基于生物信息学分析筛选重性抑郁障碍关键基因及构建诊断模型  

Identification of key genes and construction of diagnostic model of major depressive disorder using bioinfor⁃matic analysis

作　　者：刘丹 张敏[1] 和申 彭代辉 肖泽萍[1] LIU Dan;ZHANG Min;HE Shen;PENG Daihui;XIAO Zeping(Shanghai Mental Health Center,Shang-hai Jiao Tong University School of Medicine,Shanghai 200030,China)

机构地区：[1]上海交通大学医学院附属精神卫生中心,上海200030

出　　处：《中国神经精神疾病杂志》2023年第8期468-474,共7页Chinese Journal of Nervous and Mental Diseases

基　　金：上海市科委项目(编号:YDZX20213100001003);上海市精神卫生中心院级课题(编号:2022zd02,2020-YJ11)。

摘　　要：目的通过整合分析外周血转录组数据探究重性抑郁障碍(major depressive disorder,MDD)关键基因并创建诊断模型。方法检索基因表达汇编(Gene Expression Omnibus,GEO)公共数据库得到5个MDD外周血相关数据集。使用R limma包及稳健排序聚合(robust rank aggregation,RRA)算法筛选出差异表达基因。以包含最大样本量的GSE98793为训练集,使用Boruta算法进行关键基因筛选,使用logistic回归分析关键基因表达水平与抑郁症的关系。使用Bootstrap法进行内部验证,将剩余4个数据集作为外部验证集,使用受试者工作特征(receiver operating characteristic,ROC)曲线评估诊断模型的诊断性能。结果分析共得到31个差异表达基因,其中上调基因20个,下调基因11个,从中筛选出7个基因为关键基因,分别为MMP8、TDRD9、FAM3B、LCN2、ARG1、NPTN和FANCF。将7个基因纳入多因素logistic回归分析构建诊断模型,绘制ROC曲线,曲线下面积(area under curve,AUC)为0.803(95%CI:0.740~0.867),说明该模型在训练集具有较好的预测能力。Bootstrap重抽样法内部验证结果显示AUC为0.804(95%CI:0.757~0.851),模型的校准曲线显示一致性良好。同时,在4个外部验证数据集中,该模型也表现出较好的诊断性能,AUC值分别为0.781(GSE76826)、0.901(GSE38206)、0.722(GSE39653)、0.725(GSE52790)。结论本文通过对现有MDD外周血转录组数据进行整合分析,筛选出7个MDD关键基因并构建出具有较好诊断能力的诊断模型,为基于生物标志物的MDD诊断提供了依据。Objective To identify key genes related to major depressive disorder(MDD)and develop a diagnostic model by analyzing peripheral blood transcriptome data.Methods We identified five MDD-related datasets meeting our criteria from the Gene Expression Omnibus(GEO)database.Differential gene expression analysis was conducted using the R limma package and RRA algorithm.The dataset with the largest sample size,GSE98793,was used as the training set,and a logistic regression model was developed for predicting MDD after feature selection with the Boruta algorithm.The model was internally validated using Bootstrap,while the remaining four datasets were used as external validation sets.The model's diagnostic performance was assessed using receiver operating characteristic(ROC)curves.Results We identified 31 differentially expressed genes,among which 20 are upregulated and 11 are downregulated.From these,we selected 7 genes as key genes,namely MMP8,TDRD9,FAM3B,LCN2,ARG1,NPTN,and FANCF.We included these 7 genes in a multiple logistic regression analysis to construct a diagnostic model and plotted a ROC curve.The area under curve(AUC)was 0.803(95%CI:0.740~0.867),indicating the model's good predictive ability in the training set.Internal validation using bootstrap resampling showed an AUC of 0.804(95%CI:0.757~0.851),and the calibration curve of the model demonstrated good consistency.The diagnostic performance of the model was also good in the external validation datasets.The AUC values were 0.781(GSE76826),0.901(GSE38206),0.722(GSE39653),and 0.725(GSE52790),respectively.Conclusion This study integrated and analyzed public transcriptome data on MDD to identify 7 key genes related to MDD and construct a diagnostic model.The model showed good diagnostic ability in both internal and external validation datasets.These findings provide a basis for further understanding the potential pathological mechanisms of MDD and developing biomarker-based diagnostic methods.

关 键 词：重性抑郁障碍 生物信息学 生物标志物 高通量测序 诊断 外周血 受试者工作特征曲线 

分 类 号：R749.4[医药卫生—神经病学与精神病学]