Using AI to design a universal coronavirus vaccine

HostIt feels like we have been stuck in a game of catch up for years now. Every time a new version of a virus shows up, we have to tweak our shots and hope they still work against whatever comes next.

HostI have been reading about a group of researchers who want to stop that cycle by making one single shot that works against every coronavirus out there, even the ones we haven't seen yet. How can a computer design something that covers so much ground at once?

GuestIt's a huge shift in how we think about medicine. Usually, when we make a vaccine, we're looking at the past. We take a piece of a virus that's already making people sick and we show it to the body so the body knows what to look for. The problem is that coronaviruses are like a massive family with hundreds of cousins. Some cause the common cold, some caused the first SARS scare years ago, and others caused the big one we just went through. They're all slightly different, so a shot for one doesn't always work for the others.

GuestTo get around this, scientists used AI to build what they call a mosaic. Think of a tiny soccer ball. Instead of having just one type of spike sticking out of it, this ball has eight or more different types of spikes from all across the virus family tree. The AI's job was to figure out how to glue all those different spikes onto that one ball so they stay stable and show the body the right shapes.

HostWait, so this isn't just a better version of the shots we already have?

GuestNot at all. It's meant to be a shield against the whole family. The goal is to teach your body to find the parts that all these viruses have in common. If you only see one spike, your body might focus on a part of the virus that changes easily. But when the AI puts all those different spikes in front of your defenses at once, your body is forced to look for the bits that are the same on every single one of them. We call those the conserved parts. They're like the anchor of the virus. If the virus changes that anchor, it can't function anymore. So, by hitting that one spot, you can potentially stop viruses that haven't even jumped from animals to humans yet.

HostThat sounds like a lot of guesswork for a computer. I mean, if these viruses haven't even shown up in humans, how does the AI know what they're going to look like?

GuestIt's less about guesswork and more about math. Scientists fed the AI the blueprints for hundreds of different coronaviruses found in bats and other animals. The AI then looks at how the proteins in those viruses fold. Proteins are the building blocks of the virus, and they fold into very specific, messy shapes. For a human to figure out how to design a new protein that can hold eight different spikes while staying perfectly stable, it would take years of trial and error in a lab. The AI can run through millions of those shapes in a few hours. It finds the one perfect design that can act as a base for all those different spikes.

HostBut we have had vaccines for a long time without needing a computer to do the heavy lifting. Why is the AI part so vital for this specific problem?

GuestBecause the shapes we're talking about are too complex for us to just eye ball. When you try to stick eight different spikes onto one tiny ball, they usually just fall apart or get tangled. It's a bit like trying to build a bridge out of pieces that were never meant to fit together. The AI acts like a super fast architect. It designs new, synthetic proteins that don't exist in nature. These proteins are built specifically to be the glue. Without the AI, we would be stuck using whatever shapes nature gave us, which just aren't strong enough or flexible enough to hold all those different virus parts at once.

HostSo if this works, we wouldn't need to keep getting new boosters every time a new version of a virus pops up?

GuestThat's the hope. Instead of reacting to a new version of the virus after it starts spreading, we would already have the protection in our systems. In the lab, they have already seen this work in animals. They gave them this mosaic shot, and the animals were protected not just against the versions of the virus we know, but also against older ones like the first SARS from two decades ago. It showed that the body really did learn to find that common anchor point.

HostIt's a bit scary to think about how many of these viruses are still out there in the wild just waiting to make the jump.

GuestIt's a constant threat. There are thousands of these viruses in bats and birds. Most of them will never bother us, but it only takes one. The old way of making vaccines was like waiting for a fire to start and then trying to invent a new kind of water to put it out. This new way is more like building a house out of stuff that simply can't burn in the first place. We're moving from being reactive to being proactive.

HostIs there a limit to what the AI can pack onto that little soccer ball? I mean, could you just keep adding more and more spikes until it covers every germ we know?

GuestThere's a physical limit. If you crowd the ball too much, the body’s defenses might get confused and not learn anything well. That's why the AI is so important. It picks the most diverse group of spikes possible. It's like picking eight people from all over the world to represent everyone, rather than just picking eight people from the same town. The AI finds the most spread out points on the virus family tree to make sure the coverage is as wide as it can get.

GuestThe real test is happening right now in labs where they're seeing if this can stop viruses that haven't even moved into humans yet.

HostThe soccer ball approach really changes things, especially if we can have that shield ready before the next family member even shows up at the door.