基于转录组学鉴定糖尿病视网膜病变中血管生成相关的生物标志物及其影响糖尿病视网膜病变的机制  

Identification of angiogenesis-related biomarkers in diabetic retinopathy based on transcriptomics and exploration of their mechanisms affecting diabetic retinopathy

作　　者：岳凤娟[1] 冯戎 张辉 李晓东[1] YUE Fengjuan;FENG Rong;ZHANG Hui;LI Xiaodong(Medical Examination Central,Guizhou Provincial People’s Hospital,Guiyang 520002,Guizhou Province,China)

机构地区：[1]贵州省人民医院健康体检中心,贵州省贵阳市520002

出　　处：《眼科新进展》2024年第12期972-980,共9页Recent Advances in Ophthalmology

基　　金：贵州省科技计划项目[编号:黔科合成果-LC(2023)028];贵州省卫生健康委科学技术基金项目(编号:gzwkj2023-228)。

摘　　要：目的基于转录组学挖掘糖尿病视网膜病变(DR)中与血管生成基因(ARGs)相关的生物标志物。方法通过差异表达分析筛选出DR患者群体与健康对照群体间的差异表达基因(DEGs)。运用加权基因共表达网络分析识别出与ARGs最显著相关的关键模块。将DEGs与此关键模块中的基因做交集处理,得到了一系列的候选基因。其次,操纵3款不同的机器学习算法从这些候选基因中挑选出有潜力的DR生物标志物并基于这些标志物构建出用于DR诊断的列线图模型,通过绘制并分析接受者操作特征曲线(ROC),校准曲线以及决策曲线分析曲线,进一步验证列线图模型预测的准确性。此外,研究DR组患者和对照组之间的免疫细胞浸润情况,同时,预测针对DR治疗的潜在靶向生物标志物的化学药物。结果基于生物信息学分析,共鉴定出153个DEGs和969个关键模块基因。以上两种结果取交集获得19个候选基因。3种机器学习算法确定了4个生物标志物[血清淀粉样蛋白A2(SAA2)、内质网驻留蛋白27(ERP27)、谷胱甘肽过氧化物酶8(GPX8)和RAB蛋白家族成员RAB7B],ROC曲线证实此4个生物标志物具有有效区分DR患者和正常人的能力。根据此4种生物标志物构建的列线图模型,能够有效预测DR的患病概率。免疫细胞浸润结果表明,4种生物标志物与28种免疫细胞之间均展现出较强的相关性。药物预测发现,分别靶向SAA2和GPX8的地塞米松、二硫化谷胱甘肽和谷胱甘肽可能是治疗DR的潜在药物。结论SAA2、ERP27、GPX8和RAB7B可建立诊断DR的可靠模型,这些生物标志物主要参与补体和凝血级联反应,以及调控免疫细胞浸润等关键生物学过程。Objective To explore the biomarkers associated with angiogenesis-related genes(ARGs)in diabetic retinopathy(DR)based on transcriptomics.Methods The differentially expressed genes(DEGs)between the DR patients(DR group)and the healthy population(control group)were identified through differential expression analysis.Weighted gene co-expression network analysis(WGCNA)was used to obtain the key module genes most significantly related to ARGs.Candidate genes were determined by taking the intersection of DEGs and these key module genes obtained.Then,three machine learning algorithms were used to screen potential DR biomarkers from these candidate genes,a nomogram model used for DR diagnosis was constructed based on these biomarkers,and the accuracy of nomogram model prediction was further verified by drawing and analyzing receiver operating characteristic(ROC)curves,calibration curves and decision curve analysis curves.Furthermore,the immune cell infiltration in the DR group and the control group was compared,and the potential biomarker-targeted chemotherapeutics for DR treatment was predicted.Results Based on bioinformatics analysis,a total of 153 DEGs and 969 key module genes were identified.Based on the intersection of the above two results,19 candidate genes were determined.Three machine learning algorithms identified four biomarkers[serum amyloid A2(SAA2),endoplasmic reticulum protein 27(ERP27),glutathione peroxidase 8(GPX8),and RAB protein family member (RAB7B)],and the ROC curves confirmed that these four biomarkers have the ability to effectively distinguish DR patients from healthy people.The nomogram model constructed based on these four biomarkers could effectively predict the probability of DR.Immune cell infiltration showed that these four biomarkers were significantly correlated with 28 immune cells.Drug prediction found that dexamethasone,glutathione disulfide and glutathione,which target SAA2 and GPX8,were potential drugs for DR treatment.Conclusion A reliable model for DR diagnosis can be established ba

关 键 词：糖尿病视网膜病变 血管生成 机器学习 生物标志物 免疫细胞浸润 

分 类 号：R774[医药卫生—眼科]