Anthropic: Claude designed protein binders vs scorpion toxins, 17 of 211 validated
nathanbenaich · x · 2026-10-23
Anthropic released a technical report on using Claude to accelerate drug discovery's 'hardest last 20%'.
- Task: an internal research model designed minibinders against scorpion neurotoxins, aiming toward a future antivenom. Citrini Research asked the core question: can AI drug discovery let motivated nonexperts tackle a real global health problem?
- Context: scorpions sting 1.2M+ people yearly, killing 3,250+, mostly children in underserved regions; a neglected tropical disease with no commercial incentive
- Process: Claude generated designs from one prompt plus four follow-ups; Adaptyv Bio and Twist Bioscience tested 211 designs experimentally, 17 met binding criteria; all 211 designs and data released openly
- Caveat: Alasdair Russell notes these are early lab results, not an effective antivenom — translation remains the bottleneck even with AI acceleration
More from Research
- NTU's CodeActionBench tests whether vision-and-code agents can close the robot loop — jiqizhixin · 2026-10-23
- Same nets, 3 seeds: Adam hits 0.033 test loss vs SGD's 0.442 on regression — cneuralnetwork · 2026-10-23
- Event sequence forecasting as generative modeling reveals a realism-accuracy trade-off — Dmitry Osin · 2026-10-23
- NanoStories: 50M-story TinyStories-style dataset generated by rules in 10 minutes — Tall_Abrocoma_3533 · 2026-10-23
- FORC 2027 call for papers: differential privacy, mechanism design, AI alignment — Aaroth · 2026-10-23
- FORC 2027 Call for Papers: Foundations of Responsible Computing — Aaroth · 2026-10-23