The Pulse
Anthropic Tests 211 Claude-Designed Scorpion Toxin Binders
Anthropic says 17 of 211 Claude-designed protein candidates met its laboratory criteria for binding scorpion neurotoxins. The company has not shown that the candidates neutralize venom or protect animals.
AI.info Team · Written with AI · fact-checked ·
On October 20, Anthropic reported that an internal Claude research model helped produce 17 protein designs that met laboratory criteria for binding scorpion neurotoxins. Researchers at Adaptyv Bio and Twist Bioscience tested 211 candidates; the result is an early-stage finding, not evidence of a working antivenom.
318,484 designs in a 64-hour campaign
Anthropic’s team set Claude the task of designing minibinders—small proteins intended to attach to a target—against five neurotoxins from scorpions associated with Mexico, Brazil, North Africa and the Middle East. The targets were Cn2, Ts1, AaH I, AaH II and LqhIII. The researchers aimed the designs at sites where the toxins bind to sodium channels, which help nerve cells fire.
Claude generated 318,484 candidate designs and selected 180 for testing. Anthropic researchers added another 31 with Claude’s help, bringing the total tested to 211. The model worked mainly autonomously for 64 hours, but the team supplied four follow-up prompts, including directions to pursue designs that could target multiple toxins.
The company says its evaluation required evidence of binding in every replicate, a dissociation constant of 1 micromolar or tighter, and no signal against CLEC12A, an unrelated human protein used to check for off-target binding. By those criteria, 17 designs bound a single toxin. None of the candidates designed to target multiple toxins met the criteria for binding more than one.
Binding is not venom neutralization
Anthropic’s two strongest candidates, Cn2-B1 and Cn2-B2, bind Cn2, a toxin from the Mexican scorpion Centruroides noxius. The company reports single-digit nanomolar binding measurements for both in tests by Adaptyv Bio and Twist Bioscience. Competition experiments also suggest they attach at or near the site used by LR, an antibody previously reported to neutralize Cn2 in mice.
That evidence does not establish that either candidate stops the toxin from harming cells or protects an animal. Anthropic says it has not tested the binders for neutralization in cells or efficacy in living animals, and the experiments used recombinant toxins rather than toxin purified from venom. The company describes testing against purified venom, measuring effects on sodium channels and then testing in mice as steps needed to assess a possible treatment.
A research result, not a treatment
Anthropic frames the work as a possible way to speed early discovery for a neglected health problem. The company says scorpion stings affect at least 1.2 million people a year and cause more than 3,250 deaths, mostly among children in places with limited access to medical care. Existing antivenoms are generally made using antibodies from immunized animals; Anthropic says they can be costly, hard to produce and limited in the toxins they cover.
Timothy P. Jenkins, an associate professor at the Technical University of Denmark, called the work a promising step toward exploring designs across several toxins. Susana Vázquez Torres, a postdoctoral researcher in the Structural Biology Programme at Spain’s National Cancer Research Centre, wrote: “The challenge is to use these extraordinary tools to develop molecules that can become real drugs, ones that are safe, affordable, and widely accessible.”
Anthropic says it is releasing all 211 designs and their data for other researchers. The next concrete test is whether a candidate such as Cn2-B1 can bind toxin from venom and prevent it from disrupting sodium channels; until those experiments are done, the results show binding, not an antivenom.