Claude discovers novel enzyme system ART with CRISPR-like repeats
Anthropic's Claude autonomously discovered a novel enzyme system, Array-Associated Reverse Transcriptases (ART), in bacteriophages, demonstrating AI's potential to accelerate fundamental biological discovery by identifying complex molecular machines in DNA.
The Discovery of Array-Associated Reverse Transcriptases (ART)
The system discovered by Claude, named Array-Associated Reverse Transcriptases (ART), is a previously uncharacterized enzyme system found primarily in bacteriophages (viruses that infect bacteria).
ART consists of three defining components: a reverse transcriptase (RT), an associated partner gene, and a long array of evenly spaced DNA repeat sequences. The arrangement of the DNA repeats in ART strongly resembles a CRISPR array, which is known to hold a bank of distinct RNA sequences that make CRISPR-Cas systems programmable biotechnological tools.
While the underlying reverse transcriptase (RT) had been identified in prior studies from a jumbo phage, Claude was the first to notice the system’s defining features: the associated array of non-coding DNA sequences and an additional accessory protein of unknown function. Early experiments indicate that the ART array is expressed as a set of distinct short RNAs, suggesting an analogous function to known programmable systems.
Claude's Methodology and Process
The discovery was the result of a systematic, AI-driven search through massive genomic databases, which dramatically reduced the time required for initial candidate identification.
Anthropic provided Claude with a prompt to search a large database of DNA sequences for interesting new examples of RTs. Claude agents then combed through the data, investigated distinct RT families, and used their own judgment to identify compelling candidates.
This process involved gathering over 200,000 RTs and narrowing them down to 3,500 new candidate systems. Claude agents further analyzed these candidates, producing human-readable reports for the 20 most compelling candidates. For an expert scientist, this type of initial analysis typically requires weeks to months of work.
Specifically, after 21 hours of searching the data using approximately 950 agents and 210 million tokens, one agent detected a remarkable pattern: a repeating tandem array of DNA sequences located next to an unusual RT gene. Following this detection, Claude proceeded like a human scientist, counting the repeats, measuring their spacing, and comparing the layout to known RT systems before filing a report for human review.
Context and Implications for Life Sciences
This finding highlights a new paradigm for biological research, where general AI models collaborate with human scientists at every stage of the discovery process. The ability of Claude to autonomously detect anomalies and drive analyses to initiate biological discoveries demonstrates significant acceleration potential in life sciences.
Historically, major biological breakthroughs—such as the discovery of restriction enzymes or Taq polymerase—started with a scientist noticing an unusual feature in the molecular diversity of nature. CRISPR itself was first noticed as an unusual repeat sequence in bacterial DNA, launching modern gene editing.
By applying AI to genomics, Anthropic's research group aims to systematize and accelerate such discoveries. The team's approach involves Claude surveying protein families, reproducing established public data, and then searching for family members or genomic neighbors that fit no described system. When a candidate survives human review, it is tested in the laboratory, with Claude assisting in interpreting the resulting data.