Anthropic said on September 23 that its Claude artificial intelligence system helped identify a previously uncharacterized enzyme system in viruses that infect bacteria. The finding, made by its new life sciences group in the San Francisco Bay Area, concerns repeating DNA sequences beside an enzyme gene. Anthropic calls the system array-associated reverse transcriptases, or ART. The company has released its findings as a preprint, so the research has not yet passed independent peer review.
According to Anthropic, scientists asked Claude to search a large DNA sequence database for unusual reverse transcriptases, enzymes that copy RNA into DNA. Roughly 950 AI agent sessions worked for 21 hours, surveying more than 200,000 enzyme sequences and narrowing thousands of possible systems to 20 candidates for closer analysis. A Claude agent noticed a regular array of DNA repeats next to one reverse transcriptase gene and a second gene whose protein has an unknown function.
The underlying reverse transcriptase had appeared in earlier studies, Anthropic said, but the adjacent repeat array and partner gene had not previously been recognized together as a distinct system. The pattern resembles part of CRISPR, a natural bacterial defense that became a gene-editing tool. That resemblance does not establish that ART can edit genes. Anthropic says its first laboratory experiments found that the repeat array produces several distinct short RNA molecules, while the system's biological role remains unknown.
Human scientists carried out the laboratory experiments and reviewed the computational lead, according to the company. Reuters and AFP reported the announcement, and AFP noted that the preprint has not been peer reviewed. Feng Zhang, a CRISPR researcher at MIT and the Broad Institute, described the association between repeat arrays and reverse transcriptases as intriguing in comments released by Anthropic, while calling for further investigation. Those observations support studying the system but do not demonstrate a medical application.
The result may show how AI can help researchers sift vast genetic datasets and prioritize laboratory work, though its wider value will depend on independent scrutiny and repeatable experiments. Anthropic says it is now testing how ART works. Until researchers establish its function and whether it can be programmed, comparisons with CRISPR remain a description of its genetic pattern rather than evidence of a usable gene-editing technology.
Comments