推文

@InGodWeBelieveU · 2026-10-14 06:26

The target was found with generative AI. So was the molecule. Then it went into patients, in a randomized, placebo-controlled trial. Here's how I'd read the result, in 4 steps: ➊ What the trial was built to test
↳ Phase 2a, 𝟕𝟏 patients with idiopathic pulmonary fibrosis (IPF)
↳ 12 weeks, four arms: 30 mg once daily, 30 mg twice daily, 60 mg once daily, placebo
↳ Primary endpoint: safety, not efficacy ➋ Safety first
↳ Any adverse event: 𝟕𝟐–𝟖𝟑% on rentosertib vs 𝟕𝟏% on placebo
↳ Treatment stops were mostly liver toxicity or diarrhea ➌ Then the signal (a secondary endpoint)
↳ Lung capacity (FVC), 60 mg once daily: +𝟗𝟖.𝟒 ml (95% CI 10.9 to 185.9)
↳ Placebo: −𝟐𝟎.𝟑 ml (95% CI −116.1 to 75.6)
→ Promising direction. Wide, overlapping intervals. ➍ Then the limits
↳ About 18 patients per arm, 12 weeks
↳ Many authors work at the company that developed the drug
↳ The authors themselves call for larger, longer trials In cancer research, a phase 2a asks "is it safe, and is there a hint?" It doesn't ask "does it work?" "AI-discovered" is a story about how the molecule was found. The trial design is what tells you whether it helps. I'm not a clinician; this is a research read, not medical advice. Would you trust an AI-discovered drug more, less, or the same once it clears a proper randomized trial? Follow @InGodWeBelieveU for plain-English reads of AI and medicine news, with the paper named in every post so you can check it yourself. #AI #DrugDiscovery #ClinicalTrials #Science

曝光 23 · 评论 1 · 点赞 0 · 书签 0 · 曝光/时 1.134507256607482