Anthropic claims AI found new enzyme

On September twenty-third, Anthropic announced that its large language model Claude identified an enzyme system with properties reminiscent of Crispr, the Nobel Prize-winning gene-editing tool.

Claude found it fast

About nine hundred fifty Claude agents running simultaneously found the interesting genetic sequences in twenty-one point five hours, according to Anthropic's announcement.

Scientists split on the finding

Fyodor Urnov, a gene-editing expert at the University of California, Berkeley, praised Anthropic for announcing the discovery. But Le Cong, a Stanford professor, expressed skepticism about the timing.

“The experiments are still in the queue. The PR is already live.”

Anthropic admits uncertainty

Anthropic said in an X post that it does not yet understand what this system does, but only a handful of known systems share its features and all can cut, copy, and paste DNA.

“We don't yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA.”

The discovery needs lab proof

One physical experiment was included in a technical report that has not been peer-reviewed. Much more work is needed to determine whether this enzyme system can be used as a gene-editing tool.

Claude searched for reverse transcriptases

Anthropic researchers prompted Claude to search through huge genomic databases for interesting new examples of reverse transcriptases, proteins that copy RNA into DNA, the opposite of the normal process in cells.

Claude found ART in jumbo phages

Claude identified more than two hundred thousand possible reverse transcriptases, narrowed to several thousand new ones, then found an unusual family with repeat DNA sequences resembling Crispr. Anthropic calls it ART, found in jumbo phages.

The enzyme may be a retron instead

An AI agent acknowledged the system could be a retron, a bacterial immune system like Crispr but not the same multi-tool. Retrons have some utility in gene editing, which may explain why Anthropic emphasized the Crispr comparison.

The enzyme was found before

Jason Gill, a microbiologist at Texas A and M University, and his colleagues identified this same reverse transcriptase in a two thousand twenty-one paper on jumbo phages. Claude's novel contribution was identifying repeats around it.

AI found patterns humans might miss

Gill says Claude's ability to find patterns in large datasets in a day is impressive, but cautions against calling it an autonomous discovery. Scientists were involved in the study and ART has no obvious relationship to known Crispr systems.

“These models are good at finding patterns, better than a person staring at it with their eyeballs can.”

Crispr took twenty-five years to develop

It took twenty-five years from the initial discovery of Crispr sequences in bacteria in nineteen eighty-seven until Jennifer Doudna and Emmanuelle Charpentier demonstrated it could cut DNA in two thousand twelve. Only one drug using Crispr was approved by late two thousand twenty-three.