基于生物信息学和机器学习算法鉴定PDR铁死亡关键基因

Identification of Ferroptosis-Related Hub Genes in Proliferative Diabetic Retinopathy by Integrating Bioinformatics and Machine Learning

目的:联合生物信息学分析方法和机器学习算法筛选增殖性糖尿病视网膜病变(PDR)中铁死亡关键基因和治疗靶点。方法:从GEO数据库和FerrDb数据库中下载PDR转录组测序数据集、基因芯片数据集和铁死亡相关基因列表;筛选出铁死亡相关差异基因并行功能富集分析;通过WGCNA分析方法筛选PDR中铁死亡相关疾病特征基因;综合LASSO算法与SVM-RFE算法筛选出铁死亡关键基因并通过受试者工作特征曲线评估诊断能力;对铁死亡关键基因行GSEA分析;通过DGIdb数据库构建铁死亡关键基因的药物调控网络;应用CIBERSORT算法分析PDR中免疫细胞浸润分布特征。组间比较采用独立样本t检验或Mann-Whitney U检验。Pearson相关分析用于不同基因模块与临床性状之间的相关性分析。结果:从3678个差异基因中筛选出83个铁死亡相关差异基因。功能富集分析结果显示,这些基因参与铁死亡、长寿调节、自噬等信号通路。通过WGCNA分析方法从中筛选出17个铁死亡相关疾病特征基因。联合LASSO算法与SVM-RFE算法从17个疾病特征基因中筛选出3个铁死亡关键基因为PPARA、ABCC5、TSC1,并证明这些基因有很高的诊断效能。单基因GSEA分别展示了3个关键基因高低表达组中富集到的信号通路。从DGIdb数据库中筛选出20个相关药物并构建3个关键基因的药物调控网络。CIBERSORT结果显示,M1巨噬细胞、中性粒细胞等在PDR样本中浸润程度较高。结论:本研究分析鉴定出PDR中3个铁死亡关键基因,为进一步研究PDR的病理分子机制和诊断治疗靶点提供了新的思路与参考。

Objective:This study aimed to identify ferroptosis-related hub genes in proliferative diabetic retinopathy(PDR)by integrating bioinformatics analysis method and machine learning alorithms.Methods:We downloaded RNA-seq dataset of PDR and gene lists related to ferroptosis from GEO and FerrDb databases to search for differentially expressed ferroptosis-related genes(DEFRGs)in PDR.Gene ontology analysis and Kyoto encyclopedia of genes and genomes analysis were further performed for DEFRGs in PDR.Weighted gene co-expression network analysis(WGCNA)was utilized to identify PDR-related feature genes from normalized RNA-seq dataset.Ferroptosis-related PDR feature genes were intersected genes between DEFRGs and PDR-related feature genes.Then we combined LASSO algorithm and SVM-RFE algorithm to screen ferroptosis-related hub genes which were tested the diagnostic ability by receiver operating characteristic(ROC)curves.Gene set enrichment analysis(GSEA)was used to explore potential signal pathways for each hub gene in PDR group.Through DGIdb database,we retrieved relevant drugs and constructed a gene-drug interaction network.Finally,CIBERSORT analysis was applied to detect immune cell infiltration characteristics between PDR samples and normal retina samples.The differences between continuous variables with a normal distribution were examined with Student's t-test or Mann-Whitney U test.Pearson correlation analysis was used to calculate correlations between different gene modules and clinical phenotypic.Results:We found 83 DEFRGs from 3678 differentially expressed genes and functional enrichment analysis showed that these genes were involved in ferroptosis,longevity regulating pathway and autophagy pathway significantly.Based on WGCNA,17 PDR-related feature genes were screened.Combining LASSO algorithm and SVM-RFE algorithm,we identified 3 hub genes(PPARA,ABCC5 and TSC1)from 17 ferroptosis-related PDR feature genes.ROC curves proved that 3 hub genes have reliable diagnostic ability.Outcomes in GSEA showed that fructose and mannose metabolism and proximal tubule bicarbonate reclamation pathways were significantly enriched in the subgroup with high expression level of PPARA and ABCC5 in PDR.From DGIdb database,we retrieved 20 relevant drugs and constructed a gene-drug interaction network.CIBERSORT analysis showed that M1 macrophages,neutrophils infiltrated significantly in PDR samples.Conclusion:This study identified 3 ferroptosis-related hub genes in PDR,which provides new directions for the study of molecular mechanism and clinical therapies of PDR.