When scientists first sequenced the human genome in 2003, they revealed the full set of DNA instructions that make a person. But we still didn’t know what all those 3 billion genetic letters actually do. 

Now Google’s DeepMind division says it’s made a leap in trying to understand the code with AlphaGenome, an AI model that predicts what effects small changes in DNA will have on an array of molecular processes, such as whether a gene’s activity will go up or down. It’s just the sort of question biologists regularly assess in lab experiments.

“We have, for the first time, created a single model that unifies many different challenges that come with understanding the genome,” says Pushmeet Kohli, a vice president for research at DeepMind.

Five years ago, the Google AI division released AlphaFold, a technology for predicting the 3D shape of proteins. That work was honored with a Nobel Prize last year and spawned a drug-discovery spinout, Isomorphic Labs, and a boom of companies that hope AI will be able to propose new drugs.

AlphaGenome is an attempt to further smooth biologists’ work by answering basic questions about how changing DNA letters alters gene activity and, eventually, how genetic mutations affect our health. 

“We have these 3 billion letters of DNA that make up a human genome, but every person is slightly different, and we don’t fully understand what those differences do,” says Caleb Lareau, a computational biologist at Memorial Sloan Kettering Cancer Center who has had early access to AlphaGenome. “This is the most powerful tool to date to model that.”

Google says AlphaGenome will be free for noncommercial users and plans to release full details of the model in the future. According to Kohli, the company is exploring ways to “enable use of this model by commercial entities” such as biotech companies. 

Lareau says AlphaGenome will allow certain types of experiments now done in the lab to be carried out virtually, on a computer. For instance, studies of people who’ve donated their DNA for research often turn up thousands of genetic differences, each slightly raising or lowering the chance a person gets a disease such as Alzheimer’s.

Lareau says DeepMind’s software could be used to quickly make predictions about how each of those variants works at a molecular level, something that would otherwise require time-consuming lab experiments. “You’ll get this list of gene variants, but then I want to understand which of those are actually doing something, and where can I intervene,” he says. “This system pushes us closer to a good first guess about what any variant will be doing when we observe it in a human.”

Don’t expect AlphaGenome to predict very much about individual people, however. It offers clues to nitty-gritty molecular details of gene activity, not 23andMe-type revelations of a person’s traits or ancestry. 

“We haven’t designed or validated AlphaGenome for personal genome prediction, a known challenge for AI models,” Google said in a statement.

Underlying the AI system is the so-called transformer architecture invented at Google that also powers large language models like GPT-4. This one was trained on troves of experimental data produced by public scientific projects.

Lareau says the system will not broadly change how his lab works day to day but could permit new types of research. For instance, sometimes doctors encounter patients with ultra-rare cancers, bristling with unfamiliar mutations. AlphaGenome could suggest which of those mutations are really causing the root problem, possibly pointing to a treatment.

“A hallmark of cancer is that specific mutations in DNA make the wrong genes express in the wrong context,” says Julien Gagneur, a professor of computational medicine at the Technical University of Munich. “This type of tool is instrumental in narrowing down which ones mess up proper gene expression.” 

The same approach could apply to patients with rare genetic disease, many of whom never learn the source of their condition, even if their DNA has been decoded. “We can obtain their genomes, but we are clueless as to which genetic alterations cause the disease,” says Gagneur. He thinks AlphaGenome could give medical scientists a new way to diagnose such cases. 

Eventually, some researchers aspire to use AI to design entire genomes from the ground up and create new life forms. Others think the models will be used to create a fully virtual laboratory for drug studies. “My dream would be to simulate a virtual cell,” Demis Hassabis, CEO of Google DeepMind, said this year. 

Kohli calls AlphaGenome a “milestone” on the road to that kind of system. “AlphaGenome may not model the whole cell in its entirety … but it’s starting to sort of shed light on the broader semantics of DNA,” he says.

When you think of AI’s contributions to science, you probably think of AlphaFold, the Google DeepMind protein-folding program that earned its creator a Nobel Prize last year.

Now OpenAI says it’s getting into the science game too—with a model for engineering proteins.

The company says it has developed a language model that dreams up proteins capable of turning regular cells into stem cells—and that it has handily beat humans at the task.

The work represents OpenAI’s first model focused on biological data and its first public claim that its models can deliver unexpected scientific results. As such, it is a step toward determining whether or not AI can make true discoveries, which some argue is a major test on the pathway to “artificial general intelligence.”

