基于生物信息学和机器学习算法的非酒精性脂肪性肝炎细胞衰老关键基因的分析与验证

Analysis and Validation of Key Cellular Senescence Genes in Non-Alcoholic Steatohepatitis Based on Bioinformatics and Machine Learning Algorithm

摘要: 目的：非酒精性脂肪性肝炎(Non-alcoholic Steatohepatitis, NASH)是发病机制涉及多重病理生理过程，其中细胞衰老对NASH的发生发展起着关键作用。基于此，针对细胞衰老相关基因及其调控网络的深入研究，可能为NASH的靶向治疗提供新的干预策略。方法：从GEO数据库获取NASH相关转录组测序数据集GSE89632和GSE37031。采用WGCNA筛选出与NASH高度相关的基因模块，对其进行富集分析，以阐明其潜在生物学功能。将差异表达基因(Differentially expressed genes, DEGs)与CellAge数据库中的细胞衰老基因集取交集，筛选出细胞衰老DEGs。对细胞衰老DEGs进行功能富集分析，系统揭示其在NASH中的分子调控网络。随后，整合机器学习算法筛选Hub基因。最后，构建ROC曲线并计算AUC评估Hub基因的诊断效能。结果：WGCNA分析表明，共有10个基因模块被识别，其中黑色模块基因与NASH呈现最显著相关。富集分析显示，关键模块的基因与细胞衰老和免疫炎症密切相关。对训练基因集进行差异分析，共获得细胞衰老相关DEGs共51个。进一步GO和KEGG分析显示，细胞衰老DEGs主要与衰老和免疫炎症相关。对细胞衰老DEGs进行三种机器学习算法，综合对比后，FOS、MYC和PIM1被确定为关键基因。通过绘制ROC曲线检验在训练集和测试集中的诊断效能。关键基因的ROC曲线下面积均大于0.7且这三个基因较对照组均显著低表达。结论：本研究通过整合生物信息学分析和机器学习算法，最终鉴定出FOS、MYC和PIM1三个关键调控基因，均在NASH患者中显著低表达，且与免疫炎症反应密切相关。

Abstract: Objective: Non-alcoholic steatohepatitis (NASH) is a pathogenesis involving multiple pathophysiological processes, in which cellular senescence plays a key role in the development of NASH. Based on this, in-depth studies of cellular senescence-related genes and their regulatory networks may provide new intervention strategies for targeted therapy of NASH. Methods: NASH-related transcriptome sequencing datasets GSE89632 and GSE37031 were obtained from the GEO database, and WGCNA was used to screen out the gene modules that were highly related to NASH, and they were enriched and analyzed to elucidate their potential biological functions. Differentially expressed genes (DEGs) were intersected with the cellular senescence gene sets in the CellAge database to screen the cellular senescence DEGs. Functional enrichment analysis was performed on the cellular senescence DEGs to systematically reveal their molecular regulatory networks in NASH. Subsequently, machine learning algorithms were integrated to screen Hub genes. Finally, ROC curves were constructed and AUC was calculated to evaluate the diagnostic efficacy of Hub genes. Results: WGCNA analysis showed that a total of 10 gene modules were identified, with the black module genes showing the most significant association with NASH. Enrichment analysis showed that the key module genes were closely associated with cellular senescence and immune inflammation. Differential analysis of the training gene set yielded a total of 51 cellular senescence-related DEGs. Further GO and KEGG analysis showed that cellular senescence DEGs were mainly associated with senescence and immune inflammation. Three machine learning algorithms were performed on the cellular senescence DEGs, and FOS, MYC and PIM1 were identified as key genes after comprehensive comparison. Diagnostic efficacy in the training and test sets was examined by plotting ROC curves. The area under the ROC curve for the key genes was greater than 0.7 and all three genes were significantly under expressed compared to the control. Conclusion: By integrating bioinformatics analysis and machine learning algorithms, the present study finally identified three key regulatory genes, FOS, MYC and PIM1, which were significantly under expressed in NASH patients and closely associated with immune-inflammatory responses.