The Royal Swedish Academy of Sciences honoured University of Washington professor David Baker and two scientists from Google DeepMind, CEO Demis Hassabis and senior research scientist John Jumper.
Hassabis and Jumper were recognized for winning a decades-long race to use computers to predict a protein’s structure based on only on its genetic code. Baker’s prize nods to his use of computers to invent never-before-seen proteins.
In 2020, DeepMind offered its first big advance, showing its AI algorithm AlphaFold could correctly ‘solve’ the structures of proteins that scientists had already worked out in the lab.
Just a couple of years later, it offered a more stunning feat with the release of snapshots of nearly every protein in existence—all 200 million building blocks for humans, animals, plants, fungi, bacteria and more.
Consider that not so long ago, capturing an image of a single protein was a very painstaking process that sometimes didn’t even work. So AlphaFold may be new, but it’s already changing the industry.
For the last few years, whenever I visit pharma companies to learn about their latest science, it’s become my habit to ask: And what do you think about AlphaFold? Are you using it?
The answer, always, is ‘yes.’ In 2022, Jay Bradner, then head of research at Novartis (now chief scientific officer at Amgen) told me, “I’m on it more than Spotify.”
Why does protein structure matter so much? Proteins are the targets of most medicines. At times, proteins themselves can be a medicine. Chemists want to be able to ‘see’ them in intricate detail to design drugs that can wedge into just the right spot—a hidden pocket or a sticky bit — to turn them off, on, or even tune their activity to address a disease.
Yet, if you flip through a biochemistry textbook, you’ll see the challenge: They are amazingly varied in structure, from blobs to squiggles to pinwheels.
No, we aren’t in a world where someone presses a button on a computer and it spits out the blueprint for a novel drug—and I doubt that we’ll ever get there.
Nevertheless, biotech and pharma companies have quietly integrated the protein prediction technology into their daily work in ways that trim time from the difficult, slow process of inventing new medicines.
And DeepMind scientists continue to make the technology even more useful. This spring, the company unveiled a version of the technology that can predict the interactions between proteins and other key players in the cell, whether that’s DNA, RNA, small molecules or other proteins.
And last year, scientists analysed the structures within AlphaFold to figure out which changes in a protein are harmful and which are benign, a tool that can help researchers much more easily pinpoint the cause of rare genetic diseases.
Many more advances are needed. Not all the structures in AlphaFold all perfect— some far from it—and so further refinement is in order. And people in the field of AI drug design would like to be able to create a drug that not only locks into its target, but is also safe and has the kind of properties that make it a viable commercial product (for example, one that sticks around in the body long enough to do its job and can be packed into a pill for easy consumption). All of that is still a work in progress.
The other winner of the Nobel Prize is University of Washington’s David Baker, who has spent his career trying to do something quite extraordinary: designing proteins not found in nature.
That first required understanding how proteins fold (as the DeepMind work underscores, no easy task), and then tinkering with the genetic sequences to come up with new structures.
While the applications of this approach span many disciplines, in medicine, that could mean anything from tiny tweaks in just the right spot to address an existing drug’s shortcomings, or dreaming up something never before seen.
How promising is this approach? Last spring, investors promised to sink more than $1 billion into Xaira Therapeutics, a biotech firm whose foundational technology came from Baker’s lab. That puts the company into the upper echelons of not only AI-focused companies, but all biotech startups.
What’s incredible about the work done across both Baker’s lab and DeepMind is the pace of progress. We’re getting closer to a place where the drug-discovery process is more efficient and successful. That’s something to celebrate. ©bloomberg