Anthropics Claude Science system generated 318,484 protein binder designs to counteract scorpion neurotoxins without direct human intervention for the first twenty-four hours. The AI targeted five distinct toxins from four scorpion species responsible for over 3,250 deaths annually. Researchers subsequently reviewed the output and selected 211 candidates for laboratory testing. Of these, 17 met the criteria for functional binders, with the two strongest specimens docking tightly to Cn2, a toxin from the Mexican scorpion.
The significance lies in the shift from traditional antibody engineering to autonomous protein design. Independent labs confirmed the results, validating that the AI could identify structures capable of neutralising specific toxins. This approach bypasses the typical 10 to 15 percent success rate of standard methods to achieve hit rates between 22 and 35 percent.
* Claude operated autonomously for 24 hours before receiving further instructions
* External experts Susana Vázquez Torres and Timothy P. Jenkins called it a promising first step
* The strongest binders dock at the same site as an existing antibody known to neutralise Cn2 in mice
