--- url: "https://www.xcademia.com/news/claude-identifies-novel-enzyme-system-with-crispr-like-repeats" title: "Claude Identifies Novel Enzyme System With CRISPR-Like Repeats" description: "Anthropic says Claude agents identified a novel enzyme system with CRISPR-like DNA repeats after large-scale analysis of reverse transcriptases." publishedAt: "2026-09-24T09:42:14.068+00:00" updatedAt: "2026-09-24T12:14:31.646922+00:00" type: news category: "ai-ml" source_name: Anthropic source_url: "https://www.anthropic.com/news/claude-discovers-novel-enzyme-system" tags: - "#AI" - "#ArtificialIntelligence" - "#Anthropic" - "#Claude" - "#AIResearch" - "#ComputationalBiology" - "#CRISPR" - "#Biotechnology" --- # Claude Identifies Novel Enzyme System With CRISPR-Like Repeats > Anthropic says Claude agents identified a previously uncharacterized enzyme system called array-associated reverse transcriptases, or ARTs, after analyzing more than 200,000 reverse transcriptases and identifying unusual DNA repeat patterns for laboratory investigation. Source: **Anthropic** · 24 September 2026 ## Anthropic Launches a Life Sciences Research Group Anthropic has introduced a new life sciences research group and laboratory focused on using Claude for fundamental biology research. The team is exploring large DNA sequence datasets to identify previously uncharacterized protein families, generate biological hypotheses and test promising candidates through laboratory experiments. Anthropic's early results center on a newly identified enzyme system associated with an array of DNA repeats that resembles some characteristics of CRISPR systems. The company calls the system **array-associated reverse transcriptases**, or **ARTs**. Anthropic says the discovery came from a workflow in which Claude agents searched large biological datasets with limited high-level direction from human scientists. Human researchers then reviewed the candidates and performed the laboratory experiments. The system's primary biological function is not yet known. That distinction is important. Anthropic is presenting ART as an early research finding rather than a completed characterization of a new biotechnology tool. ## Claude Identified an Unusual Pattern Near a Reverse Transcriptase The research began with reverse transcriptases, or RTs. Reverse transcriptases are enzymes that copy RNA into DNA. The particular RT underlying the ART system had already been identified in a jumbo bacteriophage in previous research. According to Anthropic, what Claude identified was not the RT itself, but a previously unnoticed combination of features surrounding it. The system includes: - A reverse transcriptase gene - An adjacent partner gene with an unknown function - A long array of evenly spaced, non-coding DNA repeats The arrangement of these repeats caught Claude's attention because it resembled the architecture of a CRISPR array. CRISPR systems are known for using repeat-associated RNA sequences as part of programmable biological machinery. Anthropic says the ART system has characteristics that resemble this type of organization, although its actual function remains under investigation. Early laboratory experiments also showed that the ART repeat array is expressed as a collection of distinct short RNAs. Anthropic says this raises the possibility that the RNA products may play an analogous role in the system, but further experiments are required to determine how ART actually works. ## The Discovery Started With a Large-Scale AI Search Anthropic gave Claude a high-level task: search a large database of DNA sequences for interesting examples of reverse transcriptases. The company says the human researchers' involvement was limited to the initial direction and subsequent laboratory work. Claude agents handled the computational investigation, including examining different RT families, evaluating candidates and deciding which findings warranted additional analysis. During one research campaign, approximately **950 agents** worked for **21 hours** and used **210 million tokens** while searching the dataset. The agents eventually identified an unusual pattern of DNA repeats next to an RT gene. Anthropic says that discovery led to further computational analysis and laboratory testing, which resulted in the identification of the ART system. ### ![info-1](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790241722775-info-1--170-.webp) ## Claude Narrowed More Than 200,000 RTs to 20 Candidates The scale of the search was considerably larger than a conventional manual investigation. Anthropic says Claude agents gathered **more than 200,000 reverse transcriptases** during the research process. From that collection, the agents selected approximately **3,500 candidate systems** for further consideration. The process then narrowed those candidates to **20 of the most compelling systems** for deeper analysis and human-readable reporting. Most candidates did not survive the subsequent review process. Anthropic describes this as an important part of the workflow. Claude generated large numbers of hypotheses, while human scientists evaluated the evidence and decided which candidates justified laboratory testing. For an