基于生物信息学方法对颅内动脉瘤破裂铁死亡相关特征基因的综合分析

Comprehensive analysis on characteristic genes related to ferroptosis in ruptured intracranial aneurysms based on bioinformatics method

目的通过生物信息学方法综合分析颅内动脉瘤(IA)破裂特征基因,构建IA破裂风险的基因预测模型及药物调控靶点预测。方法自美国国家生物信息中心基因表达综合(GEO)数据库中下载GSE13353和GSE122897数据集,分别用于筛选IA破裂的特征基因和验证。对IA破裂铁死亡相关的差异基因进行基因本体(GO)富集分析和京都基因与基因组百科全书(KEGG)富集分析;通过LASSO回归和机器学习-支持向量机递归特征消除算法筛选出IA破裂的特征基因;通过受试者工作特征(ROC)曲线,并计算曲线下面积(AUC),评估特征基因模型对IA破裂预测的准确性;基因集富集分析(GSEA)量化特征基因和相关通路活性;对IA组织中免疫细胞的浸润情况以及与特征基因相关性进行分析;使用药物基因相互作用数据库(DGIDB)和Cytoscape软件构建IA破裂特征基因的药物调控网络。结果自GEO数据库中下载了GSE13353数据集,其中未破裂IA患者8例,破裂IA患者11例。从FerrDb数据库下载铁死亡相关基因731个,去除非人类种属335个,最终获得铁死亡相关基因396个,对其进行差异分析后得到90个差异基因(P<0.05),其中58个表达上调,32个表达下调。铁死亡相关的IA破裂差异基因在氧化应激等生物学过程显著富集(P<0.05);对LASSO回归和机器学习-支持向量机递归特征消除算法筛选出特征基因取交集,最终获得G6PD、MTOR、NF2、POR、SMPD15个IA破裂特征基因,将其联合对GSE122897数据集中IA破裂预测进行ROC分析,结果显示,AUC为0.884(95%CI:0.773~0.973,P<0.05)。GSEA分析获得了特征基因相关的富集通路;免疫浸润分析发现,滤泡辅助性T细胞、自然杀伤细胞的激活、M0型巨噬细胞和肥大细胞激活在IA未破裂与IA破裂中表达的差异有统计学意义(均P<0.05)。构建特征基因-药物调控网路显示,多种药物与5个IA破裂特征基因有相互作用。结论成功构建铁死亡相关特征基因的IA破裂风险模型并显示出较好的预测性,通过构建基因-药物调控网络对患者IA破裂个体化治疗提供了新思路。本研究结果尚需大样本数据以及动物实验进一步验证。

Objective To comprehensively analyze the characteristic genes of intracranial aneurysm(IA)rupture,and to construct the prediction model for IA rupture risk and drug regulatory targets.Methods The GSE13353 and GSE122897 datasets were downloaded from the gene expression omnibus(GEO)datasets,with GSE13353 as the model construction analysis group and GSE122897 as the model testing group.The datasets information was subsequently annotated and normalized.Ferroptosis-related genes were downloaded from the FerrDb database to screen for differentially expressed genes associated with ferroptosis between ruptured and unruptured intracranial aneurysms.Gene ontology(GO)and Kyoto encyclopedia of genes and genomes(KEGG)enrichment analysis were performed for the differentially expressed genes related to ferroptosis in IA rupture.LASSO regression and machine learning-support vector machine recursive feature elimination methods were applied to screen for IA rupture characteristic genes.The accuracy of the characteristic gene model for predicting IA rupture was constructed and assessed by receiver operating characteristic(ROC)curves,and the area under the curve(AUC)was calculated.Gene set enrichment analysis(GSEA)was performed to quantify the activity of characteristic genes and related pathways.Correlation analysis was used to evaluate the infiltration of immune cells in IA tissues and its correlation with characteristic genes.Data information on characteristic gene-drug interactions associated with ferroptosis for ruptured intracranial aneurysms was obtained from the drug-gene interaction database(DGIDB).Cytoscape software was used to construct the drug regulatory network for IA rupture characteristic genes.Results GSE13353 datasets were downloaded from GEO datasets,including 8 patients with unruptured IA and 11 patients with ruptured IA.A total of 731 genes related to ferroptosis were downloaded from FerrDb database,and 335 genes related to human species were removed.Finally,396 genes related to ferroptosis were obtained.The difference in expression of 90 ferroptosis-related genes between ruptured and unruptured intracranial aneurysm tissue was statistically significant(P<0.05),with 58 genes expressed up-regulated and 32 genes expressed down-regulated.The IA rupture differentially expressed genes related to ferroptosis were significantly enriched in biological processes such as oxidative stress(P<0.05).The LASSO regression and machine learning-support vector machine recursive feature elimination were applied to screen intersection of genes,and five characteristic genes of intracranial aneurysm rupture were screened out,including G6PD,MTOR,NF2,POR and SMPD1,and a predictive model of ferroptosis-related intracranial aneurysm rupture was successfully constructed.Combined with the prediction of IA rupture in GSE122897 dataset,the ROC analysis showed that the AUC was 0.884(95%CI 0.773-0.973,P<0.05).GSEA analysis revealed the enrichment pathways associated with characteristic genes.Immuno-infiltration analysis showed that the activation of follicular T helper cells,natural killer cells,M0 macrophage and mast cell were different in IA unruptured and IA ruptured samples(all P<0.05).A drug regulatory network of characteristic genes was constructed to reveal targeted drugs that may be related to characteristic genes.Conclusions The IA rupture risk model of ferroptosis-related characteristic genes was successfully constructed and showed good predictive properties.The drug regulatory network was constructed to provide a new idea for the prediction of IA rupture in patients and individualized treatment.The results of this study need to be further verified by large sample data and animal experiments.