Table_2_Machine learning-based identification of tumor-infiltrating immune cell-associated model with appealing implications in improving prognosis and immunotherapy response in bladder cancer patients.xlsx

BackgroundImmune cells are crucial components of the tumor microenvironment (TME) and regulate cancer cell development. Nevertheless, the clinical implications of immune cell infiltration-related mRNAs for bladder cancer (BCa) are still unclear.
MethodsA 10-fold cross-validation framework with 101 combinations of 10 machine-learning algorithms was employed to develop a consensus immune cell infiltration-related signature (IRS). The predictive performance of IRS in terms of prognosis and immunotherapy was comprehensively evaluated.
ResultsThe IRS demonstrated high accuracy and stable performance in prognosis prediction across multiple datasets including TCGA-BLCA, eight independent GEO datasets, our in-house cohort (PUMCH_Uro), and thirteen immune checkpoint inhibitors (ICIs) cohorts. Additionally, IRS was superior to traditional clinicopathological features (e.g., stage and grade) and 94 published signatures. Furthermore, IRS was an independent risk factor for overall survival in TCGA-BLCA and several GEO datasets, and for recurrence-free survival in PUMCH_Uro. In the PUMCH_Uro cohort, patients in the high-IRS group were characterized by upregulated CD8A and PD-L1 and TME of inflamed and immunosuppressive phenotypes. As predicted, these patients should benefit from ICI therapy and chemotherapy. Furthermore, in the ICI cohorts, the high-IRS group was related to a favorable prognosis and responders have dramatically higher IRS compared to non-responders.
ConclusionsGenerally, these indicators suggested the promising application of IRS in urological practices for the early identification of high-risk patients and potential candidates for ICI application to prolong the survival of individual BCa patients.

背景 免疫细胞是肿瘤微环境（tumor microenvironment, TME）的关键组成成分，可调控癌细胞的发生发展。然而，与免疫细胞浸润相关的mRNA在膀胱癌（bladder cancer, BCa）中的临床意义仍有待阐明。
方法 本研究采用包含10种机器学习算法、共计101种组合的10折交叉验证框架，构建了共识性免疫细胞浸润相关特征（immune cell infiltration-related signature, IRS）。随后全面评估了IRS在预后预测与免疫治疗响应预测中的预测性能。
结果 IRS在多队列中展现出高精度且稳定的预后预测能力，涵盖TCGA-BLCA数据集、8个独立GEO数据集、本研究自建队列（PUMCH_Uro）以及13个免疫检查点抑制剂（immune checkpoint inhibitors, ICIs）队列。相较于传统临床病理特征（如分期与分级）及94项已发表的特征标签，IRS的预测性能更优。进一步分析证实，IRS是TCGA-BLCA队列与部分GEO队列中总生存期的独立风险因素，同时也是PUMCH_Uro队列中无复发生存期的独立风险因素。在PUMCH_Uro队列中，高IRS评分组患者的CD8A与PD-L1表达水平上调，其肿瘤微环境呈现炎性免疫抑制表型。如预期所示，该组患者可从免疫检查点抑制剂治疗与化疗中获益。此外，在ICI队列中，高IRS评分组患者预后更佳，且免疫治疗响应者的IRS评分显著高于无响应者。
结论 综上，本研究结果提示IRS在泌尿外科临床实践中具备良好的应用前景，可用于早期识别高危膀胱癌患者，以及筛选适合免疫检查点抑制剂治疗的潜在人群，进而延长膀胱癌患者的总体生存时间。