基于胰腺癌糖酵解异常构建LASSO-Cox预后模型  

Constructing a LASSO-Cox prognostic model based on aberrant glycolysis in pancreatic cancer

作　　者：李文嘉 杜岩 李昕 周文策[1,2] LI Wenjia;DU Yan;LI Xin;ZHOU Wence(The Second Clinical Medical School,Lanzhou University,Lanzhou 730030,China;Department of General Surgery,The Second Hospital of Lanzhou University,Lanzhou 730030,China)

机构地区：[1]兰州大学第二临床医学院,甘肃兰州730030 [2]兰州大学第二医院普外科,甘肃兰州730030

出　　处：《兰州大学学报(医学版)》2025年第3期21-29,共9页Journal of Lanzhou University(Medical Sciences)

基　　金：国家自然科学基金资助项目(82260555);甘肃省科技重大专项资助项目(22ZD6FA021-4);甘肃省拔尖领军人才资助项目[(2023)9];甘肃省青年科技基金资助项目(24JRRA374)。

摘　　要：目的 基于糖酵解相关基因构建胰腺癌预后模型,帮助指导临床风险分层和个性化治疗。方法 基于TCGA和GTEx数据库,筛选胰腺癌上调基因与糖酵解通路基因的交集。将TCGA-PAAD数据集178例患者分为训练集和验证集,用训练集构建糖酵解基因集Cox风险模型,以LASSO回归筛选关键基因防止过拟合,并采用多变量Cox回归确定最终预后模型。通过绘制风险、生存、受试者操作特征曲线评估模型预后预测性能,结合功能富集、京都基因与基因组百科全书和基因本体论分析探究模型涉及的肿瘤生物学机制。结果 差异分析显示5 542个基因在胰腺癌组织中显著上调,相关性分析提示7 752个基因与糖酵解代谢密切相关,最终筛选出3 092个糖酵解基因集。LASSO回归分析确定12个特征基因对患者预后贡献最大,拟合得到含8个糖酵解相关基因的预后模型,按训练集中位风险评分阈值分组。风险曲线表明风险评分越高,结局事件越频繁;Kaplan-Meier生存曲线证实高风险组预后更差,受试者操作特征曲线证实模型对患者生存情况预测能力良好。功能富集分析显示高风险组肿瘤经典信号和免疫相关通路显著激活,突变分析揭示风险评分与基因突变频率有关。结论 本研究构建的糖酵解相关基因预后模型对胰腺癌患者生存情况具有良好的预测效能。Objectives A prognostic model for pancreatic cancer based on glycolysis-related genes will be con-structed to aid in clinical risk stratification and personalized treatment.Methods Based on the TCGA and GTEx databases,the intersection of the upregulated genes in pancreatic cancer and the genes in the glycolysis pathway was taken to obtain the glycolysis gene set.The 178 patients in the TCGA-PAAD dataset were divid-ed into a training set and a validation set.The former was used to construct a Cox risk model for the glycolysis gene set.LASSO regression was applied to screen genes to prevent overfitting,and a multivariate Cox regres-sion was used to determine the final prognostic model.Risk curves,survival curves,and receiver operating characteristic curves were plotted to evaluate the prognostic prediction performance of the model.Functional enrichment analysis,as well as Kyoto Encyclopedia of Genes and Genomes and Gene Ontology analyses,were conducted to explore the tumor biological mechanisms involved in the model.Results Differential anal-ysis showed that 5542 genes were significantly upregulated in pancreatic cancer tissues.The correlation analysis indicated that 7752 genes were closely related to glycolytic metabolism.By taking the intersection,a glycolytic gene set containing 3092 genes was obtained.LASSO regression analysis determined that 12 characteristic genes contributed the most to the prognosis of patients.A prognostic model containing 8 glycol-ysis-related genes was fitted,and the patients were grouped according to the median risk score in the training set.The risk curve showed that the higher the risk score,the more frequent the outcome events.The Kaplan-Meier survival curve demonstrated that the prognosis of the high-risk group was worse.The receiver operating characteristic curve confirmed that the model had a good ability in predicting the long-term survival of patients.Enrichment analysis showed that the classical tumor signaling and immune-related pathways were significantly enriched in the high

关 键 词：胰腺癌 糖酵解 代谢重编程 预后 预测模型 

分 类 号：R735.9[医药卫生—肿瘤]