Scorpion venom: a Claude model designs 17 mini-proteins that bind to neurotoxins, out of 211 tested in the laboratory

Anthropic and Citrini Research let an internal model work for 64 hours on binders against five scorpion toxins. The first step towards an antivenom, without demonstrated neutralization.

Anthropic published the results of a protein design campaign conducted almost independently by an internal research model, in the Claude Science environment. The aim, chosen with the analytical company Citrini Research, was to design small proteins capable of clinging to scorpion neurotoxins, with a future antivenin in line. Scorpion bites affect more than 1.2 million people per year and cause more than 3,250 deaths, especially in children in poorly equipped areas, and current treatments, produced from immune horse antibodies, are expensive and cover only a small number of toxins.

In 64 hours, guided by an initial prompt and four revivals, the model generated 318,484 candidates and selected 180; the team added 31. The Adaptyv Bio and Twist Bioscience laboratories tested the 211 designs: 17 bind to one of four targeted toxins from species from Mexico, Brazil, North Africa and the Middle East.

Anthropic specifies that it has not verified that these binders actually neutralize venom in a cell environment and therefore does not claim antivenin. Two external experts see this as a promising first step for neglected tropical diseases. The 211 sequences and their data are made public.