Anthropic releases 211 protein designs from scorpion toxin research
Anthropic published 211 protein designs and laboratory results, with 17 meeting its toxin-binding criteria but none established as an antivenom.
AnthropicAnthropic released 211 experimental protein designs and their laboratory data on October 20 following a Claude-assisted study of scorpion toxins. The company said 17 designs passed its requirements for binding to a toxin. The results do not establish that the proteins can treat scorpion stings.
An internal research version of Claude generated 318,484 candidates during 64 hours of computational work. Researchers guided it with an initial prompt and four follow-ups. Adaptyv Bio and Twist Bioscience tested the selected proteins, and the successful designs bound individually to four toxins.
The project examined whether AI could help develop proteins for an area with limited commercial investment. None of the designs intended to target several toxins passed the company's binding criteria for multiple targets. Attaching to a toxin also does not establish that a protein blocks its harmful effects.
Anthropic has made the designs and measurements available for further research. The study did not test toxin neutralization in cells or effectiveness in animals, leaving those questions unresolved before any treatment could be established.
The day's biggest AI news in one short email, every morning. Plus an instant heads-up when a reset or a big release lands.
More on Anthropic
- Oct 20Eric Schmidt reportedly holds about 1% of Anthropic ahead of planned IPORumor
- Oct 20Anthropic reportedly pitches AI safety bill with limits on government interventionRumor
- Oct 19Claude Code 2.1.304 is outUpdate
- Oct 19Nscale and Anthropic reportedly discuss second $45 billion computing dealStrong rumor
- Oct 19Musk pledges SpaceX stock for every US child in 2027Big story
- Oct 18Anthropic proposes 5% equity allocation for new public benefit fundBig story