Anthropic's Claude designs 17 lab-validated protein binders against scorpion neurotoxins

What Anthropic did

Anthropic asked an internal Claude research model to design minibinders — small proteins built to attach tightly to a chosen target — against neurotoxins found in scorpion venom, then sent the results to outside laboratories for testing. According to Anthropic, teams at Adaptyv Bio and Twist Bioscience, two contract research organizations, screened 211 designs, and 17 met the criteria for a binder. The company is releasing all 211 designs and their data as a public resource for the field.

The project began when Citrini Research, a financial investment research firm, approached Anthropic with a question: could AI drug discovery let a group of motivated nonexperts tackle a genuine human health problem? Anthropic's life-science team, brainstorming with Claude, settled on scorpion envenomation, with the hope that the resulting binders could one day seed a new antivenom.

A 64-hour, largely autonomous design run

Anthropic started from a prompt adapted from an earlier single-prompt binder experiment it described in August, modified together with Citrini Research. Over roughly 64 hours, an internal research model running within Anthropic's Claude Science environment worked mostly on its own, steered by four follow-up messages. It produced 318,484 candidate protein designs, selected 180 of them for testing, and Anthropic's researchers added 31 more with Claude's help, bringing the total to 211. Under human supervision, Claude also analyzed the results, generated the figures and drafted the final report.

The 17 confirmed binders each attach to one of four scorpion neurotoxins known to be dangerous to humans.

Why scorpion venom

Scorpions sting 1.2 million people or more every year, and more than 3,250 die, mostly children in regions with limited access to medical care, according to Anthropic. Because new treatments are costly to develop and unlikely to earn that money back, scorpion envenomation has been described as a neglected tropical disease. The relevant toxins attach to sodium channels in nerve-cell membranes, causing nerves to fire at the wrong time and releasing stress chemicals such as adrenaline; severe cases can lead to heart failure and fluid buildup in the lungs.

Today's antivenoms are made by immunizing horses with venom milked from thousands of scorpions, then purifying antibodies from the animals' blood. The process takes months, usually covers only a small number of neurotoxins, and the products are expensive and often scarce where stings are most common.

The campaign targeted five toxins from four regions: Cn2 from Mexico, Ts1 from Brazil, AaH I and AaH II from North Africa, and LqhIII from the Middle East. All five share one overall fold, but any two are only 26% to 62% identical in amino-acid sequence, which made them a demanding test for multitoxin design — fewer molecules would mean a cheaper, more broadly useful antivenom cocktail. Claude chose toxin structures by searching the Protein Data Bank and the AlphaFold database, picked five binding hotspots on each toxin, and was asked for four classes of binders: one per individual toxin, one for the beta-family pair Ts1 and Cn2, one for the alpha-family trio AaH II, AaH I and LqhIII, and one aimed at all five simultaneously.

What the result does and does not show

Anthropic is careful about scope. The designs have not been tested for whether they neutralize venom in a cellular environment, and the company explicitly does not claim to have produced an antivenom; binding was the endpoint that was measured.

Two outside experts who reviewed the work — Susana Vázquez Torres, a postdoctoral researcher at the Spanish National Cancer Research Centre, and Timothy P. Jenkins, an associate professor at the Technical University of Denmark — described it as a promising first step, with Jenkins arguing that credible binders against five toxins designed in parallel within days illustrates how AI is shifting therapeutic discovery, especially for conditions long starved of commercial investment. Anthropic also notes precedent and novelty: an earlier project designed de novo proteins that neutralized snake neurotoxins and protected mice, and, to its knowledge, no de novo binders against scorpion neurotoxins had been described before this one.

Why it matters

Anthropic's own conclusion is that obtaining tight binders quickly is no longer the hard part — binding is no longer where drug discovery gets stuck. The difficult work lies further down the pipeline: proving the molecules neutralize venom, then making them safe, affordable and available in the regions where children are dying. For neglected diseases, where low expected returns have suppressed innovation for decades, compressing months of design work into days changes the economics of even starting a project. And the origin of this one — a financial research firm asking a frontier AI lab a side question that turned into a full experimental campaign — suggests the expertise needed to launch credible drug-design work is starting to spread beyond traditional pharmaceutical labs.