Anyone with an Internet connection and a computer can share their unused computing power to help in the fight against the coronavirus pandemic through Folding@Home.
Folding@Home is a distributed computing project that relies on the CPUs and GPUs of hundreds of thousands of home computers (when they are not in use) to perform the complex calculations that simulate protein dynamics. The project uses the idle processing resources of these personal computers owned by volunteers who have installed the software on their systems. Free clients for Linux, Mac, and Windows are available for download to connect PCs to Folding@Home's crowdsourced network.
Folding refers to the way proteins fold, in the cells that make up the human body. There are tens of thousands of proteins within a typical human cell. A protein starts off as a chain of amino acids that assemble, or fold, in different configurations in order to perform specific tasks, such as building muscles, binding molecules with a cell, or acting as a catalyst to help cells grow. Proteins are the workhorses of human cells, keeping our bodies running properly and maintaining our health. If proteins do not fold properly, things go wrong.
According to Greg Bowman, director of Folding@Home and associate professor of Biochemistry and Molecular Biophysics at the Washington University School of Medicine, protein folding is how the linear chain of chemicals that make up a protein spontaneously fold up into a reasonably specific three-dimensional structure. Bowman suggests protein folding is akin to disassembling all the parts of a car, laying them out on a road, and having them suddenly spring together on their own to form a functioning automobile.
The complex computing models of the coronavirus created by Folding@Home will contribute to efforts to find a cure. The data generated will be quickly and openly disseminated as part of an open science collaboration of multiple laboratories around the world, giving researchers new tools to fight Covid-19.
The Folding@Home project was started at Stanford University in 2000 by Vijay Pande, a Stanford professor teaching chemistry, structural biology, and computer science, who also served as director of Stanford's program in biophysics. Pande's research centered on cloud computing simulation techniques that address problems in chemical biology; he pioneered the distributed computing methodology that resulted in the launch of Folding@Home. Pande's intent was to use distributed computing models for disease research by applying computer science techniques, such as distributed systems and machine learning, to biology and medicine, for research purposes and for the development of new therapeutics.
Pande, now a general partner at the Andreessen Horowitz venture capital firm in Silicon Valley, led Folding@Home until 2019, when he turned its stewardship over to Bowman, who studied under Pande at Stanford. The project is now run out of Bowman's lab at Washington University's School of Medicine, with several other labs heavily involved, including the Chodera lab at the Sloan-Kettering Institute, and the Vincent Voelz lab in the chemistry department of Temple University.
"The original idea behind the project was to study protein dynamics," says Bowman. He adds the emphasis of Folding@Home has shifted over time, to understanding how the motions of the different components of a protein contribute to its function (or its malfunction), and then providing insight into things like drug design and news therapeutics.
In their book The Life Science: Current Ideas of Biology, biologists Jean and Peter Medawar describe a virus as "simply a piece of bad news wrapped up in protein," and the coronavirus has proved to be the worst news possible.
"We had just published a paper where we uncovered a new way to potentially combat Ebola virus infections," Bowman says, "so when Covid-19 came on our radar, we thought we should try and do the same thing."
Currently, Folding@Home is simulating the dynamics of Covid-19 proteins to hunt for new therapeutic opportunities. With so many distributed computers working towards the same goal, the aim is to help develop a therapeutic remedy as quickly as possible, Bowman says.
"To help tackle coronavirus, we want to understand how these viral proteins work and how we can design therapeutics to stop them," Bowman says. Folding@Home's specialty is using computer simulations to understand proteins' moving parts; watching how the atoms in a protein move relative to one another provides valuable information inaccessible by any other means, according to Bowman.
Folding@Home's massive computing capacity is crucial to understanding protein folding, but rather than a single gargantuan machine, Folding@Home is a distributed computing network. Hundreds of thousands of volunteers all over the world have downloaded Folding@Home's software onto their computing devices and contributed their idle computing power, putting Folding@Home's distributed network on a par with the largest supercomputers in the world.
Distributed networks of personal computers can be quite powerful, says Jack Dongarra, American University Distinguished Professor of Computer Science in the Electrical Engineering and Computer Science Department at the University of Tennessee, who compiles the Top500 ranking of the fastest supercomputers in the world.
Regarding Folding@Home, Dongarra said, "You divide the problem into smaller problems, work on them on individual PCs, and when you put it all back together you can solve the tasks." In essence, problems are split up into many independent chunks which can be sent to hundreds of PCs at the same time. Calculations are run in parallel, and as each small job is completed, they are pooled together to create a larger framework. These are shared with participating scientists, exponentially speeding up the time it takes to perform complex calculations.
"Supercomputers differ in how they put it together," Dongarra says, explaining that Summit, the fastest supercomputer in the world, situated at Oak Ridge National Laboratory (ORNL), has 2.3 million processors which can be quickly coordinated to solve a task. "A virtual computer can't pass information quickly between processors like in a supercomputer, and that's the difference between a supercomputer and a PC."
The fight against Covid-19 has led to a rapid acceleration in volunteers signing up, providing a phenomenal boost in computing power for Folding@Home. Bowman says volunteers jumped from about 30,000 in February to over 400,000 by mid-March.
"There are over one million devices running Folding@Home right now, with about 600,000 CPUs and about 400,000 GPUs that are running our software," he says.
The additional computing capacity has allowed Folding@Home to surpass an exaflop of distributed computational performance, making it the first to attain exascale computing. That's more than five times the theoretical peak operations per second of ORNL's Summit, which is capable of 200 petaFLOPS.
This rapid influx of capacity has also created operational challenges. So many people have joined to support Folding@Home's work on Covid-19 that its servers are struggling to keep up. Bowman says they have been scaling up their server-side infrastructure as quickly as they can, but a few of Folding@Home's users have experienced some idle time.
However, "Anyone encountering idle time should not come to the erroneous conclusion that we don't need volunteers," Bowman says. "There is still plenty of science to be done."
John Delaney is a freelance writer based in New York City, NY, USA.
No entries found