Guo Wei, Chen Hao, Wang Feng, Chi Yingjiao, Zhang Wei, Wang Shan, Chen Kezhu, Chen Hong
Department of Pediatrics, First Affiliated Hospital, Heilongjiang University of Chinese Medicine, Harbin, China.
Department of Surgery, Heilongjiang Academy of Traditional Chinese Medicine, Harbin, China.
Front Immunol. 2025 Jun 17;16:1586584. doi: 10.3389/fimmu.2025.1586584. eCollection 2025.
Septic shock in children is an infectious disease caused by low immunity, and its mortality is very high. Early prediction of the risk of death in children with septic shock is helpful for clinicians to judge the severity of the disease, take active treatment measures, and improve the adverse outcomes of patients. However, the mechanism of death from sepsis in children remains unclear. This study aims to use bioinformatics and machine learning algorithms to identify key genes and pathways associated with fatal sepsis in children, and provide theoretical basis for rational drug use in follow-up TCM treatment.
Gene expression profiles were obtained from the GEO database (GSE4607) for 15 blank patients and 14 children with sepsis death. Differentially expressed genes (DEGs) were enriched by GO and KEGG pathways. Construct and visualize protein-protein interaction (PPI) networks to identify candidate genes responsible for fatal sepsis in children. Three kinds of machine learning models were established, and the candidate genes were screened by intersection to obtain the core genes with diagnostic value. ROC curve was drawn for core genes to clarify the diagnostic value of genetic markers.
Analysis of differences in the preprocessed dataset identified 83 genes, including 78 up-regulated genes and 5 down-regulated genes. 17 candidate genes were screened by protein interaction network analysis. Three machine learning algorithms LASSO, random forest (RF), and support vector machine recursive feature elimination (SVM-RFE) were used to finally screen out three core genes: CD163, MCEMP1 and RETN. CD163, MCEMP1 and RETN may jointly regulate complement and coagulation cascades, toll like receptor signaling pathway, graft versus host disease, type I diabetes mellitus.
In this study, three core genes (CD163, MCEMP1 and RETN) that lead to sepsis death in children were screened out, providing a new understanding of the lethal mechanism of sepsis in children and a promising new therapeutic approach.
儿童脓毒性休克是一种因免疫力低下引起的感染性疾病，死亡率极高。早期预测儿童脓毒性休克的死亡风险有助于临床医生判断疾病严重程度，采取积极治疗措施，改善患者不良结局。然而，儿童脓毒症的死亡机制仍不清楚。本研究旨在利用生物信息学和机器学习算法识别与儿童致命性脓毒症相关的关键基因和通路，为后续中医治疗合理用药提供理论依据。
从GEO数据库（GSE4607）获取15例空白对照患者和14例脓毒症死亡儿童的基因表达谱。通过GO和KEGG通路对差异表达基因（DEGs）进行富集。构建并可视化蛋白质-蛋白质相互作用（PPI）网络，以识别导致儿童致命性脓毒症的候选基因。建立三种机器学习模型，并通过交集筛选候选基因，以获得具有诊断价值的核心基因。绘制核心基因的ROC曲线，以阐明基因标志物的诊断价值。
对预处理数据集的差异分析确定了83个基因，其中78个上调基因和5个下调基因。通过蛋白质相互作用网络分析筛选出17个候选基因。使用三种机器学习算法LASSO、随机森林（RF）和支持向量机递归特征消除（SVM-RFE）最终筛选出三个核心基因：CD163、MCEMP1和RETN。CD163、MCEMP1和RETN可能共同调节补体和凝血级联反应、Toll样受体信号通路、移植物抗宿主病、I型糖尿病。
本研究筛选出导致儿童脓毒症死亡的三个核心基因（CD163、MCEMP1和RETN），为儿童脓毒症的致死机制提供了新的认识和一种有前景的新治疗方法。
