基于氧化应激与乳酸代谢相关基因的胰腺癌预后预测模型的构建与验证

Construction and validation of a prognostic model for pancreatic cancer based on oxidative stress and lactate metabolism-related genes

摘要:

背景与目的 胰腺癌是一种恶性程度极高、预后极差的消化系统肿瘤，其复杂的肿瘤微环境在疾病进展中发挥重要作用。氧化应激与乳酸代谢是影响肿瘤发生、发展及免疫逃逸的关键生物过程，二者在胰腺癌预后中的协同效应尚不明确。本研究旨在整合氧化应激和乳酸代谢相关基因，构建并验证一个具有临床价值的胰腺癌预后模型，为揭示其潜在的分子机制提供依据。方法 从肿瘤基因组图谱数据库获取177例胰腺癌患者的转录组测序数据和临床资料，并结合基因表达全景图谱（GEO）数据库（GSE57495）进行外部验证。通过“limma”R包筛选胰腺癌中氧化应激与乳酸代谢相关的差异表达基因，采用单变量Cox回归分析筛选预后相关基因，进一步经LASSO回归建立多基因风险评分模型。利用Kaplan-Meier生存曲线、受试者工作特征（ROC）曲线、一致性指数（C指数）、列线图及决策曲线评估模型的预测能力和临床应用价值。最后，运用CIBERSORT算法和ssGSEA分析不同风险组间免疫细胞浸润差异及免疫状态特征。结果 最终筛选出6个关键基因（MUC1、KRT18、SDC1、AREG、DDC、ATPAF2）构建胰腺癌预后模型。根据模型计算的风险评分，患者被划分为高、低风险组。Kaplan-Meier分析显示高风险组患者总生存率明显低于低风险组（P<0.01）。ROC曲线分析显示该模型在1、2、3年预测中的曲线下面积分别为0.710、0.674、0.649，C指数和列线图验证均表明模型具有良好的区分性与校准性。多因素Cox分析提示风险评分是胰腺癌的独立预后因素。外部GEO数据集验证了模型的稳健性。免疫浸润分析显示，高风险组中M0型巨噬细胞比例明显增加，而效应T细胞及CD8⁺ T细胞浸润减少，提示免疫抑制性微环境可能与预后不良相关。结论 本研究构建了一个基于氧化应激与乳酸代谢相关基因的胰腺癌预后预测模型，能够独立预测患者生存，并揭示不同风险组的免疫微环境差异。该模型为胰腺癌的预后评估、分层管理及个体化治疗策略的制定提供了新的分子依据与潜在靶点。

Abstract:

Background and Aims Pancreatic cancer is a highly malignant digestive system tumor characterized by poor prognosis and limited therapeutic response. The tumor microenvironment plays a crucial role in its progression, where oxidative stress and lactate metabolism are two tightly interconnected processes influencing tumor growth, immune escape, and therapeutic resistance. However, their combined prognostic impact remains poorly understood. This study aimed to integrate oxidative stress– and lactate metabolism-related genes to establish and validate a robust prognostic model for pancreatic cancer, and to explore its association with immune microenvironment characteristics.Methods Transcriptomic and clinical data of 177 pancreatic cancer patients were obtained from TCGA database and an external validation was performed using the GEO dataset (GSE57495). Differentially expressed genes associated with oxidative stress and lactate metabolism were identified using the "limma" package. Univariate Cox regression was used to screen prognostic genes, followed by LASSO regression to construct a multi-gene risk model. Model performance was evaluated by Kaplan-Meier survival analysis, receiver operating characteristic (ROC) curves, concordance index (C-index), nomogram calibration, and decision curve analysis (DCA). The CIBERSORT and ssGSEA algorithms were used to analyze immune cell infiltration and immune functional differences between risk groups.Results A six-gene signature (MUC1, KRT18, SDC1, AREG, DDC, and ATPAF2) was identified to construct the prognostic model. Based on the calculated risk score, patients were stratified into high- and low-risk groups. Kaplan-Meier analysis revealed significantly worse overall survival in the high-risk group (P<0.01). The model showed good predictive accuracy with 1-, 2-, and 3-year AUCs of 0.710, 0.674, and 0.649, respectively. The C-index and calibration curves confirmed its reliability, and multivariate Cox regression indicated that the risk score was an independent prognostic factor. External validation using GEO data demonstrated consistent predictive performance. Immune infiltration analysis revealed that M0 macrophages were markedly enriched in the high-risk group, while cytotoxic and effector T-cell populations were reduced, suggesting that an immunosuppressive microenvironment may contribute to poor outcomes.Conclusion This study developed and validated a novel prognostic model for pancreatic cancer based on oxidative stress and lactate metabolism-related genes. The model accurately predicts patient survival, reflects immune microenvironment heterogeneity, and provides new molecular insights for risk stratification and individualized therapeutic strategies in pancreatic cancer management.