Anthropic says Claude designed binders for scorpion neurotoxins

News

Anthropic says Claude helped design 211 minibinders against scorpion neurotoxins, with 17 meeting binding criteria in lab tests.

Anthropic said on Oct. 20, 2026 that an internal Claude research model helped design protein minibinders against scorpion neurotoxins, with 17 of 211 lab-tested designs meeting the company’s binding criteria. The company said the work is a research demonstration rather than evidence of an antivenom, but argued it shows how AI could speed early-stage discovery for neglected diseases.

Evidence and context

In the research announcement, Anthropic said Claude worked from one initial prompt and four follow-up prompts over 64 hours, operating mainly autonomously within Claude Science, while the company’s life science team supervised the campaign. The system generated 318,484 protein designs, selected 180 for testing, and the team added 31 more with Claude’s help, bringing the total to 211 designs. Adaptyv Bio and Twist Bioscience tested the designs for binding.

The campaign targeted five scorpion neurotoxins across four regions: Cn2 from Mexico, Ts1 from Brazil, AaH I and AaH II from North Africa, and LqhIII from the Middle East. Anthropic said the designs were aimed at each toxin’s channel-binding site, and that Claude also proposed multitoxin strategies, although none of the multitoxin designs met the company’s binder criteria for more than one toxin.

According to Anthropic, the 17 binders that met its threshold all bound single toxins, with seven against Cn2, four against AaH II, four against AaH I and two against LqhIII. The company said 11 of the 60 single-toxin designs were binders, a hit rate of 18%. It also said its strongest Cn2 binders, Cn2-B1 and Cn2-B2, bound in the single-digit nanomolar range and appeared in competition experiments to attach on or near the same site as an antibody reported to neutralize Cn2 in mice.

Anthropic said the study does not establish neutralization, safety or clinical efficacy. The company noted that it has not tested the strongest binders in cells or animals, and that the results were measured against recombinant toxins rather than toxin purified from venom. It said further work would be needed to confirm activity in vitro and in vivo before any therapeutic claim could be made.

The announcement follows earlier Anthropic research on AI-assisted protein design and is part of a broader series of company reports on Claude in scientific discovery. Anthropic said it is releasing all 211 designs and binding data for other researchers to use.