Citrini Research used Anthropic’s Claude to test AI-designed proteins that bind scorpion toxin, a project the firm said could reveal investment demand for laboratory infrastructure beyond computing, BlockBeats reported on Oct. 22. Anthropic described the research in a research post.
Protein designs still needed laboratory testing
According to the report, five researchers worked with an unreleased Anthropic model for 64 hours and generated about 320,000 protein designs. They narrowed the output to 180 candidates, then sent those and 31 additional designs to two laboratories. Seventeen proteins bound the target scorpion toxin, but none had been shown to bind several toxins or to neutralize venom.
BlockBeats said the computational work took days, while laboratory validation took nearly two months. Anthropic covered about $50,000 in computing costs, and the laboratory bills approached $150,000. Citrini said the gap could increase demand for synthesis, testing and automation services as AI produces more drug candidates. The firm pointed to Charles River, Revvity and Tecan as examples of companies serving that work; the report did not make a claim about their share-price performance.