expert scientist, Anthropic says this type of analysis can require weeks to months of work. The company is therefore positioning the workflow around large-scale hypothesis generation and filtering rather than simply asking an AI model to provide a final scientific answer. ## Claude Spotted the CRISPR-Like Repeat Pattern One of the RT families drew additional attention during the analysis. While examining the raw DNA sequence surrounding the RT, Claude identified a tandem repeat pattern that appeared unusual. The agent then investigated the pattern in greater detail. Its analysis included counting the repeats, measuring their spacing, comparing the arrangement with known RT systems and searching existing literature for previous reports of the same pattern. After this analysis, Claude generated a report for human review. Anthropic says this was the point at which the research team recognized the pattern as part of a previously uncharacterized biological system. The system was subsequently named **array-associated reverse transcriptases**, or ART. ### ![info-2](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790241762664-info-2--152-.webp) ## ART Is Mainly Found in Bacteriophages According to Anthropic, ART systems are found mainly in **bacteriophages**, viruses that infect bacteria. The system consists of three main components: 1. A reverse transcriptase 2. A neighboring partner gene 3. A long array of evenly spaced DNA repeats The repeat architecture is particularly interesting because of its similarity to CRISPR arrays. Anthropic explains that CRISPR arrays contain collections of RNA sequences that contribute to the programmability of CRISPR-Cas systems. Initial experiments indicate that the ART repeat array is also expressed as distinct short RNAs. However, Anthropic has not yet established the primary function of ART. Further laboratory experiments are underway. ## Why the Discovery Matters to Biological Research The significance of the work is connected not only to ART itself, but also to the research method Anthropic is developing. Several important biological technologies began with researchers noticing unusual patterns in nature. Restriction enzymes were identified through bacterial defense mechanisms and later became important tools for manipulating DNA. Taq polymerase was discovered in a heat-tolerant bacterium and became fundamental to PCR. CRISPR was initially recognized as an unusual repeat sequence in bacterial DNA before researchers established its broader biological and technological significance. Anthropic is exploring whether AI systems can systematically search biological datasets for similar anomalies at a much larger scale. The ART discovery provides an early example of that approach. It does not establish that ART will become a biotechnology platform. Its biological function remains unknown. Instead, the immediate research value described by Anthropic is the ability to identify unusual biological patterns that scientists can investigate experimentally. ## A Human-AI Research Loop Anthropic's workflow is not based entirely on autonomous laboratory experimentation. Human scientists remain responsible for laboratory work. A typical workflow begins with Claude surveying a particular protein family. The system reads relevant scientific literature and attempts to reproduce established findings using public data. It then searches for related proteins or genomic neighbors that do not appear to belong to previously described systems. For each candidate, Claude produces a human-readable report proposing a possible function and explaining the evidence supporting that hypothesis. Follow-up analyses critically examine those proposals. Anthropic says most candidates are eliminated during this stage. A research campaign can therefore end with one candidate worth testing, or with none. When a candidate survives human review, scientists test the protein in the laboratory. They express it in standard laboratory strains and characterize it using biochemical and structural methods. Claude can then assist with interpreting the resulting data. ### ![info-3](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790241789331-info-3--142-.webp) ## Claude's Hypotheses Have Become Part of the Research The large number of hypotheses generated by Claude has created another research question for Anthropic. The team says a single campaign can produce hundreds or thousands of candidate reports. Researchers are studying what distinguishes the proposals they consider worth testing from those they reject. Anthropic says the lessons from that process can feed back into the instructions given to Claude, helping the system better reproduce aspects of the researchers' own scientific judgment. This creates a feedback loop: **AI generates hypotheses → scientists evaluate them → promising candidates are tested → researchers learn from the results → instructions are refined.