基因组学与应用生物学

本研究旨在基于多源转录组数据与机器学习算法，筛选出能准确识别阿尔茨海默病(Alzheimer′s disease, AD)的特征基因并构建竞争性内源RNA(competing endogenous RNA, ceRNA)调控网络，揭示其在AD病例进程中的表达机制。通过整合GEO数据库中AD人脑组织转录组数据(训练集：GSE37263、 GSE122063、 GSE138260;验证集：GSE5281),采用主成分分析(principal component analysis, PCA)进行聚类并筛选差异表达基因(differentially expressed genes, DEGs),联合最小绝对值收缩和选择算子(least absolute shrinkage and selection operator, LASSO)回归、支持向量机(support vector machine, SVM)与随机森林(random forest, RF)机器学习算法，筛选出AD相关特征基因，并进行GO和KEGG分析，显示其富集的功能通路。同时对特征基因CARTPT高/低表达组DEGs进行基因集富集分析(gene set enrichment analysis, GSEA)与基因集变异分析(gene set variation analysis, GSVA),得出高/低表达组中富集通路的差异。最后进行免疫细胞浸润分析，构建ceRNA调控网络。鉴定出4个AD相关特征基因(CARTPT、FCGBP、NPTX2、RBM3),多基因联合诊断模型AUC达0.945(95%CI:0.904～0.977)。其中，CARTPT在独立验证集中表现出最优的疾病识别效能，CARTPT基因调控网络与AD病理进程之间密切相关。CARTPT可能通过调节内源性大麻素信号、神经活性配体信号通路、核苷酸代谢、嘌呤代谢等，减缓AD进展。基于GSEA和GSVA的多维度解析显示：CARTPT高表达组与促炎、组织重构及NOTCH信号驱动的细胞命运调控相关；而低表达组呈现氧化磷酸化代谢亢进与帕金森病相关通路富集，导致神经代谢紊乱和能量代谢抑制。通过免疫微环境分析，发现CARTPT的表达水平与初始B细胞丰度呈显著负相关，而仅有静息肥大细胞在AD病理进程中存在差异性，CARTPT可能并非通过调控免疫微环境参与AD的病理进程。最后，通过构建ceRNA调控网络，从系统层面解析基因表达的调控机制。本研究整合多组学数据与机器学习算法，筛选出4个AD潜在关联基因(CARTPT、FCGBP、NPTX2、RBM3),构建的多基因诊断模型在训练集中显示优异性能(AUC=0.945)。其中CARTPT在独立验证集(AUC=0.793)中展现诊断潜力，其参与的神经递质与代谢通路(如内源性大麻素信号、嘌呤代谢)为探索AD机制提供有价值的参考。

This study aims to screen signature genes capable of accurately identifying Alzheimer′s disease(AD) and construct a competing endogenous RNA(ceRNA) regulatory network based on multi-source transcriptomic data and machine learning algorithms, thereby revealing their expression mechanisms during AD disease progression. By integrating AD human brain tissue transcriptomic data from the GEO database(training sets: GSE37263, GSE122063, GSE138260; validation set: GSE5281), principal component analysis(PCA) was used for clustering and screening differentially expressed genes(DEGs), least absolute shrinkage and selection operator(LASSO) regression, support vector machine(SVM), and random forest(RF) machine learning algorithms were combined to screen AD-related feature genes. Gene ontology(GO) and Kyoto encyclopedia of genes and genomes(KEGG) analyses revealed enriched functional pathways. Gene set enrichment analysis(GSEA) and gene set variation analysis(GSVA) of DEGs in high/low CARTPT expression groups identified differentially enriched pathways. Immune cell infiltration analysis was performed, and a ceRNA regulatory network was constructed. Four AD-related feature genes(CARTPT, FCGBP, NPTX2, RBM3) were identified, with the multigene diagnostic model achieving an AUC of 0.945(95% CI: 0.904～0.977). CARTPT demonstrated optimal disease-discrimination perfor-mance in the independent validation set, and its regulatory network was closely associated with AD pathology. CARTPT may alleviate AD progression by regulating endocannabinoid signaling, neuroactive ligand-receptor interactions, nucleotide metabolism, and purine metabolism. Multidimensional GSEA and GSVA analyses showed: high CARTPT expression correlated with pro-inflammatory responses, tissue remodeling, and NOTCH signaling-driven cell fate regulation, while low expression exhibited hyperactive oxidative phosphorylation, Parkinson′s disease-related pathway enrichment, neurometabolic dysregulation, and suppressed energy metabolism. Immune microenvironment analysis revealed a significant negative correlation between CARTPT expression and na6ve B-cell abundance. Only resting mast cells showed differential infiltration in AD pathology, suggesting CARTPT may not participate in AD pathogenesis via immune microenvironment regulation. Finally, a ceRNA regulatory network was built to systematically analyze gene expression regulation mechanisms. This study integrates multi-omics data with machine learning algorithms to identify four genes potentially associated with Alzheimer′s disease(CARTPT, FCGBP, NPTX2, RBM3). The constructed polygenic diagnostic model demonstrated excellent performance in the training set(AUC=0.945). Among these, CARTPT demonstrated potential diagnostic utility in an independent validation set(AUC=0.793). Its involvement in neurotransmitter and metabolic pathways(e.g., endocannabinoid signaling, purine metabolism) provides valuable insights for exploring the mechanisms of AD.