AI News AI资讯 4h ago Updated 1h ago 更新于 1小时前 51

AI-designed drug appears to turn back the body's biological clock in early trial AI设计药物在早期试验中似乎能逆转人体生物钟

Insilico Medicine's AI-designed drug rentosertib, originally developed for idiopathic pulmonary fibrosis (IPF), showed signs of reversing biological aging markers in a 42-patient clinical trial Six independent AI aging clocks from institutions including Harvard, Oxford, and Beijing all predicted lower biological age in treated patients, with reductions of up to six years The drug was developed using two AI systems: one identifying disease-relevant protein targets (TNIK) and another generating ma Insilico Medicine使用生成式AI开发的药物rentosertib在42名特发性肺纤维化(IPF)患者试验中显示可逆转生物衰老标志物 6个独立AI衰老时钟一致预测治疗组患者生物学年龄降低3-6年,且该效果与肺功能改善的最佳剂量不同,暗示存在独立于肺功能的机制 药物通过AI系统发现TNIK蛋白靶点并设计分子,从靶点到候选药物仅耗时18个月,目前已进入III期临床试验阶段 专家肯定结果但强调样本量有限且缺乏健康人群数据,衰老时钟的可靠性仍需进一步验证

75
Hot 热度
70
Quality 质量
72
Impact 影响力

Analysis 深度分析

TL;DR

  • Insilico Medicine's AI-designed drug rentosertib, originally developed for idiopathic pulmonary fibrosis (IPF), showed signs of reversing biological aging markers in a 42-patient clinical trial
  • Six independent AI aging clocks from institutions including Harvard, Oxford, and Beijing all predicted lower biological age in treated patients, with reductions of up to six years
  • The drug was developed using two AI systems: one identifying disease-relevant protein targets (TNIK) and another generating matching molecules, completing the process in approximately 18 months
  • Results were published in Nature Biotechnology, with protein profile comparisons against 55,000+ UK Biobank samples showing reversal of age-related protein changes
  • Experts caution that small sample size, lack of trials in healthy individuals, and the indirect nature of aging clock measurements limit definitive conclusions

Why It Matters

This represents a significant convergence of generative AI drug discovery and the longevity research field, demonstrating that an AI-designed molecule can produce measurable shifts in biological age markers. For AI practitioners and pharma companies, it validates the potential of generative AI to identify novel therapeutic targets with pleiotropic effects beyond their original indication. The study also highlights the growing role of AI aging clocks as surrogate endpoints in clinical research, though their reliability remains debated.

Technical Details

  • Drug development pipeline: Insilico used two AI systems—one scanning health data and scientific literature for disease-relevant proteins, and another analyzing protein structure to generate matching molecules. The target protein identified was TNIK, implicated in both aging and pulmonary fibrosis.
  • Aging clock methodology: Six independent AI models, developed by separate teams at Harvard, Oxford, Beijing, and Insilico, were applied to blood protein data. These models do not share features or training data, making their consensus findings more robust.
  • Clinical trial design: The 42-patient IPF trial compared rentosertib against placebo, with blood samples collected for proteomic analysis. The optimal dose for lung function (60 mg once daily) differed from the optimal dose for aging clock reduction (30 mg twice daily), suggesting an effect partially independent of pulmonary improvement.
  • Validation approach: Treated patients' protein profiles were compared against over 55,000 UK Biobank profiles tracking age-related protein changes, showing reversal of typical aging signatures.
  • Current status: Rentosertib has advanced to Phase III trials for IPF, with at least 28 AI-designed drug candidates in Insilico's pipeline as of March 2026.

Industry Insight

  • The differentiation between optimal dosing for disease treatment versus aging marker reversal suggests AI-designed drugs may have complex, multi-dimensional efficacy profiles that require careful optimization—a consideration for clinical trial design in longevity research.
  • Major pharma investment (e.g., Eli Lilly's stake in Insilico) signals growing institutional confidence in AI-driven drug discovery, particularly for targets with broad therapeutic potential beyond single indications.
  • The use of consensus across multiple independent aging clocks addresses a key criticism in the field, but larger trials in healthy populations remain essential before aging clocks can be accepted as validated surrogate endpoints by regulatory bodies.

TL;DR

  • Insilico Medicine使用生成式AI开发的药物rentosertib在42名特发性肺纤维化(IPF)患者试验中显示可逆转生物衰老标志物
  • 6个独立AI衰老时钟一致预测治疗组患者生物学年龄降低3-6年,且该效果与肺功能改善的最佳剂量不同,暗示存在独立于肺功能的机制
  • 药物通过AI系统发现TNIK蛋白靶点并设计分子,从靶点到候选药物仅耗时18个月,目前已进入III期临床试验阶段
  • 专家肯定结果但强调样本量有限且缺乏健康人群数据,衰老时钟的可靠性仍需进一步验证

为什么值得看

这项研究首次通过多模型交叉验证展示了AI药物可逆转生物衰老标志物,为衰老干预领域提供了重要临床证据。同时验证了生成式AI在药物发现中的加速能力,从靶点发现到临床阶段仅需18个月,对制药行业具有示范意义。

技术解析

  • 多时钟验证方法:研究使用6个独立开发的AI衰老时钟(来自哈佛、牛津、北京和Insilico团队),这些模型不共享特征或训练数据,但一致预测治疗组患者蛋白质组模式更年轻,增强了结果可信度
  • 药物设计流程:Insilico使用两个AI系统——一个从健康数据和科学文献中筛选疾病相关蛋白靶点,另一个分析蛋白结构并生成匹配分子,最终锁定TNIK蛋白作为靶点
  • 蛋白质组学分析:将治疗组患者血液蛋白质与UK Biobank中55,000多个年龄相关蛋白质谱进行对比,发现rentosertib逆转了典型的衰老相关蛋白质变化模式
  • 剂量差异发现:肺功能改善的最佳剂量为60mg每日一次,而降低预测生物学年龄的最佳剂量为30mg每日两次,提示抗衰老效应可能独立于肺保护机制

行业启示

  • AI药物发现进入临床验证阶段:从靶点发现到III期临床试验仅18个月,验证了生成式AI可大幅压缩药物研发周期,传统药企应加速AI药物发现平台的布局
  • 多模型交叉验证成为新标准:使用多个独立衰老时钟验证结果,为生物标志物研究提供了更严谨的方法论框架,单一模型验证可能产生偏差
  • 老药新用与靶点重定位机会:TNIK同时关联肺纤维化和衰老,显示AI可发现疾病的共同分子机制,为现有药物拓展适应症提供新路径

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Healthcare AI 医疗AI Research 科学研究 LLM 大模型