基于机器学习与孟德尔随机化筛选脓毒症的潜在靶点  

Screening potential targets for sepsis based on machine learning and Mendelian randomization

作　　者：杨丹丹 Yang Dandan(School of Clinical Medicine,Anhui Medical University,Hefei,Anhui 230012,China)

机构地区：[1]安徽医科大学临床医学院,安徽合肥230012

出　　处：《齐齐哈尔医学院学报》2025年第6期509-517,共9页Journal of Qiqihar Medical University

摘　　要：目的应用机器学习与孟德尔随机化方法,筛选脓毒症的潜在生物标志物,并探索其在免疫调节中的分子机制。方法本研究利用公开的脓毒症基因表达数据,首先通过差异表达分析和加权基因共表达网络分析(WGCNA)构建基因共表达网络,识别与脓毒症相关的基因模块。随后,采用多种机器学习算法筛选关键诊断基因,并运用CIBERSORT算法分析免疫细胞浸润特征。最终,通过孟德尔随机化分析验证候选基因与脓毒症风险之间的因果关系。结果差异表达分析和WGCNA分析揭示了多个与脓毒症显著相关的基因模块。通过113种机器学习算法的筛选,最终确定16个核心诊断基因。免疫细胞浸润分析结果显示脓毒症患者的免疫细胞组成发生显著变化。孟德尔随机化分析进一步确认了GRB10基因与脓毒症风险之间存在显著的因果关联(OR=1.459,P=0.034)。结论综合运用多种生物信息学分析方法,筛选出脓毒症的潜在生物标志物。GRB10基因与脓毒症风险的因果关联为其作为治疗靶点提供了新的科学依据,同时研究还揭示了免疫细胞浸润在脓毒症发生发展中的关键作用。这些发现为脓毒症的早期诊断和个体化治疗提供了新的视角和潜在靶点。Objective To apply machine learning and Mendelian randomization to screen potential biomarkers of sepsis and explore their molecular mechanisms in immune regulation.Methods Using publicly available sepsis gene expression data,this study first constructed gene co-expression networks by differential expression analysis and weighted gene co-expression network analysis(WGCNA)to identify gene modules associated with sepsis.Subsequently,multiple machine learning algorithms were used to screen key diagnostic genes,and the CIBERSORT algorithm was applied to analyze immune cell infiltration characteristics.Finally,the causal relationship between candidate genes and sepsis risk was verified by Mendelian randomization analysis.Results Differential expression analysis and WGCNA analysis revealed multiple gene modules significantly associated with sepsis.Sixteen core diagnostic genes were finalized by screening with 113 machine learning algorithms.Immune cell infiltration analysis revealed significant changes in immune cell composition in sepsis patients.Mendelian randomization analysis further confirmed a significant causal association between the GRB10 gene and sepsis risk(OR=1.459,P=0.034).Conclusions The present study utilized a variety of bioinformatics analysis methods to screen potential biomarkers for sepsis.The causal association between GRB10 gene and sepsis risk provides a new scientific rationale for its use as a therapeutic target,and the study also revealed the critical role of immune cell infiltration in the development of sepsis.These findings provide new perspectives and potential targets for early diagnosis and individualized treatment of sepsis.

关 键 词：脓毒症 机器学习 孟德尔随机化 GRB10 免疫浸润 

分 类 号：R73[医药卫生—肿瘤]