Last week, OpenAI CEO Sam Altman said he was “confident” his company knows how to build an AGI, adding that “superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own.” 

The protein engineering project started a year ago when Retro Biosciences, a longevity research company based in San Francisco, approached OpenAI about working together.

That link-up did not happen by chance. Sam Altman, the CEO of OpenAI, personally funded Retro with $180 million, as MIT Technology Review first reported in 2023.

Retro has the goal of extending the normal human lifespan by 10 years. For that, it studies what are called Yamanaka factors. Those are a set of proteins that, when added to a human skin cell, will cause it to morph into a young-seeming stem cell, a type that can produce any other tissue in the body. 

It’s a phenomenon that researchers at Retro, and at richly funded companies like Altos Labs, see as the possible starting point for rejuvenating animals, building human organs, or providing supplies of replacement cells.

But such cell “reprogramming” is not very efficient. It takes several weeks, and less than 1% of cells treated in a lab dish will complete the rejuvenation journey.

OpenAI’s new model, called GPT-4b micro, was trained to suggest ways to re-engineer the protein factors to increase their function. According to OpenAI, researchers used the model’s suggestions to change two of the Yamanaka factors to be more than 50 times as effective—at least according to some preliminary measures. 

“Just across the board, the proteins seem better than what the scientists were able to produce by themselves,” says John Hallman, an OpenAI researcher.

Hallman and OpenAI’s Aaron Jaech, as well as Rico Meinl from Retro, were the model’s lead developers.

Outside scientists won’t be able to tell if the results are real until they’re published, something the companies say they are planning. Nor is the model available for wider use—it’s still a bespoke demonstration, not an official product launch.

“This project is meant to show that we’re serious about contributing to science,” says Jaech. “But whether those capabilities will come out to the world as a separate model or whether they’ll be rolled into our mainline reasoning models—that’s still to be determined.”

The model does not work the same way as Google’s AlphaFold, which predicts what shape proteins will take. Since the Yamanaka factors are unusually floppy and unstructured proteins, OpenAI said, they called for a different approach, which its large language models were suited to.

The model was trained on examples of protein sequences from many species, as well as information on which proteins tend to interact with one another. While that’s a lot of data, it’s just a fraction of what OpenAI’s flagship chatbots were trained on, making GPT-4b an example of a “small language model” that works with a focused data set.

Once Retro scientists were given the model, they tried to steer it to suggest possible redesigns of the Yamanaka proteins. The prompting tactic used is similar to the “few-shot” method, in which a user queries a chatbot by providing a series of examples with answers, followed by an example for the bot to respond to.

Although genetic engineers have ways to direct evolution of molecules in the lab, they can usually test only so many possibilities. And even a protein of typical length can be changed in nearly infinite ways (since they’re built from hundreds of amino acids, and each acid comes in 20 possible varieties).

OpenAI’s model, however, often spits out suggestions in which a third of the amino acids in the proteins were changed.

“We threw this model into the lab immediately and we got real-world results,” says Retro’s CEO, Joe Betts-Lacroix. He says the model’s ideas were unusually good, leading to improvements over the original Yamanaka factors in a substantial fraction of cases.

Vadim Gladyshev, a Harvard University aging researcher who consults with Retro, says better ways of making stem cells are needed. “For us, it would be extremely useful. [Skin cells] are easy to reprogram, but other cells are not,” he says. “And to do it in a new species—it’s often extremely different, and you don’t get anything.” 

How exactly the GPT-4b arrives at its guesses is still not clear—as is often the case with AI models. “It’s like when AlphaGo crushed the best human at Go, but it took a long time to find out why,” says Betts-Lacroix. “We are still figuring out what it does, and we think the way we apply this is only scratching the surface.”

OpenAI says no money changed hands in the collaboration. But because the work could benefit Retro—whose biggest investor is Altman—the announcement may add to questions swirling around the OpenAI CEO’s side projects.

Last year, the Wall Street Journal said Altman’s wide-ranging investments in private tech startups amount to an “opaque investment empire” that is “creating a mounting list of potential conflicts,” since some of these companies also do business with OpenAI.

In Retro’s case, simply being associated with Altman, OpenAI, and the race toward AGI could boost its profile and increase its ability to hire staff and raise funds. Betts-Lacroix did not answer questions about whether the early-stage company is currently in fundraising mode. 

OpenAI says Altman was not directly involved in the work and that it never makes decisions based on Altman’s other investments.