AI News AI资讯 3d ago Updated 3d ago 更新于 3天前 53

Insilico Medicine advances AI drug for IPF to Phase III trials 因赛力特医学将AI药物推进至IPF三期临床试验

Insilico Medicine has advanced its AI-discovered drug, rentosertib, for idiopathic pulmonary fibrosis (IPF) into Phase III human trials, marking a significant milestone for computational drug discovery. The drug was identified using the Pharma.AI platform, specifically leveraging PandaOmics for target discovery (TNIK) and Chemistry42 for generative molecular design. Phase I results showed a mean forced vital capacity gain of +98.4 mL in the 60 mg cohort compared to a 20.3 mL loss in the placebo Insilico Medicine的AI药物Rentosertib针对特发性肺纤维化(IPF)进入III期临床试验,标志着AI药物从早期安全评估迈向晚期疗效验证的关键突破。 该药物通过Pharma.AI平台发现,靶向TNIK激酶,在II期试验中使60mg剂量组的平均用力肺活量增加98.4mL,显著优于安慰剂组。 研发流程结合了PandaOmics的多组学因果推断靶点发现和Chemistry42的生成式分子设计,将临床前候选药物确定时间缩短至18个月。 临床试验整合了多种蛋白质组学生物年龄时钟(如ProtAge, OrganAge),验证了药物具有衰老修饰活性并减少细胞外基质重塑。 完整的发现到

75
Hot 热度
70
Quality 质量
85
Impact 影响力

Analysis 深度分析

TL;DR

  • Insilico Medicine has advanced its AI-discovered drug, rentosertib, for idiopathic pulmonary fibrosis (IPF) into Phase III human trials, marking a significant milestone for computational drug discovery.
  • The drug was identified using the Pharma.AI platform, specifically leveraging PandaOmics for target discovery (TNIK) and Chemistry42 for generative molecular design.
  • Phase I results showed a mean forced vital capacity gain of +98.4 mL in the 60 mg cohort compared to a 20.3 mL loss in the placebo group, with a manageable safety profile.
  • The development utilized a biology-first, aging-informed AI workflow that integrated multi-omics data and generative tensorial reinforcement learning to create novel chemical entities.
  • This progression provides empirical validation for AI-driven pipelines, demonstrating the ability to move from target identification to late-stage clinical efficacy validation efficiently.

Why It Matters

This development serves as a critical empirical case study for the computational drug discovery sector, proving that AI-generated candidates can successfully navigate early safety evaluations and demonstrate efficacy in late-stage clinical settings. For researchers and industry professionals, it validates the integration of causal inference and generative chemistry in identifying novel therapeutic targets outside conventional pathways, such as bypassing receptor tyrosine kinases in IPF treatment.

Technical Details

  • Target Discovery (PandaOmics): Utilized causal inference mechanisms on vast biological datasets (genomics, literature, patents) to identify TNIK as a central node regulating fibrosis and inflammation via multiple signaling channels (Wnt, TGF-β, Hippo/YAP-TAZ, etc.), integrating a hallmarks-of-aging framework.
  • Generative Design (Chemistry42): Employed Generative Tensorial Reinforcement Learning (GENTRL) to build molecules physically aligned with the TNIK protein pocket, synthesizing 79 physical molecules and selecting the 55th iteration for preclinical testing, reducing the timeline to 18 months.
  • Clinical Trial Data: A randomized trial of 71 patients across 22 Chinese sites tested 30 mg and 60 mg doses over 12 weeks; the 60 mg dose yielded a +98.4 mL mean forced vital capacity gain versus -20.3 mL in the placebo group.
  • Proteomic Validation: Deployed internal proteomic aging-clock frameworks (ProtAge, OrganAgechrono, ipfP3GPT, PAOPAC) and mortality-risk clocks (PAC, OrganAgemortality) to track biological age changes and senomorphic activity, confirming reductions in extracellular matrix remodeling.
  • Regulatory Status: Received FDA Orphan Drug Designation in February 2023, with full discovery-to-clinic progression documented in peer-reviewed publications in Nature Biotechnology and The Journal of Medicinal Chemistry.

