Claude Designs Proteins That Neutralize Scorpion Venom
An internal Claude-based model designed protein binders against scorpion venom with almost no human help, and 17 of 211 tested designs actually worked.
Anthropic, the company behind the Claude models, showed how an internal research model built on the Claude architecture designed protein binder molecules against scorpion venom. A protein binder is a small protein engineered to latch tightly onto one specific target. In this case, the targets were venom neurotoxins. The model worked with almost no human help. It ran for 64 hours and needed only four clarifying prompts from human supervisors. It generated 318,484 candidate designs. Two contract research organizations, specialized labs that build and test drug candidates, Adaptyv Bio and Twist Bioscience, tested 211 of them in the lab. Binding to the toxin was confirmed in 17. The two strongest binders target Cn2, a toxin from the Mexican scorpion Centruroides noxius. Their binding strength, measured in nanomolar units, where lower means tighter binding, came in at single-digit values. That is close to the best known antibodies, which bind below 1 nanomole.
This is not a scorpion-venom cure yet. The protection has not been tested in living cells or in animals. None of the 211 designs worked against more than one type of toxin at once. Still, the speed is striking. Work like this usually takes months of manual lab research. Here, the model ran almost the entire process on its own. Anthropic also published all 211 designs and their binding data in the open, rather than just announcing the results.
Source: www.anthropic.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides

Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles