Claude discovers a novel enzyme system with CRISPR-like repeats
Anthropic ● Covered by 2 sources
Claude found a new enzyme system by scanning DNA, then lab tests backed it up. Anthropic says it hints AI can spot biology humans miss.
Based on reporting by Anthropic — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has opened a new life sciences research group and lab, and its first public result is unusually concrete: Claude appears to have found a previously uncharacterized enzyme system in bacteriophages, the viruses that infect bacteria. The company says the discovery came from a workflow that started with a simple prompt, then pushed a lot of the search work onto Claude agents while human scientists handled the lab side.
The target was a huge DNA database full of proteins with no known function. According to Anthropic, roughly 950 agents spent 21 hours searching through 210 million tokens of data, gathering more than 200,000 reverse transcriptases and narrowing them to 3,500 candidate systems, then to 20 that got human-readable reports. One of those agents noticed a repeating DNA pattern sitting next to an odd-looking reverse transcriptase and treated it as something worth chasing.
What made the system interesting was not just the enzyme itself. The setup included a reverse transcriptase, a partner gene, and a long array of evenly spaced DNA repeats. That arrangement reminded the team of CRISPR, which is built around repeat arrays and has become the basis for gene-editing tools, but Anthropic says the new system’s function is still unknown. Early experiments suggest the repeat array is expressed as distinct short RNAs, which raises the possibility that it works in a programmable way, too.
The reverse transcriptase at the center of the find had already been seen in a jumbo phage, but Anthropic says Claude was the first to connect the rest of the pattern: the non-coding DNA repeats and the accessory protein of unknown function. Feng Zhang, who helped pioneer CRISPR genome editing, called the result an exciting example of AI agents contributing to biological discovery and said the RNA-repeat arrays deserve further investigation.
Anthropic is careful not to oversell the lab setup. The work is done by human scientists in a Bay Area lab operating at BSL-1 and BSL-2, and the company says it does not handle human pathogens. The larger point, though, is clear enough: it wants Claude to do the boring-but-brainy part of genomics mining, where a machine can keep staring at sequence data long after a person would have moved on.
My take — AI-written commentary, not fact-checked reporting
This is the kind of AI story that actually deserves attention: narrow, testable, and a lot less silly than the usual “revolution” talk. The real tell is that Anthropic is using Claude to generate hypotheses, then making humans prove them in the lab; that’s much healthier than letting a model freestyle biology and call it science. If more AI companies copied that restraint, the field might have fewer fireworks and more useful results.
Read more about this at: Anthropic