Last week, Anthropic announced a discovery that could potentially accelerate the pace of biological research. According to the company, its Claude large language model, operating via 950 simultaneous agents, managed to identify an enzyme system with properties "reminiscent of CRISPR," the Nobel Prize-winning gene-editing tool, in just 21.5 hours.
The Speed of Discovery
The system, which Anthropic has named ART (array-associated reverse transcriptases), was found in "jumbo phages" — large viruses that infect bacteria. The process involved searching through massive genomic databases for "interesting new examples" of reverse transcriptases, proteins that copy RNA into DNA.
While identifying such enzymes can take months of manual mining, Claude narrowed down hundreds of thousands of possibilities to an "unusual" family containing long regions of repeat DNA sequences similar to CRISPR. "I can see by eye a tandem repeat array … that's a CRISPR-like … repeat array?!" one AI agent wrote during the process.
Scientific Skepticism and Transparency
Despite the excitement, the scientific community remains cautious. Le Cong, a professor at Stanford University, noted pointedly: "The experiments are still in the queue. The PR is already live." Although Anthropic has established a wet lab for drug discovery, its technical report has not yet been peer-reviewed.
Furthermore, it emerged that the same reverse transcriptase had been previously identified in 2021 by microbiologist Jason Gill. Claude's novel contribution appears to be identifying the repeats around it, hinting at its function as a gene-editing system. It remains unknown whether ART can actually be utilized as an effective gene-editing tool.
Questions have also been raised regarding the transparency of the model's training data. Anthropic has not released its code or data, making it difficult for independent researchers to reproduce the findings in an era where major journals demand such openness.
