基于机器学习探究子宫内膜癌免疫微环境与基因表达的关联及其对预后的预测价值  

Investigation of the association between endometrial cancer immune microenvironment and gene expression based on machine learning and its predictive value for prognosis

作　　者：林海宏 郭苑莉 潘如 雷南香 曾维红[1] LIN Haihong;GUO Yuanli;PAN Ru;LEI Nanxiang;ZENG Weihong(Department of Gynaecology,Meizhou People's Hospital,Meizhou 514130,China;Department of Gynaecology,the First Affiliated Hospital of Guangdong Pharmaceutical University,Guangzhou 510699,China)

机构地区：[1]梅州市人民医院妇科,广东梅州514130 [2]广东药科大学附属第一医院妇产科,广东广州510699

出　　处：《中国医学物理学杂志》2024年第12期1568-1577,共10页Chinese Journal of Medical Physics

基　　金：广东省医学科研基金(A2022318);广东省基础与应用基础研究基金(2023A1515220157);梅州市社会发展科学计划(2022C0301003)。

摘　　要：目的:探究子宫内膜癌(Endometrial Cancer,EC)免疫微环境与基因表达的关联及其对预后的预测价值,通过生物信息学分析与机器学习技术,识别关键的免疫相关基因,构建预后模型,以期为EC的个性化治疗提供新的方向。方法:基于TCGA数据库,采用DESeq2、edgeR和limma工具筛选差异表达基因,结合ImmPort数据库筛选免疫相关基因。利用Lasso回归、单变量特征选择、Boruta和随机森林等机器学习算法进行特征基因筛选。通过单因素和多因素Cox回归分析评估基因的预后价值,并构建风险评分模型。采用CIBERSORT算法分析肿瘤免疫浸润,通过免疫组化验证关键基因表达。结果:通过3种差异分析结果与免疫相关基因的交集,确定62个差异表达的免疫基因,并使用多种机器学习模型筛选得到25个潜在生物标志物,其中被机器学习筛选为预后相关基因。单因素和多因素Cox回归分析证实,INHBE、SLURP1和TNFSF11基因与EC患者的生存期显著相关。构建的风险评分模型能够有效区分不同预后组别的生存率,且与免疫细胞浸润程度相关。免疫组化分析进一步验证这些基因在肿瘤与正常组织间的表达差异。结论:INHBE、SLURP1和TNFSF11是EC免疫微环境中关键的预后生物标志物,其表达水平与免疫细胞浸润及患者生存率密切相关,为EC的精准医疗提供理论基础。Objective To investigate the association between the immune microenvironment and gene expression in endometrial cancer(EC)and discuss its predictive value for prognosis,and to identify critical immune-related genes through bioinformatics analysis and machine learning techniques and construct a prognostic model for providing new directions for personalized treatment of EC.Methods Based on data from the Cancer Genome Atlas(TCGA)program,tools such as DESeq2,edgeR,and limma were utilized to screen for differentially expressed genes.Immune-related genes were selected by integrating data from the Immunology Database and Analysis Portal(ImmPort).Machine learning algorithms including Lasso regression,univariate feature selection,Boruta and random forest were employed to refine the selection of feature genes.Univariate and multivariate Cox regression analyses were conducted to assess the prognostic value of the genes,followed by construction of a risk score model.Additionally,tumor immune infiltration was analyzed using CIBERSORT algorithm,and key gene expressions were validated through immunohistochemistry.Results The intersection of 3 difference analysis results and immune-related genes identified 62 differentially expressed immune genes,and 25 potential biomarkers which were selected by a variety of machine learning models were considered as prognosis related genes.Univariate and multivariate Cox regression analyses confirmed that INHBE,SLURP1 and TNFSF11 genes were significantly associated with survival in EC patients.The constructed risk score model effectively distinguished the survival rate of different prognostic groups,and was related to the degree of immune cell infiltration.Immunohistochemical analysis further verified the differences in the expression of these genes between tumor and normal tissues.Conclusion INHBE,SLURP1 and TNFSF11 are key prognostic biomarkers in EC immune microenvironment,and their expression levels are closely associated with immune cell infiltration and patient survival rate,providing th

关 键 词：子宫内膜癌 机器学习 肿瘤基因组图谱 免疫基因 预后 

分 类 号：R318[医药卫生—生物医学工程] R711[医药卫生—基础医学]