Anthropic details Claude's role in six pharma partnerships, commenters dig into case studies
akbirthko · x · 2026-10-28
Anthropic announced deep research partnerships with six organizations developing new medicines and has begun drug discovery efforts of its own, releasing case studies that @owlposting reviewed in detail:
- BigHat: Claude (mainly Fable 5) read papers, suggested mutations, analyzed results, and helped plan experiments. Public protein models (Boltz-2, Protenix, ESMFold2) were unreliable; smaller models trained on the team's own data worked better. Most interesting: Claude helped diagnose that an instrument's "melting temperature" readings actually reflected proteins clumping out of solution.
- Divergence: Claude Mythos 5.1 helped design a two-part antibody targeting IgE; 3 of 57 anchoring components bound IgE, but the modified antibodies didn't clearly outperform the parent, and the final construct went untested.
- Generate Biomedicines: redesigned a CAR receptor activation domain using Claude Opus/Sonnet 4.6, GPT-5.2, Kimi K2.5, and MiniMax M2.5 early on, later Opus 4.7/4.8. Specialized structure models performed poorly, while LLMs iterating on experimental results produced a design needing 1/5 the RNA to halve donor B cells. The commenter notes a simple membrane-distance rule competed with several tested models.
Overall take: LLM fluid intelligence can contribute to drug development without fine-tuning, but this is more timeline compression than unlocked new capabilities.
More from Models
- Rumors: OpenAI and Anthropic internal models gain reasoning via RL in games and simulated environments — imjustnewatai · 2026-10-28
- Users notice ChatGPT almost always ends replies with "One question:" — Angaisb_ · 2026-10-28
- Community wishlist for next Gemma: better multilingual, OCR, fewer refusals — MaziyarPanahi · 2026-10-28
- Mystery model Space Bunny revealed as early MiniMax M3.1 Flash, free on EvoX Agent Oct 28–30 — lxfater · 2026-10-28
- Accusations fly that Artificial Analysis rankings systematically underrate open-weight models — aiamblichus · 2026-10-28
- AI News Roundup: Gemini 4 Hydrogen Leak, Genie 4 World Model, DeepSeek V5 Rumors — WorldofAI · 2026-10-28