AI BriefingAnthropicPress Releases22:30
AI summarized from verified sources
Claude designs protein binders with high precision to accelerate drug discovery research
Reduces trial-and-error in protein design and streamlines early drug discovery research.
SOURCE CHECK
2 sources
Sources
Key Points
- 1Successful binders for 14 of 15 targets
- 222-35% success rate (over 2x typical)
- 3Some exceed prior best affinities
- 4Used Opus 5 and Mythos Preview
Anthropic ran experiments where Claude designed binders for 15 protein targets. Wet lab results showed success on 14 targets with 22-35% success rates, exceeding the typical 10-15%. Some designs bound more tightly than prior bests.
What happened
Anthropic had Claude autonomously run protein design tasks and validated results in external labs. The model completed work that typically takes experts weeks, achieving high success rates.
Impact for users
Life sciences researchers can expect major time savings in the design process. Future expanded access for scientists should make it even more usable.