trillion, floating-point operations per second) and comprises 4,608 nodes, each with two IBM POWER9 processors and six Nvidia Volta GPUs. Or it could be one of NASA’s supercomputers, the biggest of which can handle 8.32 PFLOPS, or a commercial data center that has donated runtime (these include Google Cloud, Microsoft Azure, and Amazon Web Services [AWS]). The U.S. project enabled creation of the first 3D, atomic-scale map of the spike protein — the part of the virus that attaches to human cells. The map can be used in the search for drugs that bind to the spike, rendering it ineffective.
Even those without a hyperscale data center or a supercomputer at their disposal have been helping the effort against the virus by joining a project called folding@home, which has already been used to identify a druggable site on the Ebola virus.
This global project uses home computers to run simulations of how virus proteins move. Such information is absolutely critical in the search for effective drugs; while there are many experimental methods available for determining a protein’s structure, the molecules have lots of moving parts, and their behavior is characterized in part by this movement. However, as you’d imagine, simulating this kind of movement is very compute-intensive because there are so many permutations.
That’s where consumer devices come in. Taken together, hundreds of thousands of home computers can effectively be used as a supercomputer by breaking the required calculations into tiny chunks and sending them across the internet for idle laptops and gaming PCs to compute. As of this writing, the project’s 700,000+ computers’ combined power have surpassed an exaFLOPS (1,000 PFLOPS), making them collectively more powerful than the next top 100 supercomputers combined.
Taken together, the world has enough sheer compute power to attack this complex problem aggressively and quickly. The scale of the challenge requires the collaboration of governments, industry, scientific and medical researchers, and members of the public. And their efforts are paying off.
At press time, work enabled by the Covid-19 High Performance Computing Consortium had already resulted in the discovery of 77 drug compounds that may be effective in treating Covid-19 patients, and the project folding@home had captured the opening of one of the three receptor-binding domains of the Covid-19 spike, the first part of the mechanism by which it binds to human cells.
Unless a treatment or vaccine is developed soon, this shadow will hang over us for years to come. While we can’t be sure that any of these projects will lead to treatments or vaccines, we should remain dedicated to fighting the coronavirus with every weapon we have.