Claude found CRISPR-like repeats beside a phage reverse transcriptase gene during a search for something else
Anthropic says Claude found a CRISPR-like repeat array beside a reverse transcriptase gene in a bacteriophage, function unknown. The company's researchers say they have no idea yet what it does, so the finding is a lead for lab validation.
The Scientist · Science desk
What happened
- The prompt explicitly asked Claude to look for accessory proteins riding along with reverse transcriptase genes, and Claude returned an array of repeats in noncoding RNA instead.
- Kauderer-Abrams described the prompt as fairly generic and said an amateur biologist would not have written it.
- Anthropic has run comparable searches through other families of DNA polymerases, Kauderer-Abrams said.
- He called this the first discovery in Anthropic's lab that Claude made on its own.
Why it matters
- What it limits constraint With the function unknown, nothing yet supports calling the array a tool. That needs experiments on the enzyme and the repeats.
- What's now possible capability An agent asked about accessory proteins surfaced a different structure in the same sequences. That is useful behaviour for sequence mining, though a single instance cannot show how often it happens.
- What it costs cost Any lab tempted to follow up would be starting its own validation, since the only account so far is an interview with the company's own staff.
- What it sets up precedent Anthropic names mining nature for tools as its first self-directed scientific problem, so we'd expect more candidates from the same pipeline, each facing the gap between a flagged sequence and a demonstrated function.
The account of the find is an interview with two Anthropic staff: Eric Kauderer-Abrams, who leads the company's life sciences work, and Nicholas Perry, one of its researchers [11]. It does not say how many searches Claude has run, how many returned a candidate, or whether anyone has tested the array at the bench. One find from an unknown number of runs shows that the agent can produce a lead.
Kauderer-Abrams was careful about scale. He had cited PCR tests and CRISPR as discoveries that came from looking around in nature [8], and then drew a line: "we are by no means claiming that what Claude has discovered here is of that status," he said [9]. On the array itself he said it has "interesting properties that may be suggestive of having some potential as a tool" but that "we have no idea what its function is yet" [3].
As described, the prompt set no test that a repeat array could pass [5]. That makes the result an interesting surprise and a hard one to evaluate. Whether a standard sequence-analysis pipeline, run over the same sequences, would have flagged the same array is an open question. Without that comparison, what Claude added is hard to separate from what the data already held. Kauderer-Abrams, judging his own company's model, said Claude showed "exactly the type of logic that the really talented computational biologists that we work with would go after" [14].
We think the finding is best filed as a candidate for validation. A reverse transcriptase copies RNA into DNA [13], and Scientific American calls the neighbouring repeats "a hint that this discovery may have some useful properties" [15]. Both scientists say much more work is needed to understand the region, and that Claude is no replacement for biologists [10].
What to watch
- A wet-lab result on the enzyme and the repeats, from Anthropic or an outside group, showing what the array actually does.
- Disclosure of the phage, the number of searches run and the number of candidates returned, which would give a hit rate.
- A side-by-side run of a standard sequence-analysis pipeline on the same sequences, to show what Claude added.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence25
- Adoption
- Not scored
- Hype gap−10
- Incentives
- Not scored
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic claimed it had discovered a stretch of DNA in a bacteriophage that repeats over and over, a feature similar to that of CRISPR.
- [2]
The newly discovered DNA repeats sit next to a gene for a reverse transcriptase enzyme.
- [3]
Eric Kauderer-Abrams said: "We know that it has these interesting properties that may be suggestive of having some potential as a tool, but we have no idea what its function is yet, and we're at the very beginning of the research."
- [4]
Anthropic gave Claude a fairly generic prompt to look for particular patterns in DNA sequences associated with reverse transcriptase enzymes, and Kauderer-Abrams said it was not a prompt an amateur biologist would have written.
- [5]
The prompt explicitly asked Claude to look for accessory proteins that might be hitching a ride along with reverse transcriptase genes; Claude instead found an array of repeats in noncoding RNA.
- [6]
Anthropic has run these sorts of searches through other families of DNA polymerases.
- [7]
Kauderer-Abrams said: "This was the first discovery in our lab that Claude made on its own."
- [8]
Kauderer-Abrams cited PCR tests and CRISPR as examples of discoveries that came from looking around in nature and finding odd things.
- [9]
Kauderer-Abrams said: "we are by no means claiming that what Claude has discovered here is of that status."
- [10]
The scientists say much more work is needed to understand what this region of the genome does, and that Claude is no replacement for biologists.
- [11]
Scientific American interviewed Eric Kauderer-Abrams, Anthropic's head of life sciences, and Nicholas Perry, a researcher at the company.
- [12]
Kauderer-Abrams said the first scientific problem Anthropic decided to pursue itself was to find, systematize and scale up the finding of useful new things that already exist in nature and turn them into tools.
- [13]
A reverse transcriptase is a type of enzyme that copies RNA into DNA.
- [14]
Kauderer-Abrams said Claude showed "exactly the type of logic that the really talented computational biologists that we work with would go after."
- [15]
Scientific American describes the adjacency of the repeats to the reverse transcriptase gene as "a hint that this discovery may have some useful properties."
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.