Industry Insight

  • The successful transition of an AI-discovered drug to Phase III validates the "biology-first" AI approach, suggesting that computational pipelines can effectively identify novel targets beyond traditional screening methods, potentially expanding the druggable genome.
  • The integration of proteomic aging clocks into clinical trials offers a new paradigm for measuring treatment efficacy in age-related diseases, providing orthogonal data streams that complement standard clinical endpoints.
  • The 18-month timeline from project initiation to preclinical candidate nomination highlights the potential for AI to significantly compress drug discovery cycles, offering a competitive advantage in reducing capital-intensive trial-and-error processes.

TL;DR

  • Insilico Medicine的AI药物Rentosertib针对特发性肺纤维化(IPF)进入III期临床试验,标志着AI药物从早期安全评估迈向晚期疗效验证的关键突破。
  • 该药物通过Pharma.AI平台发现,靶向TNIK激酶,在II期试验中使60mg剂量组的平均用力肺活量增加98.4mL,显著优于安慰剂组。
  • 研发流程结合了PandaOmics的多组学因果推断靶点发现和Chemistry42的生成式分子设计,将临床前候选药物确定时间缩短至18个月。
  • 临床试验整合了多种蛋白质组学生物年龄时钟(如ProtAge, OrganAge),验证了药物具有衰老修饰活性并减少细胞外基质重塑。
  • 完整的发现到临床进展已在《Nature Biotechnology》等期刊发表,为AI制药提供了可复现、同行评审的实证案例。

为什么值得看

本文展示了AI在药物发现全链条中的实际落地能力,特别是从靶点发现到分子生成的端到端闭环,为行业提供了宝贵的实证数据。它揭示了多组学与生成式化学结合如何加速罕见病药物的研发进程,并对传统制药模式构成挑战。对于关注AI制药商业化路径和技术可行性的从业者而言,这是理解计算生物学如何转化为临床价值的重要参考。

技术解析

  • 靶点发现引擎 (PandaOmics):利用基因组、临床试验结果、文献和专利数据构建生物网络模型,应用因果推断机制识别隐藏的疾病关联。成功绕过传统的受体酪氨酸激酶通路,锁定TNIK作为IPF干预的核心靶点,并结合“衰老特征”框架进行评分。
  • 分子生成引擎 (Chemistry42):采用生成张量强化学习(GENTRL方法)而非传统高通量筛选,直接构建与靶蛋白口袋物理对齐的新分子。在生成79个分子后选定第55代迭代进入临床前测试,极大提高了研发效率。
  • 蛋白质组学验证体系:在临床评估中部署内部蛋白质组学生物年龄时钟(包括ProtAge, OrganAgechrono, ipfP3GPT, PAOPAC等),追踪干预后的生物学年龄变化,并结合UK Biobank数据进行对比,以量化药物的抗衰老和衰老修饰效应。
  • 临床数据结构:II期随机试验包含71名患者,分为安慰剂和活性治疗组(30mg或60mg每日一次),观察期为12周。结果显示高剂量组肺功能显著改善,且安全性良好,不良事件发生率与基线预期一致。

行业启示

  • AI制药需建立标准化验证体系:Insilico通过发表详细的技术路线和临床数据,证明了AI药物发现的透明度和可重复性,这将成为未来监管审批和行业信任建立的关键标准。
  • 多组学与衰老科学的融合是新方向:将生物学靶点选择与“衰老特征”框架及蛋白质组学生物年龄时钟结合,不仅提升了靶点选择的科学性,也为评估药物对复杂退行性疾病的长期影响提供了新维度。
  • 研发周期的大幅压缩具备商业竞争力:从项目启动到临床前候选药物仅需18个月,这种速度优势使得AI制药在应对罕见病或未被满足的临床需求时,相比传统制药模式具有显著的资本和时间效率优势。

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 Product Launch 产品发布