Decoding corona's protein structure

Guessing the movements of the various amino acids and their interactions could enable the creation of drugs that combat the virus

One of the unforeseen side effects of the coronavirus has been the development of what could be described as an “uber-super-computer” through a mode of crowd-sourcing. The Vijay Pande Lab at Stanford University has been running a distributed computing project for many years.

This project, Folding@home, asks for volunteers to donate spare time on their computers. These resources are used to investigate the combinatorially complex mechanics of the ways in which various proteins can be folded and how those proteins work. Protein chains acquire their 3-dimensional structure through folding and this affects their functionalities.

Misfoldings can be the cause of mutations and of many genetic diseases and health conditions. Several types of cancers and other diseases like the Parkinson’s disease are attributed to misfolded proteins. The possible ways in which proteins can be folded run into very large numbers and there are underlying rules which are still not well-understood.

Indeed, there have been games designed around protein-folding to try and discover new rules, by induction of the ingenuity of the gamer community. Alphazero, the famous AI algorithm from DeepMind — which taught itself to play chess, Go, and Shogi — is also working on protein-folding.

The Folding@home uses distributed resources from volunteers who donate time on their machines, to study protein folding problems. The project sets those machines to work through possible variations. As far as the volunteers are concerned, it just means a small download. This allows the Folding@home computations to run, whenever the computer concerned has free resources.

In mid-March, the project announced that it is going to devote some of its resources to studying the structure of the proteins in the Sars-CoV-2 virus, which causes Covid-19 and it was also going to try and hunt for drugs which might fight the coronavirus. That led to a huge burst of enthusiasm for the project with a lot of signups — close to 400,000 new signups happened within the first week.  

The enthusiasm for new signups continued. By last week, the project was capable of running 2.4 ExaFLOPS of combined performance. One exa-flop is 10 to the power of 18 floating point operations per second (a rough description of a “flop” would be one mathematical calculation). That made it by far, the most powerful computer in the world, almost 16 times as fast as the IBM Summit, which the single most powerful supercomputer in existence. It also meant that Folding@Home was performing more calculations per second than the 500 fastest computers in the world combined!

Managing this sort of distributed or parallel processing effort efficiently is tricky. Mathematical problems have to be broken up into chunks that can be worked out effectively by different machines simultaneously and then the pieces have to be stitched back together again.

The Folding@Home website has a covid section which uses an interesting analogy to explain what the project is trying to do. Imagine having a snapshot of an American Football team in its starting lineup and having to guess how the players will move. The Pande lab is attempting to decode the protein structure of the coronavirus and then guess at the movements of the various amino acids and their interactions. The lab specialises in simulations that guess at the movements. This has in the past, helped it develop insights that could enable the creation of drugs that combat the Ebola virus.

The coronavirus has to have one very important movement — the spike or “corona” (which means crown) has to open up and dig into a host human cell to anchor the virus which ten invades the cell. Folding@Home has already produced a possible simulation of how this happens. The project hopes that it will be able to develop new insights into the virus’ movements that help it find “druggable” locations where the virus could be vulnerable to attack by new drugs. The pitch for new volunteers is “Each simulation you run is like buying a lottery ticket. The more tickets we buy, the better our chances of hitting the jackpot”.

Folding@Home uses a distributed computing concept that’s similar to the famous SETI (Search for Extra-Terrestrial Intelligence) project. SETI used a network of volunteer home computers to study radio signals from space in the hopes that it could decipher any sort of intelligent message. SETI was managed out of University of California’s Berkeley Campus and ran from 1999 till March 2020. While it did not find intelligence, it did generate a lot of useful data before it was shelved.

SETI inspired Folding@Home, and a successful distributed computing effort at Climate Prediction and other similar science-oriented efforts where large chunks of data had to be handled. It also inspired the creation of bitcoin and other cryptocurrencies which use similar distributed “mining” efforts to generate coins.