dailynewsnblog
Anthropic’s Claude made headlines last week by discovering an enzyme system that humans had never seen before. That doesn’t happen every day, and Anthropic is quite proud of the work its AI did in finding it. The question is, is Anthropic’s discovery all hype, or could this really be the next big thing in biological research?
The system in question is a CRISPR-like enzyme that was discovered while Claude was scouring through about 200,000 reverse transcriptase samples. This same level of research would’ve taken humans months to conduct, and so Claude is being heralded for doing it a lot faster and actually finding a result worth studying.
Anthropic says human involvement was limited to the prompts and lab work while 950 AI agents scoured through publicly available databases to identify something unique. It took about 21 hours to spot the enzyme, and then it was confirmed by human scientists to be a novel enzyme, meaning that it had never been observed by science before.
“This is an exciting example of how AI agents can contribute to biological discovery,” Feng Zhang, professor at MIT and the Broad Institute, said in a statement. “The identification of RNA-repeat arrays associated with reverse transcriptases is genuinely intriguing and merits further investigation.”
What is CRISPR?
There are a lot of big words here, so let’s pause to decipher some of them. CRISPR (clustered regularly interspaced short palindromic repeats) is a defense system found in bacteria. This defense system is used to fight off viral infections. That’s right, bacteria can get sick too.
The way it works is that when a bacterium survives a viral infection, it copies a snippet of that virus’s DNA into its own DNA, and that library of snippets builds up over generations. When a virus attacks the bacterium, the bacterium summons an RNA with the appropriate viral snippet. The RNA uses that snippet like a paint-by-numbers to guide a cutting protein that cuts the viral DNA in just the right places so that it can’t replicate, thereby killing the infection. CRISPR is this entire process from start to finish.
Researchers discovered CRISPR in 1987 and eventually figured out that they could use that system in gene therapies, thereby removing harmful or mutated DNA in living organisms using a similar method as CRISPR and replacing it with clean, nonmutated DNA. It took darn near 40 years, but the first CRISPR medicine was approved in 2023 in the UK, and then in 2024 by the US Food and Drug Administration.
And what is a reverse transcriptase?
The other glossary term you need to know is reverse transcriptase. This is an enzyme in some organisms that uses RNA as a template to build DNA. This is the reverse of how things normally go, where DNA is used as a template to build RNA, and so it has its own name and classification.
It’s a defining characteristic of many viruses, including HIV. Learning that HIV uses a reverse transcriptase enzyme to replicate helped scientists devise some of the first medicines to treat it, and the information is still used today in modern therapies that have contributed to HIV no longer being a death sentence.
However, unlike CRISPR, reverse transcriptases are in all kinds of other organisms aside from viruses.
So what exactly did Anthropic find?
One of the things about CRISPR is that it contains many repeating DNA fragments, a primary indicator that helped scientists identify it in the first place. In the decades since, scientists have only ever found CRISPR in bacteria and archaea, single-celled organisms separate from bacteria and other microbes.
This is notable because this is basically what Anthropic found with its army of Claude bots.
“What Claude noticed is another type of protein, a reverse transcriptase, sitting next to a long array of repeating DNA, which is very similar to how CRISPR is laid out,” a representative for Profluent, an AI lab that made OpenCRISPR-1, the first AI-generated gene editor, told CNET in an email. “This particular reverse transcriptase had been seen before. The array next to it and a partner protein alongside it are what Anthropic is sharing.”
The enzyme was found in a bacteriophage, which is a virus that specifically attacks bacteria. Viruses with large DNA genomes like this are commonly known as jumbo phages. Anthropic named this previously uncharacterized system “array-associated reverse transcriptases,” or ART for short.
How big of a deal is this discovery?
This is the real conversation. CRISPR is one of the biggest scientific discoveries in modern history, and its ability to rewrite DNA has a ton of potentially viable uses in medical science. So, finding something like CRISPR, as Anthropic did, is definitely going to warrant some discussion.
The problem is twofold. One, it took scientists about 20 years to figure out how CRISPR even worked, and then nearly another two decades to build the first medicine with it. That means it could take a while before we find out if this discovery is important or not.
“Scientists are finding interesting sequence patterns all the time, relatively speaking,” Bonnie Berger, head of the Computation and Biology group at MIT’s Computer Science and AI Lab, told CNET in an email. “The bottleneck is then figuring out the purpose and function of these systems using deeper analysis, expert biological knowledge, and complementary targeted experiments.”
Anthropic has no illusions when it comes to this side of things.
“Our work to understand the primary function of ARTs is ongoing,” Anthropic said in a statement. “However, we think it is important to share such findings early, both to demonstrate Claude’s capabilities and to give the broader community insight into what we’re working on.”
The other problem is that another group of scientists may have already found this particular enzyme. Mario Rodríguez Mestre, a computational biologist at the University of Copenhagen, told The New York Times over the weekend that his team found this enzyme back in 2022.
“They are the same systems we have been studying for years,” Mestre told the NYT. He points to a patent that contains the research, for which he is listed as a co-inventor.
Anthropic says that it was unaware of this research.
So, is this a win or not?
One thing that the greater scientific community has been able to agree on is that the speed at which this enzyme was found is actually pretty awesome. Anthropic said that it could have taken human researchers weeks or months to find this needle in the proverbial haystack. All the experts CNET spoke to agreed with this assessment.
“We have hardly scratched the surface of exploring the biological systems represented in species sequenced outside of a handful of model organisms,” Berger said. “So, many accomplished scientists could spend their careers studying interesting sequence patterns without getting to this one.”
It’s no secret that AI has racked up a laundry list of failures over the years. It’s a terrible lawyer, an even worse friend and its contributions to social media are commonly referred to as slop.
The same is not true in this case. Anthropic using Claude to find ART in 21 hours is far faster than any human would’ve done it without an overabundance of luck. If Anthropic and its peers can repeat this success over and over, AI may be adding an item to the very short list of things it’s actually good at doing. However, Berger does warn that lazy AI research can lead to “an overwhelming volume of low-impact observational findings that smothers important, targeted research on promising disease treatments.”
Mestre and his team are less convinced, however, stating that they can’t say for certain whether Claude found these novel enzymes on its own or if the bot had been trained on the treasure trove of unpublished research the scientists had been uploading to Claude for years up to this point to assist them in their own research. That research would’ve sent Claude to that enzyme like a homing beacon.
“I mean, coincidences happen,” Mestre told the Times. Even so, Mestre and his team are shutting down their projects with Claude and moving to other models, just in case. The pickings are slim, since OpenAI was accused of doing this same thing while solving the Navier-Stokes Millennium Problem, and could not rule out that the model had been trained on “deidentified data derived” from Levent Alpöge and Tristan Buckmaster, the mathematicians who actually solved it.
So, all told, Anthropic will have to wait a while longer to collect its Nobel Prize in biology, cure every disease on Earth or even score a clean win for Claude without accusations hanging in the air.
See full bio
