‘I think the only limit is our imagination’

The idea that ‘new’ proteins could be created from scratch, with structures and functions not found in nature, was once thought to be impossible. Professor David Baker has spent much of his career proving that it is possible. Now, he says, the pool of amazing proteins available to us is limited not by what evolution has developed over millennia, but by scientists’ imaginations. 

Baker first started studying how proteins form 3D shapes in the 1990s as his research interests shifted from developmental biology to the molecular structures that make cells work. Decades before the protein-structure prediction tool AlphaFold came along, Baker’s group developed software known as Rosetta to try to simulate what shape small amino acid sequences should fold into. They soon realised that running the programme backwards would enable them to find the amino acid sequence for shapes they wanted to create.

By 2003 Baker’s group had succeeded in designing a functional protein that was unlike any other found in nature, and by 2012 he had founded the Institute for Protein Design at the University of Washington in Seattle, US. Then the huge advances in AI and machine learning of the last decade accelerated his vision of being able to accurately design proteins for all manner of medical, environmental and technological challenges. 

[It may] lead to a whole new generation of hybrid materials such as bone or tooth that could function alongside new technologies

Baker’s large research group at the institute, and his vast network of former students and collaborators, are now regularly producing innovative proteins designed for specific applications in medicine, materials science, sustainability and nanotechnology – from enzymes that cleave disease-causing proteins or degrade environmental pollutants to nanopores, nanocages and switches coded in amino acids. 

In 2024 he shared the Nobel Prize in Chemistry with Demis Hassabis and John M Jumper of AlphaFold – the tool that revolutionised the way biologists deduce the structure of proteins and helped further advance Baker’s work. 

Baker spoke to The Biologist editor Tom Ireland about the rapid advances in the field and his passion for designing de novo proteins that change the world for the better. 

Hi David. What would you say is the frontier of protein design at the moment?

In nature, many proteins consume a fuel such as ATP and use it to power processes, like walking along an actin track, or unfolding and refolding the protein to power a reaction or motor.

Something that’s very exciting right now is that we’re on the cusp of being able to design these sorts of nanomachines to do predefined work at the atomic scale. Doing so successfully will be transformational to almost every aspect of daily life.

Another problem we’re close to solving is designing catalyst proteins that do things beyond binding to catalyse chemical reactions. Those are the areas that are really exciting now, as well as applying protein design to real-world problems. We’re not just designing proteins with new capabilities – folks in the lab are identifying real global problems in health, agriculture, and materials science and designing practical solutions for them in real time.

I’ve heard you say that, for a long time, your vision of designing proteins was considered to be on the ‘lunatic fringe’ of bioscience. Was it because the relationship between sequence and function was still so mysterious? Or because people thought you couldn’t improve on nature? 

It was a bit of both. One analogy I like, which is a little bit forced, is that it’s like Lord of the Rings: proteins were seen as these elven things that have been passed down from this golden age in the past and have these magical properties. People saw that all these processes of life – these really beautiful, sophisticated functions – were being carried out by different proteins, and the idea that you could make something like that from scratch, without it arising from the dawn of time and going through evolution, I guess seemed a little bit heretical.

But on the practical side, we also didn’t really understand how these sequences encode the functions, so that made it very hard to do it.

What would you say are the pros and cons of designing proteins over traditional approaches for making use of proteins? Are the old approaches – say, creating antibodies from an animal – essentially obsolete now that you can design proteins?

Where nature has worked for a billion years to optimise something, protein design is probably not going to do better. So protein design is better in cases where there was no evolutionary pressure to optimise for the property that you want.

Is creating sequences that are very different from those that have arisen in nature more likely to cause issues in terms of toxicity, immune response, biocompatibility?

We thought that might be the case, but so far we haven’t really observed it. Let’s take immunogenicity – the immune system is set up to recognise things that it hasn’t seen before. It’s known that even small modifications to a protein can make it immunogenic. If you take a natural protein such as a cytokine, for example, and you add a bunch of mutations, you could actually generate an immune response that is really catastrophic because it also reacts against your own endogenous version of that molecule. Whereas if you make something completely from scratch, it’s possible that you will elicit an immune response, but it won’t cross-react with your own proteins because it’s so different. 

In practice, we’ve had several proteins in humans and the immune responses have not been large. However, like any therapeutic, it’s going to be important to monitor immunogenicity as things proceed through the development pipeline and, in some cases, engineer them to remove epitopes. But, no – it hasn’t been the absolute showstopper you might have thought.

What kind of things are possible in designed proteins that nature has not come across or stumbled upon yet? 

Some of the cool things we’ve just been able to figure out how to do include designing proteases – proteins that cut and destroy natural proteins – except now we can design them from scratch to cut almost any target you want. That opens up many new therapeutic possibilities. Unlike using an antibody drug, where you need one antibody for every target, with a protease you can get by with much smaller amounts because each of the protease molecules can destroy many target molecules.

We can also create proteins that catalyse reactions not seen in nature. We can now make proteins with conditional functions, so they bind to a target or catalyse a chemical reaction when another factor, a gating element, is present.

In terms of what’s coming next, some of the most interesting areas are those where nature never had any interest at all. For example, there are nanopores in nature that allow things in and out of cells, but we’ve now figured out how to make nanopores that will embed themselves in silicon chips. That gives a coupling between biological sensing and solid-state electronics with a silicon chip.

