Claude autonomously discovers a novel enzyme system, ART, with CRISPR-like repeats

Claude agents autonomously carried out a genome-mining campaign from only an initial prompt: gathering over 200,000 RTs, picking out 3,500 new candidate systems, and narrowing those to the 20 most-compelling candidates analyzed into human-readable reports. Genome mining has relied on researchers searching sequence databases for uncharacterized genes and working out what they do, which the text says can take an expert weeks to months; here candidate generation and first-pass screening were handled by agents running in parallel. The text gives concrete scale: roughly 950 agents, 21 hours, 210 million tokens, over 200,000 RTs, 3,500 candidates, and 20 top candidates; this describes the workflow and its scale, with final judgment retained by human scientists.

The agent identified a previously uncharacterized three-part enzyme system, ART: a reverse transcriptase, a partner gene beside it, and a long array of evenly spaced DNA repeat sequences. The text states the underlying RT had been identified in previous studies, but Claude appears to be the first to notice the system's defining features, an associated array of non-coding DNA sequences and an additional accessory protein of unknown function. The text records the agent's reasoning: it counted the repeats, measured their spacing, compared the layout with known RT systems, searched the literature for any previous report of the pattern, and then filed a report for human review.

First laboratory experiments show the ART array is expressed as a set of distinct short RNAs, suggesting something analogous to CRISPR may be at play. CRISPR arrays hold a bank of different RNA sequences that make CRISPR-Cas systems programmable tools; the ART repeat layout resembles a CRISPR array, and the short-RNA expression result offers initial experimental support for the analogy. The text calls these 'first experiments' and states that work to understand the primary function of ARTs is ongoing, so this is early experimental evidence rather than a confirmed function.

This work speaks to researchers and life science professionals interested in how AI agents can take part in fundamental biological discovery, and it applies in a BSL-1 and BSL-2 setting, with all laboratory work performed by human scientists. The text states ART is found mainly in bacteriophages and that its function is not yet known, so the current result is a candidate system meriting further investigation rather than a ready-to-use tool. The team also says it wants to work with other scientists to extend the approach to problems in genomics and other fields.

The primary function of ART remains undetermined, and the text says further experiments are underway; the short-RNA expression result comes from first experiments and does not yet explain how the system works. This post is also an early-results introduction, so experimental details, sequence analyses, and structural characterization need to be checked in the preprint and technical report; without those details, how far the analogy to CRISPR systems extends remains an open question.