Skip to content
Neural Daily

ResearchNews

Claude designs protein binders against scorpion toxins

Anthropic used an internal research Claude model to design minibinders, small proteins that attach tightly to a target, against scorpion neurotoxins. Of 211 designs tested in the lab, 17 met strict criteria. There is no antivenom yet, but all the data is public.

By: Mikola Máté Patrik · · 2 min read

Anthropic illustration with a rock formation, a blue scorpion drawing and a black-and-white lab photo
Image: Anthropic · Press kit · Official research post social image (og:image) · Source

What happened

The idea came from investment research firm Citrini Research: could Claude help a motivated group of non-experts tackle a real health problem (Anthropic)? The team picked scorpion stings. At least 1.2 million people are stung each year and more than 3,250 die, mostly children in places with limited medical care. Existing antivenoms, made from immunized horses, are expensive and work against only a few toxins.

An internal research model running in Claude Science worked for 64 hours, mostly autonomously, with four follow-up prompts. It targeted five toxins from four regions: Mexico (Cn2), Brazil (Ts1), North Africa (AaH I and AaH II) and the Middle East (LqhIII).

The numbers

  • 12 design tools, 318,484 designs and 553,052 co-folded and screened design-toxin pairs.
  • On the order of 10,000 GPU hours.
  • Claude picked 180 designs, and the team added 31 more with Claude's help.
  • Adaptyv Bio and Twist Bioscience tested all 211. 17 qualified as binders: 7 against Cn2, 4 each against AaH II and AaH I, and 2 against LqhIII.
  • 18% of single-toxin designs worked, a typical rate for protein design campaigns.

None of the multi-toxin designs bound more than one toxin. The two strongest, Cn2-B1 and Cn2-B2, bind with single-digit nanomolar affinity. Competition experiments suggest they attach where the LR antibody does, which neutralizes Cn2 in mice.

What is still missing

Binding was measured only against recombinant toxins made in the lab. There were no cell-based neutralization tests and no animal studies, so Anthropic says plainly that it does not claim an antivenom. Two outside experts, Susana Vázquez Torres (CNIO) and Timothy P. Jenkins (DTU), called the work a promising first step.

All 211 designs and their data are on Hugging Face, with details in the technical report. The takeaway: designing a binder is no longer the bottleneck, turning it into a drug is. For neglected diseases with little funding, candidates produced in days could meaningfully speed up that first step.

    Produced with the help of AI, reviewed by an editor. How we make our articles

    Reviewed and edited by the Neural Daily desk. About

    Related articles