Claude designed protein binders against scorpion neurotoxins
An Anthropic research model autonomously generated 318,484 candidate protein designs in 64 hours, yielding 17 lab-confirmed…
Anthropic had an internal Claude research model design protein binders against five scorpion neurotoxins, working largely autonomously over 64 hours with four human follow-up prompts. It generated 318,484 candidate designs, narrowed these to 211 that were lab-tested by Adaptyv Bio and Twist Bioscience, and 17 met Anthropic's criteria as binders against individual toxins (none succeeded against multiple toxins). Anthropic explicitly states this is not a confirmed antivenom, only a step toward one, since neutralization in a cellular environment has not been verified.
Why it matters
This is a research demonstration that an AI system can autonomously run a multi-day, multi-stage scientific design campaign (target selection, hypothesis generation, large-scale candidate generation, filtering) with a CRO handoff, producing experimentally validated results in a neglected disease area that normally attracts little commercial investment. It signals a shift in how AI labs are starting to use agentic systems for real wet-lab science, not just literature synthesis or coding, though the result itself is an early-stage research finding, not a product or deployable drug.
What is verified
An internal Anthropic research model operating within Claude Science worked largely autonomously over 64 hours, directed by an initial prompt plus four follow-ups, to design protein binders against scorpion neurotoxins. [1]
Claude generated 318,484 protein designs and selected 180 for testing; the team added 31 more for a total of 211 designs tested by Adaptyv Bio and Twist Bioscience. [1]
Of the 211 designs tested, 17 met Anthropic's criteria as binders, each against a single toxin; none of the multitoxin-targeted designs met the criteria for binding more than one toxin. [1]
Anthropic states it has not confirmed whether the binders neutralize scorpion venom in a cellular environment, so it does not claim to have produced an antivenom. [1]
The design campaign used on the order of 10,000 GPU hours and co-folded/screened 553,052 design-toxin pairs. [1]
What is not
- No independent replication of the binding results beyond Anthropic's CRO partners (Adaptyv Bio, Twist Bioscience) is cited.
- No data on cost, timeline or economics compared to traditional antivenom development is provided.
- It is unclear how this internal research model relates to publicly available Claude products used in commercial or developer contexts.
- No information on whether or when this capability might be accessible outside Anthropic's internal research pipeline.
Sources
- 1Anthropic Newsanthropic.com · Primary source
Original source published 20 October 2026 at 02:00.
An AI-assisted briefing by Coding Partners. We link the evidence behind the reporting so you can explore the original work and assess the claims. How CP Signal works.