血流感染诊断生物标志物预测及免疫细胞浸润分析

Prediction of biomarkers for diagnosis of blood stream infections and analysis of immune cell infiltration

目的 通过生物信息学与机器学习识别血流感染(blood stream infections, BSI)患者的相关诊断生物标志物、发病机制及免疫细胞浸润水平,寻找新的药物靶标。方法 从高通量基因表达数据库(gene expression omnibus,GEO)中获取了BSI相关基因表达数据集。使用R语言进行差异表达基因(differentially expressed gene,DEG)筛选,然后进行基因富集分析。使用加权基因共表达网络分析(weighted correlation network analysis,WGCNA)筛选的关键模块基因。通过使用两种机器学习算法来识别中心基因。在外部数据集中使用受试者工作特征曲线(receiver operating characteristic,ROC)和箱线图模型来验证中心基因的诊断效能。通过CIBERSORT反卷积算法分析免疫细胞浸润水平。结果 本研究得到了330个加权基因共表达网络关键模块基因和DEGs的交集基因。基因富集分析结果显示,免疫和炎症相关通路被显著富集。通过机器学习和外部数据库验证共得到8个潜在生物标志物,ROC分析显示,8个潜在生物标志物曲线下面积(area under curve,AUC)均大于0.9。免疫细胞浸润分析表明,所有诊断标生物标志物都可能与免疫细胞有着不同程度的相关性。结论 通过生物信息学和机器学习方法,确定了潜在生物标志物,并构建了BSI诊断模型。本研究可以为BSI患者提供潜在的外周血诊断生物标志物,为BSI发病机制、新型治疗靶点和新药研发提供新的方向。

Objective Use bioinformatics and machine learning to find diagnostic biomarkers,pathogenesis,and immune cell infiltration levels that are relevant to people with blood stream infections(BSI)and look for new drug targets.Methods A BSI-related gene expression dataset was obtained from the High-Throughput Gene Expression Omnibus(GEO)database.Use the R language to screen differentially expressed genes(DEG),and then perform gene enrichment analysis.Key module genes were screened using weighted correlation network analysis(WGCNA).Identify central genes by using two machine learning algorithms.Use receiver operating characteristic(ROC)curves and box plot models in external datasets to validate the diagnostic efficacy of central genes.Analyze immune cell infiltration levels using the CIBERSORT deconvolution algorithm.Results This study obtained the intersection genes of 330 weighted gene co-expression network key module genes and DEGs.The results of gene enrichment analysis showed that immune and inflammation-related pathways were significantly enriched.A total of 8 potential biomarkers were obtained through machine learning and external database validation.ROC analysis showed that the area under the curve(AUC)of all 8 potential biomarkers was greater than 0.9.Immunocyte infiltration analysis indicates that all diagnostic biomarkers may have varying degrees of correlation with immune cells.Conclusion Through bioinformatics and machine learning methods,potential biomarkers were identified and a blood flow infection diagnosis model was constructed.This study can provide potential peripheral blood diagnostic biomarkers for patients with bloodstream infections and provide new directions for the pathogenesis of bloodstream infections,new treatment targets,and new drug development.