葡萄膜黑色素瘤（UM）是成人原发性眼内肿瘤，由眼部黑色素细胞产生，对视力和健康构成重大威胁。尽管UM很罕见，但由于其肝转移的可能性很高，导致检测后中位生存期约为一年，因此令人担忧。与皮肤黑色素瘤不同，UM 由于其肿瘤突变负荷和 PD-1/PD-L1 表达较低，对免疫检查点抑制 (ICI) 的反应较差。 Tebentafusp 是一种被批准用于转移性 UM 的双特异性 T 细胞接合剂 (TCE)，在临床试验中显示出潜力，但客观缓解率仍然较低。为了提高 TCE 功效，我们在本研究中探索了定量系统药理学 (QSP) 模型。通过将 TCE 模块集成到现有的 QSP 模型中并使用 UM 和 tebentafusp 的临床数据，我们的目的是识别和排序潜在的预测生物标志物以供患者选择。我们选择了 30 个重要的预测生物标志物，包括模型参数以及肿瘤和血液区室中的细胞浓度。我们使用不同的方法研究生物标志物，包括比较有反应者和无反应者的中位水平，以及基于截止值的生物标志物测试算法。肿瘤和血液中的 CD8 T 细胞密度、肿瘤中 CD8 T 细胞与调节性 T 细胞的比率以及血液中的幼稚 CD4 密度是已识别的关键生物标志物的示例。预测能力的量化表明单一治疗前生物标志物的预测能力有限，通过早期治疗生物标志物和预测生物标志物的组合可以改善这种能力。最终，这种 QSP 模型可以促进生物标志物指导的患者选择，提高临床试验效率和 UM 治疗结果。© 2024。作者。

Uveal melanoma (UM), the primary intraocular tumor in adults, arises from eye melanocytes and poses a significant threat to vision and health. Despite its rarity, UM is concerning due to its high potential for liver metastasis, resulting in a median survival of about a year after detection. Unlike cutaneous melanoma, UM responds poorly to immune checkpoint inhibition (ICI) due to its low tumor mutational burden and PD-1/PD-L1 expression. Tebentafusp, a bispecific T cell engager (TCE) approved for metastatic UM, showed potential in clinical trials, but the objective response rate remains modest. To enhance TCE efficacy, we explored quantitative systems pharmacology (QSP) modeling in this study. By integrating a TCE module into an existing QSP model and using clinical data on UM and tebentafusp, we aimed to identify and rank potential predictive biomarkers for patient selection. We selected 30 important predictive biomarkers, including model parameters and cell concentrations in tumor and blood compartments. We investigated biomarkers using different methods, including comparison of median levels in responders and non-responders, and a cutoff-based biomarker testing algorithm. CD8+ T cell density in the tumor and blood, CD8+ T cell to regulatory T cell ratio in the tumor, and naïve CD4+ density in the blood are examples of key biomarkers identified. Quantification of predictive power suggested a limited predictive power for single pre-treatment biomarkers, which was improved by early on-treatment biomarkers and combination of predictive biomarkers. Ultimately, this QSP model could facilitate biomarker-guided patient selection, improving clinical trial efficiency and UM treatment outcomes.© 2024. The Author(s).