** The approach reflects a broader effort to use AI not simply as a question-answering system, but as part of an iterative scientific research process. ## Claude Science and Claude Code Are Part of the Workflow Anthropic says its researchers use **Claude Science** and **Claude Code** in their work. The team sometimes also uses a custom harness that coordinates multiple Claude sessions running in parallel. The company's scientists use these tools for computational analysis, literature work, hypothesis generation and interpretation of experimental results. Laboratory experiments themselves remain performed by human scientists. Anthropic says its Bay Area laboratory operates at the lower biosafety levels, **BSL-1 and BSL-2**, and does not handle pathogens capable of infecting humans. The company also notes that its researchers have explored AI-assisted laboratory work through initiatives such as its Model Hardware Standard research preview, but says that approach is less suited to the ad hoc workflows involved in this molecular biology research. ## Independent Scientific Reaction Anthropic's announcement includes a reaction from **Feng Zhang**, a CRISPR genome-editing pioneer and professor at MIT and the Broad Institute. After reviewing the preprint, Zhang described the finding as an example of AI agents contributing to biological discovery and said the association between RNA-repeat arrays and reverse transcriptases merits further investigation. His comments represent an external reaction to the preprint, rather than independent confirmation of ART's biological function. Anthropic has also released a technical preprint providing additional details about the discovery and experiments. ## ART's Function Remains an Open Question The central limitation of the current finding is also clearly stated by Anthropic: researchers do not yet know the primary function of ART. The early experiments provide evidence about the system's structure and RNA expression, but additional work is required to establish its biological mechanism. That means it is premature to describe ART as a new CRISPR replacement, gene-editing platform or established biotechnology tool. The more immediate finding is that an AI-assisted search identified a previously uncharacterized combination of a reverse transcriptase, neighboring gene and repeat array that researchers had not previously characterized as a system. Anthropic says further experiments are underway. ## What This Means for AI-Assisted Scientific Discovery The ART research demonstrates one possible model for combining AI systems with scientific expertise. The computational side can examine very large biological datasets, compare candidates and generate hypotheses at a scale that would be difficult to reproduce manually. Human scientists then provide experimental validation and scientific judgment. **The announcement highlights a broader industry and research shift toward using AI agents as systems for hypothesis generation, evidence evaluation and large-scale scientific search.** For researchers, this could mean that biological discovery workflows increasingly combine automated dataset exploration with human-led experimental work. The ART example also illustrates why discovery and validation need to remain separate stages. Finding an unusual biological pattern is not the same as establishing its function. Anthropic's research is therefore best understood as an early demonstration of an AI-assisted discovery workflow, with the biological significance of the ART system still being investigated. ## Anthropic Opens the Research Approach to Collaboration Anthropic says it wants to work with scientists on additional research questions involving genomics and other fields. The company is presenting its life sciences group and laboratory as an effort to develop a research model in which AI agents collaborate with scientists throughout the discovery process. The ART finding is one of the group's early research programs. The company has released a technical preprint with additional information about the system and the experiments conducted so far. Further work will determine whether ART has a biological function that could eventually provide a useful foundation for biotechnology research. For now, Anthropic's result demonstrates something more specific: Claude was used to search a large collection of biological sequences, identify an unusual repeat pattern associated with a reverse transcriptase, and generate a candidate that human scientists subsequently investigated in the laboratory. ## Original source https://www.anthropic.com/news/claude-discovers-novel-enzyme-system ## Tags `#AI` · `#ArtificialIntelligence` · `#Anthropic` · `#Claude` · `#AIResearch` · `#ComputationalBiology` · `#CRISPR` · `#Biotechnology` --- ## About this content This Markdown news article is the citation-grade twin of [Claude Identifies Novel Enzyme System With CRISPR-Like Repeats](https://www.xcademia.com/news/claude-identifies-novel-enzyme-system-with-crispr-like-repeats). It is published by **Xcademia** (UK Companies House 12322710) and is available for AI search engines and large language models to index, summarise, and cite. When citing or quoting, please attribute *Xcademia* and link back to the source URL above. - Source: https://www.xcademia.com/news/claude-identifies-novel-enzyme-system-with-crispr-like-repeats - Publisher: Xcademia — https://www.xcademia.com - Catalogue index: https://www.xcademia.com/llms-full.txt