I’m quite optimistic about being able to design increasingly complex systems and I don’t think there’s any real hard upper limit

I mentioned nanomachines, and there the basic concept is that you have a fuel that gets hydrolysed and then the machine does some work on the environment. We can make motors that really do completely new things. I think the only limit is our imagination – and now we’re trying to think about new ways of doing very low-power computing with designed proteins, making 3D arrays of protein switches that are like miniature transistors. 

We’ve been fascinated by things like bone, tooth and shell, where proteins interact with inorganic minerals, and we’ve figured out how to design proteins that will template inorganic mineral deposition. We’ve been looking at compounds that don’t occur in nature such as zinc oxide and designing proteins that will template their deposition, which could lead to a whole new generation of hybrid materials such as bone or tooth that could function alongside new technologies. 

Is it now possible for biologists with no experience of protein design to create new proteins for the applications they need relatively easily? 

Protein design has continued to become more democratised and, yes, people in labs who have never done any protein design are successfully designing proteins. Our software is freely available, as are tools from other developers. So for doing things like designing binding proteins, anyone with access to the internet and computing can, in principle, do it. You still have to be able to test your designs experimentally – and perhaps the hardest thing is to decide what you want to design your protein to do. 

But for the easier applications, if you’re in a lab almost anywhere that has the capability to experimentally test designs and some access to computing resources, you can design proteins against targets that you’re interested in. 

There are proteins being designed in places you would never expect. It’s quite an international enterprise already.

With that level of access to a powerful technology, do you worry about biosafety? 

There are a lot of things nature has come up with that are actually quite a bit nastier than anything that can be designed. For better or for worse, the sequences of the 1918 Spanish flu are available. Things like botulinum toxin are available. I think it’s unlikely people will be making anything more dangerous than those. 

There is a very important safeguard that I’ve been lobbying for. The key step when going from the computer to the wet lab requires making a piece of DNA that encodes the new protein you’ve designed, so I think it’s very important that all DNA synthesis machines record what the sequences are and that they are uploaded to a central database. Say you were trying to make a new ultra-dangerous virus – you’re not going to get it right first time, so you’ll probably be making it, releasing it, seeing what happens and iterating. And the world will want to stop those. So when something new shows up, you want to be able to track where it came from very quickly. But, for the near future, I think the greater danger is what’s already in nature and what’s in public repositories.

If someone were to create a completely new neurotoxin, say, would a DNA synthesis company have the tools to recognise that that is coding for a neurotoxin if it was completely new to nature?

No, something totally new wouldn’t show up in a sequence comparison search, which is why it’s very important to log everything. If that neurotoxin got out, once the sequence could be determined you’d have a record of who made it. That would be a pretty good way to deter people from doing that. 

Are there physical limits on the complexity of a protein that you can design or a group of proteins that you fit together into a kind of nanomachine? Can you just keep getting more complex and multifunctional?

I think it’s a matter of your control and accuracy. As you build systems with more and more parts, it does get more complicated. But think about the ribosome in biology – it has this very long RNA, it’s got all these different parts and proteins. We can’t make things of that complexity now, but we can certainly make things that have three or four components. I think, over time, as the methods get better and better, we can also proceed hierarchically and start with a catalytic core, then build out more and more regulatory elements. I’m quite optimistic about being able to design increasingly complex systems and I don’t think there’s any real hard upper limit.I’ve seen you talk about your lab group being like a constant party. Can you tell me a bit about your laboratory and what you think is important in terms of ethos or culture?

My metaphor is that it’s like a communal brain. The idea is that you have all these really smart, energetic, motivated people who want to make the world a better place, and they’re interacting all the time. The researchers are like neurons and everyone’s really connected. 

If you walk around my lab you see everyone talking all the time. We have different free food events every day, and two group meetings and happy hours each week. It’s a pretty intense environment in terms of interacting with other scientists and sharing ideas. And it’s fun to see someone make a breakthrough.

It’s been a huge collective effort over many years. At the Nobel Prize celebration, there were 185 people who came from my group or who were previous colleagues. Nobel limits the number of people who can come to the ceremony and the banquet, so we rented out a huge ballroom at the Grand Hotel [in Stockholm, Sweden] and they had a huge party while the banquet was going on. The credit often goes to a small number of people, but really the number of people who contribute to this is fantastic. I think it really showed how communal this has been that so many people came. There’s such a concentration of amazing people here that it’s a privilege for me to work with them every day.

How important is it to you that much of the software that helps people do this is open source?

There’s been a lot of development in protein design and industry over the last couple of years, but we’re committed to making everything we develop open source and we release stuff far before papers are published, just because I’m really interested in the whole world benefitting from protein design. 

We do have advantages with having resources here, and I feel like the right thing for us to do is to share everything we have as soon as we have it. We have visitors constantly coming in – I have, I think, 10 master’s students from all over the world coming here [at any one time], and then they leave knowing how to do protein design. So we’re heavily involved not only in developing and releasing code, but in training the next generation of protein designers.

Professor David Baker is the founder and director of the Institute for Protein Design at the University of Washington, US. He is also adjunct professor of genome sciences, bioengineering, chemical engineering, computer science and physics at the University of Washington, and the recipient of numerous awards, including the 2024 Nobel Prize in Chemistry for computational protein design.

Tom Ireland is editor of The Biologist.

This article was first published in the print issue of The Biologist, the RSB's award-winning membership magazine. Subscribe by becoming a member today.