WGCNA联合机器学习算法筛选不明原因不孕症患者免疫相关的潜在生物学标志物  

Screening of immune related potential biomarkers of patients with unexplained infertility based on weighted gene co-expression network analysis and machine learning algorithms

作　　者：许婷 庄元[1] 曹华斌[1] XU Ting;ZHUANG Yuan;CAO Huabin(Jiangxi Maternal and Child Hospital,Jiangxi,Nanchang,330006,China)

机构地区：[1]江西省妇幼保健院,南昌330006

出　　处：《江西医药》2024年第12期1099-1103,1113,共6页Jiangxi Medical Journal

基　　金：江西省卫生健康委科技计划课题,编号202211101。

摘　　要：目的 运用加权基因共表达网络分析(WGCNA)、LASSO和SVM-RFE算法,探讨与不明原因不孕症(UI)免疫相关的潜在生物标志物,为UI诊治提供新思路。方法 从GEO数据库下载2个UI子宫内膜基因表达数据集,GSE165004作为芯片数据集,GSE16532作为独立验证数据集。从ImmPort数据库获得免疫相关基因,用R语言鉴定免疫相关的差异表达基因,通过WGCNA明确与UI相关的共表达模块,对最显著的模块进行功能富集分析。将获得的差异性免疫相关基因(IRGs)和关键模块基因进行重叠后,运用LASSO和SVM-RFE算法选出的核心特征基因,在GSE16532数据集中进行验证。此外,采用CIBERSORT算法进行UI免疫浸润分析。结果 获得差异性IRGs共51个。通过WGCNA获得与UI相关的黑色和黄色模块,富集分析发现这些模块富集在T细胞活化、白细胞粘附和增殖的生物学过程,以细胞因子-细胞因子受体及Th1/Th2细胞分化等通路信号上。将获得的IRGs和关键模块基因进行重叠后获得18个候选基因,通过LASSO及SVM-RFE算法筛选出特征基因并在GSE16532数据集进行检验,结果发现葡萄糖依赖性促胰岛素释放多肽受体(GIPR)、S100钙结合蛋白A5(S100A5)是UI的2个潜在生物标志物。与正常组相比,GIPR(P=0.045)和S100A5(P=0.022)在UI组中信使核糖核酸(m RNA)水平均显著下调,差异有统计学意义。另外,免疫浸润分析发现GIPR和S100A5与免疫细胞相关,其中GIPR与巨噬细胞M2正相关(P=0.003),与调节性T细胞负相关(P=0.001);S100A5与浆细胞正相关(P=0.049),与辅助性T细胞负相关(P=0.045),差异均有统计学意义。结论 GIPR和S100A5基因与UI免疫浸润相关,可能成为诊断与预测潜在生物标志物。Objective To explore immune related potential biomarkers of patients with unexplained infertility(UI)by using weighted gene co-expression network analysis(WGCNA),LASSO and SVM-RFE machine learning algorithm,so as to provide new ideas for the diagnosis and treatment of UI.Methods Two gene expression datasets of endometrial tissues in patients with UI were retrieved from the GEO database.165004 was used as the microarray dataset,and GSE16532 was used as the independent verification datasets.Immune-related genes were obtained from Immport database.The differentially expressed immune-related genes(IRGs)were identified by R language.WGCNA was used to identify the co-expression modules related to UI.Functional enrichment analysis was performed on the most significant module.After overlapping the obtained different immune-related genes and hub module genes,the hub feature gene selected by Lasso regression and SVM-RFE algorithm was used to verify it at the GSE16532 database.In addition,the CIBERSORT algorithm was used for UI immune infiltration analysis.Results A total of 51 differentially expressed IRGs were obtained.Through WGCNA analysis,black and yellow modules had the strongest correlation with UI.The enrichment analysis found that these modules are enriched on the biological process of T cell activation,white blood cell adhesion and proliferation,and on the path signals such as cytokine-cytokine receptors and Th1/TH2 cell differentiation.18 candidate genes were obtained after overlapping IRGs and key module genes.The feature genes screened by LASSO and SVM-RFE algorithm are tested at the GSE16532 dataset.It was found that GIPR and S100A5 were two potential biomarkers of UI.Compared with the normal group,GIPR(P=0.045)and S100A5(P=0.022)were significantly reduced in the mRNA level in the UI group,and the differences were statistically significant.In addition,immune-infiltration analysis found that GIPR and S100A5 were related to immune cells,of which GIPR was positively correlated with Macrophages M2(P=0.003),and neg

关 键 词：不明原因不孕症 加权基因共表达网络分析 机器学习 免疫相关基因 免疫浸润 

分 类 号：R711.6[医药卫生—妇产科学]