Anthropic's Claude AI discovers CRISPR-like enzyme system
AI autonomously identified a novel biological system with programmable DNA-editing potential
1AI-powered biology research breakthrough
Anthropic formed a dedicated life sciences research group in 2026 to explore how general AI models like Claude can accelerate biological discoveries. Their approach combines AI's pattern recognition with human scientific expertise, creating a collaborative workflow where AI generates hypotheses and humans conduct lab validation. The team specializes in computational biology and previously contributed to CRISPR research before joining Anthropic. Their lab in the Bay Area operates at biosafety levels 1 and 2, focusing on molecular biology research without human pathogens.
2The ART enzyme system discovery
Claude autonomously discovered a novel enzyme system called array-associated reverse transcriptases (ART) while analyzing DNA sequences. The system features three components: a reverse transcriptase, an accessory protein, and CRISPR-like repeating DNA sequences. Although its exact function remains unknown, ART shares structural characteristics with other programmable DNA-editing systems. The discovery occurred when Claude analyzed approximately 200,000 reverse transcriptases and identified 3,500 candidate systems before narrowing down to 20 most promising ones for detailed analysis.
3How Claude made the discovery
Researchers prompted Claude to search a massive DNA database for interesting reverse transcriptases. About 950 AI agents processed 210 million tokens over 21 hours to analyze sequences. One agent noticed unusual repeating patterns near a jumbo phage RT gene - patterns resembling CRISPR arrays. Claude then conducted detailed analysis including repeat counting, spacing measurement, and literature review before concluding it had found a new biological system. The AI's workflow mirrors human scientific methods but operates at unprecedented scale and speed.
4Technical characteristics of ART
The ART system consists of three key components: a reverse transcriptase that copies RNA to DNA, an adjacent protein of unknown function, and a long array of evenly spaced DNA repeats similar to CRISPR arrays. Early experiments show these repeats are expressed as short RNAs, suggesting programmable functionality like CRISPR. While the underlying RT was known from previous studies, Claude was first to recognize the complete system's unique configuration. ART appears primarily in bacteriophages, viruses that infect bacteria.
5Scientific validation process
After AI identification, human scientists validated ART through laboratory experiments. They expressed the proteins in standard lab strains and conducted biochemical characterization. The team uses Claude Science and Claude Code tools available to all researchers, sometimes coordinating multiple parallel Claude sessions. Their workflow involves rigorous candidate elimination - from thousands of initial AI proposals, typically only one survives to lab testing. This validation process ensures discoveries meet scientific standards before publication.
6Comparison to historic discoveries
The ART discovery parallels how major biological tools like restriction enzymes, Taq polymerase, and CRISPR were originally found - through noticing unusual patterns in nature. Restriction enzymes launched biotechnology, Taq enabled PCR diagnostics, and CRISPR revolutionized gene editing. Like these examples, ART was found by recognizing peculiar DNA repeats near an enzyme. The team believes AI can systematize such discoveries that previously relied on individual scientists' serendipitous observations.
7AI's role in hypothesis generation
Claude produces hypotheses at unprecedented scale - hundreds to thousands per research campaign. The team studies what distinguishes valuable hypotheses from less promising ones, using these insights to refine Claude's scientific judgment. AI handles initial literature review, data analysis, and report writing, while humans focus on experimental validation and interpretation. This division of labor aims to accelerate discovery while maintaining scientific rigor. The approach represents a new paradigm for biological research.
8Future implications and applications
While ART's function remains under investigation, its CRISPR-like features suggest potential as a programmable molecular tool. The discovery demonstrates AI's capability to identify biological anomalies that might elude human researchers. Anthropic aims to extend this approach to broader genomic research and collaborate with external scientists. They've released a preprint detailing ART and launched the Life Sciences Verification Program, providing biologists with specialized access to Claude models for research applications.