DataSheet_1_Identification of an Individualized Metabolism Prognostic Signature and Related Therapy Regimens in Early Stage Lung Adenocarcinoma.docx

ObjectiveThe choice of adjuvant therapy for early stage lung adenocarcinoma (LUAD) remains controversial. Identifying the metabolism characteristics leading to worse prognosis may have clinical utility in offering adjuvant therapy.
MethodsThe gene expression profiles of LUAD were collected from 22 public datasets. The patients were divided into a meta-training cohort (n = 790), meta-testing cohort (n = 716), and three independent validation cohorts (n = 345, 358, and 321). A metabolism-related gene pair index (MRGPI) was trained and validated in the cohorts. Subgroup analyses regarding tumor stage and adjuvant chemotherapy (ACT) were performed. To explore potential therapeutic targets, we performed in silico analysis of the MRGPI.
ResultsThrough machine learning, MRGPI consisting of 12 metabolism-related gene pairs was constructed. MRGPI robustly stratified patients into high- vs low-risk groups in terms of overall survival across and within subpopulations with stage I or II disease in all cohorts. Multivariable analysis confirmed that MRGPI was an independent prognostic factor. ACT could not improve prognosis in high-risk patients with stage I disease, but could improve prognosis in the high-risk patients with stage II disease. In silico analysis indicated that B3GNT3 (overexpressed in high-risk patients) and HSD17B6 (down-expressed in high-risk patients) may make synergic reaction in immune evasion by the PD-1/PD-L1 pathway. When integrated with clinical characteristics, the composite clinical and metabolism signature showed improved prognostic accuracy.
ConclusionsMRGPI could effectively predict prognosis of the patients with early stage LUAD. The patients at high risk may get survival benefit from PD-1/PD-L1 blockade (stage I) or combined with chemotherapy (stage II).

研究目的：早期肺腺癌（LUAD）的辅助治疗方案选择仍存在争议。明确与不良预后相关的代谢特征，可为临床辅助治疗决策提供应用价值。
研究方法：从22个公共数据集收集肺腺癌的基因表达谱数据。将患者划分为元训练队列（n=790）、元测试队列（n=716）以及3个独立验证队列（样本量分别为345、358、321）。在各队列中构建并验证代谢相关基因对指数（MRGPI）。开展针对肿瘤分期与辅助化疗（ACT）的亚组分析。为探索潜在治疗靶点，本研究对MRGPI进行了计算机模拟分析。
研究结果：通过机器学习方法，构建了包含12个代谢相关基因对的MRGPI。在所有队列中，无论整体人群还是I/II期亚群，MRGPI均可有效将患者分为高风险组与低风险组，两组总生存期存在显著差异。多变量分析证实，MRGPI是一项独立的预后因素。辅助化疗无法改善I期高风险患者的预后，但可提升II期高风险患者的生存获益。计算机模拟分析显示，在高风险患者中高表达的B3GNT3与低表达的HSD17B6，可能通过PD-1/PD-L1通路协同参与肿瘤免疫逃逸。将MRGPI与临床特征整合后，构建的临床-代谢联合特征的预后预测精度得到进一步提升。
研究结论：MRGPI可有效预测早期肺腺癌患者的预后。高风险患者可从PD-1/PD-L1阻断治疗（I期）或联合化疗（II期）中获得生存获益。