XNews

Claude Discovers CRISPR-like ART Enzyme System

Anthropic announced that Claude identified a new enzyme system called ART (Array-associated Reverse Transcriptase) in phage DNA, with structural similarities to CRISPR. The discovery was made autonomously by Claude agents over 21 hours, analyzing 200,000+ reverse transcriptases. The system's function is unknown but may represent a new gene editing mechanism.

Key insights

Discovery Details of ART System

ART consists of a reverse transcriptase, an unknown partner gene, and a long array of evenly spaced repeated DNA sequences, similar to CRISPR arrays. Experiments confirmed the array is transcribed into distinct short RNAs, suggesting a programmable mechanism. Only a few such systems are known, all capable of DNA manipulation.

Autonomous AI Discovery Process

Approximately 950 Claude agents ran for 21 hours, consuming 210 million tokens. They screened over 200,000 reverse transcriptases, narrowed to 3,500 candidates, then selected 20 most promising. One agent recognized the repeated array in raw sequence, compared it to known systems, and submitted the finding to human review.

Significance and Potential Applications

CRISPR pioneer Feng Zhang commented that this is a good example of AI agents contributing to biological discovery. The ART system could represent a new class of programmable DNA editing tools. Anthropic notes that similar systems have been developed into biotechnological tools, though ART's specific function is still unknown.

Anthropic's Life Sciences Research Approach

Anthropic established a life sciences team with a wet lab in the Bay Area, conducting BSL-1 and BSL-2 research. Human scientists perform experiments, while Claude proposes hypotheses and experiments. The team uses publicly available Claude Science and Claude Code, making the methodology reproducible by other labs.

Dario Amodei's Vision for AI in Biology

Anthropic CEO Dario Amodei stated that AI for biology is on an exponential trend similar to mathematics. He believes AI can accelerate fundamental biology discoveries, potentially leading to curing most diseases in 5-10 years. He emphasized that human-AI collaboration can perform experiments and validate results quickly.

Comparison to Prior and Concurrent Work

The ART system is related to prior discoveries like bridge recombinase and VIPR. A Stanford team independently described a similar RT system with a non-coding array, though the systems evolved independently. This suggests a broader class of such systems waiting to be discovered.

Source posts