How Folding@home took time off from curing cancer to tackle a pandemic

BY IAN EVENDEN

Coronavirus doesn’t stand a chance against the might of Folding@home.

© GETTY IMAGES/KATERYNA KON/SCIENCE PHOTO LIBRARY

WHEN THE Maximum PC Illustrated Bumper Book of World History is written, the years 2020 and 2021 will get their own chapter. Not because of politics but because of the global pandemic and shortage of microchips.

Folding@home will probably need a lot of coverage, too—a citizen science project contributing to biochemical research projects that is free to add to and feels extremely cool while you do it (except for the heat it pumps out of your PC case).

There’s not much we can do about the chip shortage, but the pandemic and Folding@home collided beautifully. It led to Folding@home becoming the fastest supercomputer on the planet, hitting one exaflop (a 1 followed by 18 zeros) on March 25, 2020, then peaking at 2.4 exaflops the next month. That’s as powerful as the top 500 (known) supercomputers combined. And that much power folds a lot of proteins.

But why would we want to fold proteins in the first place? Look up any image of one online, and it seems pretty well folded already—all spirals and ribbons, like your little sister’s hair on her birthday. The record was there for the taking, however, so here’s how Folding did it.

FOLDING@HOME is used to breaking records. “Back in 2010, 2012, we had a client on the PlayStation 3,” says Anton Thynell, head of collaboration and communication at Folding. “That’s where we took our first Guinness world record for the world’s strongest distributed computing network, with I think 800,000 PS3s and a couple of hundred thousand PCs. Folding@home has a very strong community, and it’s well known in tech. I think what resonated most, though, is that you can donate your computing power to research. We’re not asking for funding, we’re asking for computing power, and that power is directly connected with the simulations researchers are running.”

Since its launch back in 2000, over 200 scientific papers have been produced as a direct result of Folding@home processing, including one in June this year, published in Nature Chemistry, that announced the discovery of “cryptic pockets” on the surface of the virus that offer targeting options for new antiviral drugs, as well as epitopes—protein fragments that are capable of stimulating an immune response. This is a clear-cut breakthrough, and we dread to think how long it would have taken had Folding not been able to scale up so effectively. The data and models are being made freely publicly available.

Anton Sinitsky of the Pande Lab, Stanford, works on Folding data.

Protein folding is a complicated thing to simulate—the exascale computer Folding created for its COVID paper managed to simulate just 0.1 second in the life of the virus’s proteome (all the proteins that are, or can be, expressed by a genome)— yet the benefits from doing so are huge. Protein misfolding is a causal factor in a number of diseases, including some cancers, allergies, Alzheimer’s disease, cystic fibrosis, sickle-cell anemia, Creutzfeldt-Jakob disease, and type 2 diabetes. Then there’s the way viruses, including HIV, influenza, and coronavirus, enter our cells—they have a protein on their surface that folds to be like a key, and this attaches to a lock on the surface of the cell, opening it up for infection if the key is just the right shape.