LBPO.CL03 · 临床研究 · Late-Breaking
基于机制的临床试验匹配揭示卵巢癌中54%的靶点对齐缺口：量化精准可入组的卵巢癌患者
Mechanism-based trial matching reveals a 54% target alignment gap in ovarian cancer: Quantifying the precision-eligible ovarian cancer patients
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
**背景：** 2期肿瘤学试验有71%的失败率，部分原因在于将没有可靶向脆弱性的患者与具有机制对齐特征的患者一同入组，从而稀释了试验信号。当前的试验导航依赖入组资格筛选（组织学、分期），但并未量化肿瘤谱与药物作用机制之间的机制对齐程度。我们使用7维机制向量（DDR、MAPK、PI3K、VEGF、HER2、IO、外排）来表示患者肿瘤谱和试验药物机制，从而量化了这一缺口。
**方法：** 对试验药物（n=59）人工整理并标注作用机制。患者向量由体细胞突变和通路聚合计算得出。机制契合度使用幅度加权相似度 `(patient_vector · trial_vector) / ||trial_vector||` 计算，以防止在低负荷患者中出现假阳性。我们验证了：（1）使用高DDR参考患者谱进行的机制判别，以及（2）在真实卵巢癌队列（TCGA-OV，n=585）中的可匹配性流行率。
**结果：** 在DDR高的参考谱中，靶向DDR的试验（n=31）取得的平均契合度为**0.874**，而非DDR试验（n=17）为**0.038**，产生23倍的判别比（Δ=0.836）。应用于585例卵巢癌患者（TCGA-OV）时，尽管满足传统入组资格标准，**314例患者（53.7%）缺乏强机制对齐**——即“精准不可入组的多数”。仅**271例患者（46.3%）**拥有与当前试验对齐的可靶向脆弱性。两组间的生存结局无差异（HR=1.122，p=0.288），这符合预期——TCGA患者并非通过机制匹配入组，从而验证了可匹配性流行率而非治疗获益。
**结论：** 这项工作揭示了一个**关键的药物开发缺口**：54%的卵巢癌患者缺乏机制对齐的试验选择。三项紧迫行动：（1）针对服务不足的通路谱进行**靶点发现**，（2）**基于机制的入组**以避免无效试验，（3）在契合度低时进行**透明的咨询告知**。该框架可保护314例患者免于浪费时间、毒性和心理负担，同时暴露出一项系统性的管线缺陷。
查看英文原文 English abstract
**Background:** Phase 2 oncology trials fail 71% of the time, partly because patients without targetable vulnerabilities are enrolled alongside those with mechanism-aligned features, diluting trial signal. Current trial navigation relies on eligibility filters (histology, stage) but does not quantify mechanism alignment between tumor profiles and drug mechanisms of action. We quantified this gap using 7-dimensional mechanism vectors (DDR, MAPK, PI3K, VEGF, HER2, IO, efflux) to represent both patient tumor profiles and trial drug mechanisms.
**Methods:** Trial drugs (n=59) were manually curated with mechanism of action annotations. Patient vectors were computed from somatic mutations and pathway aggregation. Mechanism fit was calculated using magnitude-weighted similarity `(patient_vector · trial_vector) / ||trial_vector||` to prevent false positives in low-burden patients. We validated (1) mechanism discrimination using a high-DDR reference patient profile and (2) matchability prevalence in a real ovarian cancer cohort (TCGA-OV, n=585).
**Results:** In a DDR-high reference profile, DDR-targeting trials (n=31) achieved mean fit of **0.874** compared to **0.038** for non-DDR trials (n=17), yielding a 23-fold discrimination ratio (Δ=0.836). Applied to 585 ovarian cancer patients (TCGA-OV), **314 patients (53.7%) lacked strong mechanism alignment** despite meeting traditional eligibility criteria-a "precision-ineligible majority." Only **271 patients (46.3%)** possessed targetable vulnerabilities aligned with current trials. Survival outcomes did not differ between groups (HR=1.122, p=0.288), as expected-TCGA patients were not enrolled via mechanism matching, validating matchability prevalence but not treatment benefit.
**Conclusions:** This work reveals a **critical drug development gap**: 54% of ovarian cancer patients lack mechanism-aligned trial options. Three urgent actions: (1) **Target discovery** for underserved pathway profiles, (2) **Mechanism-based enrollment** to prevent futile trials, (3) **Transparent counseling** when fit is low. This framework protects 314 patients from wasted time, toxicity, and psychological burden while exposing a systemic pipeline deficiency.
利益披露 Disclosure
F. Kiani, None.