Anthropic Claude AI Doubles Protein Design Success Rate With Binders for 14 of 15 Targets
Research44d agoA new batch of lab tests confirmed that Claude can autonomously run computational protein design campaigns to produce molecules that work in a laboratory setting. The AI generated 354 successful binders from 1,320 candidates, achieving hit rates between 22% and 35% and outperforming human competitors on 4 of 6 comparable targets. These results, validated independently by Adaptyv Bio and Twist Bioscience, more than double the industry standard success rate of 10% to 15% and included working binders for 14 of 15 targets.
While high affinity binders are a necessary first step in drug development, they are not final drugs. Anthropic is using the experiment as a foundation to teach Claude to manage the end to end development of antibodies and small molecules. The company released a technical report and open sourced its data and prompts, identifying Opus 5 as its most capable model for life science research.
Key sources
- SOURCE@anthropicai“We then worked with Adaptyv Bio and Twist Bioscience, who independently built and tested the proteins Claude designed.”x.com
- SUPPORT@anthropicai“Some of its strongest designs bound several times more tightly than the best published de novo binder.”x.com
- SUPPORT@anthropicai“teaching Claude to run the entire development process end-to-end for every major type of drug molecule—from antibodies to small molecules.”x.com
- SUPPORT@anthropicai“Opus 5 remains our most capable model available for life science research.”x.com
- SUPPORT@anthropicai“open-sourcing our prompts and data here: https://t.co/OZXVQWSie2”x.com
- SUPPORT@rohanpaul_ai“A lab may no longer need a dedicated protein-design expert to manually run every computational step.”x.com