AlphaFold, DeepMind’s protein structure program, is impressive because it reveals so much fundamental information about living organisms.
Proteins are the building blocks of life, after all, and as such they are essential for life and drug development. Proteins can be drug targets, and they can themselves be drugs. In either case, it is important to know the intricate ways in which they bend into different shapes. Their coils, floppy bits, hidden pockets and sticky patches can control, for example, when a signal is sent between cells or if a process is turned on or off.
Until now, capturing an image of a protein has required painstaking work that has taken anywhere from days to months or even years – work that sometimes never came to fruition.
Since the early 1990s, scientists have been trying to train computers to predict the structure of a protein based on its genetic sequence. AlphaFold got its first taste of success in 2020, when it correctly predicted the structures of a handful of proteins. The following year, DeepMind uploaded around 365,000 proteins to its server.
Now the whole universe of proteins is at stake – in animals, plants, bacteria, fungi and other living beings. All 200 million of them.
Just as the gene-editing tool Crispr revolutionized the study of human diseases and the design of drugs to target genetic errors, AlphaFold’s exploit fundamentally changes the way new drugs can be invented.
“Anyone who might have thought that machine learning was not yet relevant to drug research must surely feel different,” said Jay Bradner, president of the Novartis Institutes for Biomedical Research, the pharmaceutical company’s research arm. . “I’m more into it than Spotify.”
Count me as one of the old skeptics. I hadn’t discounted the possibility that AI could have an impact on the pharmaceutical industry, but I was tired of the many biotech companies touting often ill-defined machine learning capabilities. Companies have often claimed that they can use AI to invent a new drug without acknowledging that the starting point – a protein structure – still needs to be engineered by a human. And until now, people had to invent drugs first for the computer to improve them.
Producing the full protein compendium is something completely different – and outside of the usual hype cycle. It’s no wonder biotech and pharmaceutical executives are widely embracing the AlphaFold revelations.
Rosana Kapeller, CEO of Rome Therapeutics, offers an example from her company’s labs. Rome is probing the “dark genome,” the repetitive part of the human genetic code believed to be largely a relic of ancient viruses. The Rome team spent more than six months refining their first image of a protein embedded in this dark genome. Just a day after capturing an initial snapshot of a second protein, DeepMind dropped its full image load. Within 24 hours, the scientists in Rome had perfected their image. “So you see,” she said, “it’s amazing.”
None of this is to say that AlphaFold will solve all drug discovery problems, or even that its 200 million protein images are perfect. They are not. Some need more work, and others are more like the scribbles of a child than the fleshed-out pictures researchers are hoping for.
Scientists in industry and university labs tell me that even when snapshots are imperfect, they contain enough information to give a rough idea of where important things are. David Liu, a professor at the Broad Institute of MIT and Harvard and founder of several biotech companies, said the technology still allows researchers in his lab to “achieve that Zen-like state of understanding” when deciding where to tinker with a protein. to change its Properties.
But proteins don’t exist as fixed snapshots either. Depending on what work they are doing at any given time, they yawn, jiggle, and twist inside the swamp of a cell. In other words, AlphaFold gives us the Instagram protein; scientists would like to have the TikTok protein or, possibly, the YouTube protein.
Even if it becomes possible one day, it is only one step in the process of creating new drugs. The most expensive part is testing this new drug in humans.
Still, AlphaFold’s images can help drugmakers get to testing faster. The DeepMind feat may have taken years of scientific exploration, but it produced something with truly monumental consequences. And it made this work freely available. (Of course, he also started his own biotech company, Isomorphic Labs, to try to capitalize on the advances.)
Finally, we get a glimpse of the potential of AI to transform the pharmaceutical industry. And now it is possible to envision what problems machine learning might solve next for science and medicine.
More other Bloomberg Opinion writers:
• The next big thing in AI is fake data: Parmy Olson
• Maybe AI technology isn’t as scary as we thought: Tyler Cowen
• AI needs a babysitter, just like the rest of us: Parmy Olson
This column does not necessarily reflect the opinion of the Editorial Board or of Bloomberg LP and its owners.
Lisa Jarvis is a Bloomberg Opinion columnist covering biotech, healthcare, and pharmaceuticals. Previously, she was the editor of Chemical & Engineering News.
More stories like this are available at bloomberg.com/opinion