Artificial intelligence is no longer just helping with text or programming. It has now also been involved in the search for previously unknown biological mechanisms. The research team at Anthropic announced that the Claude model identified an unusual enzymatic system while analyzing vast DNA databases. Its structure vaguely resembles CRISPR, a technology that has revolutionized gene editing capabilities.
The newly described system is called ART, which stands for array-associated reverse transcriptases. It consists of a reverse transcriptase, an associated gene, and a long series of regularly spaced segments of DNA. This arrangement is reminiscent of the repeating sequences known from CRISPR systems. ART is primarily found in bacteriophages, viruses that infect bacteria.
The discovery was the result of an unusually extensive digital search. Approximately 950 agents of the Claude model explored genetic databases and evaluated more than 200,000 reverse transcriptases over roughly 21 hours. From thousands of possible candidates, they gradually selected several dozen of the most interesting systems. One of them was notable for a regular pattern in the surrounding DNA, which had not been described in detail by previous research.
The discovery demonstrates how artificial intelligence can accelerate work that would take human experts weeks or months. AI can compare vast amounts of sequences, find unexpected connections, and generate hypotheses for further testing. However, scientists still play a crucial role: they must verify the results, design experiments, and determine whether the discovered mechanism actually works.
Initial laboratory observations suggest that the repeating segments of the ART system create short RNA molecules. However, it is not yet clear what function this system serves in nature or whether it can be used for precise interventions in DNA. The results have only been published as a preprint and have not yet undergone independent peer review.
Experts are therefore tempering expectations that the world has gained a new version of gene editing tools. In addition, the reverse transcriptase itself was documented in an earlier study; the novelty lies primarily in the recognition of a broader arrangement around it. The true significance of ART will only be revealed by further experiments.
This is where the greatest benefit of the report lies. Instead of replacing laboratories, AI offers a way to focus the time, money, and attention of researchers. An algorithm can process millions of records and identify several promising areas, while humans assess their meaning and safety. This combination of speed and expert judgment could contribute to new diagnostic methods, drugs, or more environmentally friendly biotechnologies in the future. However, the path from an interesting pattern to real-world application remains long.
Even with these caveats, this is an encouraging example of collaboration between humans and artificial intelligence. It is not a ready-made cure or a confirmed tool for gene editing. However, it is an interesting clue that would have been difficult to discover without rapid analysis of vast amounts of data. If the potential of the system is confirmed, it could expand the possibilities of biotechnology and also demonstrate that AI can be a valuable partner in basic scientific research.
Jaroslav Holý
