Claude designed 17 scorpion-toxin binders, Anthropic says

An internal Claude research model designed 17 protein binders against scorpion neurotoxins that met Anthropic’s lab criteria, the company said Tuesday, releasing all 211 tested designs and their data.

The campaign: Anthropic said the model worked “mainly autonomously” for 64 hours, directed by four follow-up prompts, and generated 318,484 protein designs. It selected 180 for testing, and Anthropic said its team added 31 more with Claude’s help.

Lab results: Adaptyv Bio and Twist Bioscience tested all 211 designs, and 17 met Anthropic’s criteria for a binder, each against a single toxin, the company said. Seven bound Cn2, four bound AaH II, four bound AaH I and two bound LqhIII, according to Anthropic. None of the designs made against multiple toxins met the criteria for more than one, the company said.

Strongest binders: Anthropic said its two strongest binders, Cn2-B1 and Cn2-B2, bound Cn2, a neurotoxin from the Mexican scorpion Centruroides noxius, at both labs with affinities “in the single-digit nanomolar range.” The company said competition experiments suggest they attach on or near the same site as LR, an antibody reported to neutralize Cn2 in mice.

Not an antivenom: Anthropic said it has not confirmed whether the binders neutralize scorpion venom in a cellular environment, “so we do not claim to have produced an antivenom.” It said it did not test any design in cells grown in the lab or in living animals.

Outside view: Anthropic said two outside experts, Susana Vázquez Torres, a postdoctoral researcher in the Structural Biology Programme at the Spanish National Cancer Research Centre in Madrid, and Timothy P. Jenkins, an associate professor at the Technical University of Denmark, called the work “a promising first step.” Jenkins said designing credible binders against five toxins in days “is a great showcase of the AI-driven paradigm shift we are seeing in therapeutic discovery.” Vázquez Torres said “binding itself is no longer the main bottleneck.”

What’s next: Anthropic said Cn2-B1 would be one of its strongest candidates to advance, but that it would first need to confirm binding to toxin purified from venom, then test function with patch-clamp recordings and protection in mice.