XNews

Cloude Discovers Novel ART Enzime System in Bacteriophages

Anthropic announced Claude discovered a new enzyme system called ART in bacteriophage DNA, resembling CRISPR. The system includes a reverse transcriptase, an unknown partner gene, and repeated DNA arrays. Claude independently identified it from 200,000+ candidates in 21 hours using 950 agents.

Key insights

Discovery details and structure

Claude identified a novel enzyme system named ART (Array-associated Reverse Transcriptase) in bacteriophage DNA. It consists of a reverse transcriptase, a partner gene of unknown function, and a long array of evenly spaced repeated DNA sequences. The array is transcribed into distinct short RNAs, suggesting a programmable mechanism similar to CRISPR.

AI-led discovery process

Anthropic's life sciences team gave Claude a broad prompt to find interesting reverse transcriptases in DNA databases. Approximately 950 Claude agents ran for 21 hours, consuming 210 million tokens. They filtered 200,000+ reverse transcriptases down to 3,500 candidates, then 20 most promising, and wrote readable reports. One agent recognized the repeated array directly from raw sequence.

Significance and comparison to CRISPR

The ART system shares structural features with CRISPR and other known programmable systems. CRISPR pioneer Feng Zhang commented that this is a good example of AI agents contributing to biological discovery. Anthropic notes that only a few systems with these features are known, and all have been developed into gene editing tools. The exact function of ART is not yet confirmed.

Anthropic's life sciences lab and future plans

Anthropic established a life sciences research team with a wet lab in the Bay Area, operating at BSL-1/BSL-2 levels. All experiments are conducted by human scientists. The team uses publicly available Claude Science and Claude Code, meaning other labs can replicate the AI-driven hypothesis generation and validation pipeline. Anthropic is seeking external research collaborations.

Broader implications for AI in biology

Anthropic CEO Dario Amodei stated that AI performance in biology is on an exponential trend similar to mathematics. While biology requires experimentation, humans can collaborate with AI to validate results in weeks. He believes AI can accelerate fundamental biological discoveries, potentially increasing the number of drug candidates in the pipeline, even if clinical trial times remain unchanged.

Independent parallel discovery by Stanford

A Stanford team independently described a novel RT system with a non-coding array similar to ART, though the systems evolved independently. This suggests that such systems may be more common than previously thought and that we are at the beginning of discovering and developing them into biotechnological tools.

Source posts