基于生信分析与机器学习筛选脊髓损伤铁死亡——炎症相关外周血生物标志物

Screening of ferroptosis-inflammation associated peripheral blood biomarkers in spinal cord injury based on bioinformatics analysis and machine learning

中文关键词:  脊髓损伤  生信分析  机器学习  铁死亡  炎症  生物标志物

英文关键词:Spinal cord injury  Bioinformatics analysis  Machine learning  Ferroptosis  Inflammation  Biomarkers

中文摘要:

  【摘要】 目的：筛选与铁死亡和炎症协同作用的外周血标志物，为后续深入解析其分子互作机制提供新的线索与靶点。方法：从基因表达综合数据库（Gene Expression Omnibus，GEO）获取数据集，其中训练集GSE151371包含38例脊髓损伤（spinal cord injury，SCI）与10例健康对照；验证集GSE45006包含4例脊髓损伤与4例健康对照。整合FerrDB数据库铁死亡基因集（471个）和GeneCards数据库炎症基因集（2389个），通过DESeq2软件筛选差异表达基因。对铁死亡-炎症相关差异基因进行基因本体（gene ontology，GO）功能和京都基因与基因组百科全书（Kyoto Encyclopedia of Genes and Genomes，KEGG）通路富集分析、基因集富集分析（gene set enrichment analysis，GSEA），通过构建蛋白相互作用（protein-protein interaction，PPI）网络筛选枢纽基因，再结合三种机器学习算法进一步筛选枢纽基因，并通过受试者工作特征（receiver operating characteristic，ROC）曲线下面积（area under the curve，AUC）评估其诊断效能。结果：共鉴定59个铁死亡-炎症相关差异基因（40个上调，19个下调），显著富集于铁死亡通路和Toll样受体（toll-like receptor，TLR）信号通路；经PPI网络分析初步识别得到8个枢纽基因（HIF1A、JUN、TP53、STAT3、IL1B、PTEN、MAPK3、IFNG）；通过最小绝对收缩和选择算子（least absolute shrinkage and selection operator，LASSO）回归、支持向量机（support vector machine，SVM）和随机森林（random forest，RF）三种机器学习算法筛选验证，确定肿瘤蛋白p53（tumor protein p53，TP53）、干扰素γ（interferon-gamma，IFNG）和白细胞介素1β（interleukin 1 beta，IL1B）为核心标志物，AUC显示其诊断效能显著（TP53，AUC=0.945；IFNG，AUC=0.883；IL1B，AUC=0.821）。结论：本研究筛选出了脊髓损伤外周血中与铁死亡和炎症交互作用密切相关的TP53、IFNG和IL1B三个关键基因，为深入解析SCI的病理机制提供了关键分子靶点与新的研究方向。

英文摘要:

  【Abstract】 Objectives: To screen peripheral blood biomarkers associated with the synergistic effects of ferroptosis and inflammation, and to provide novel clues and potential targets for subsequent in-depth elucidation of their underlying molecular crosstalk. Methods: The training dataset GSE151371[38 cases of spinal cord injury(SCI) vs 10 healthy controls] and validation dataset GSE45006(4 cases of SCI vs 4 healthy controls) were downloaded from the Gene Expression Omnibus(GEO) database. The ferroptosis-related gene set(471 genes) from the FerrDB database and the inflammation-related gene set(2,389 genes) from the GeneCards database were integrated, and differentially expressed genes(DEGs) were screened using the DESeq2 software. Subsequently, Gene ontology(GO) functional enrichment analysis, Kyoto Encyclopedia of Genes and Genomes(KEGG) pathway enrichment analysis and gene set enrichment analysis(GSEA) were performed on ferroptosis- and inflammation-related DEGs. Then, a protein-protein interaction(PPI) network was constructed to screen hub genes, which were further refined to identify core biomarkers using three machine learning algorithms. Finally, the diagnostic efficacy was evaluated by area under the curve(AUC) of receiver operating characteristic(ROC). Results: A total of 59 ferroptosis-inflammation-related DEGs were identified, including 40 upregulated genes and 19 downregulated genes, which were significantly enriched in the ferroptosis pathway and the toll-like receptor(TLR) signaling pathway. A total of 8 hub genes(HIF1A, JUN, TP53, STAT3, IL1B, PTEN, MAPK3, IFNG) were preliminarily identified by PPI network analysis. Further screening and validation by the three machine learning algorithms of least absolute shrinkage and selection operator(LASSO) regression, support vector machine(SVM), and random forest(RF), confirmed tumor protein p53(TP53), interferon-gamma(IFNG), and interleukin 1 beta(IL1B) as the core biomarkers. AUC analysis demonstrated their significant diagnostic efficacy(TP53, AUC=0.945; IFNG, AUC=0.883; IL1B, AUC=0.821). Conclusions: In this study, we screened three key genes, TP53, IFNG and IL1B, which are closely associated with the crosstalk between ferroptosis and inflammation in the peripheral blood of patients with SCI. This finding provides key molecular targets and new research directions for the in-depth analysis of the pathological mechanisms of SCI.