目的 基于生物信息学分析和机器学习算法探究铁死亡参与脓毒症的机制及预测治疗脓毒症的潜在中药,以期为脓毒症的临床治疗提供新的治疗方向及用药思考.方法 从gene expression omnibus(GEO)数据库下载脓毒症的基因表达数据,利用R语言筛选差异表达基因(differentially expressed gene,DEG).在FerrDb平台收集与铁死亡相关的靶点.采用weighted correlation network analysis(WCGNA)筛选关键模块基因.得到DEG、WGCNA和铁死亡基因数据集的交集基因,然后通过3种机器学习算法筛选出脓毒症相关铁死亡核心基因,利用GSE232753数据验证脓毒症相关铁死亡核心基因.通过CIBERSORT算法评估22种免疫细胞丰度进而评估脓毒症与免疫细胞浸润的相关性.最后,通过核心基因预测潜在治疗中药,根据药物性味归经拓展中医理论.结果 利用随机森林(random forest,RF)、支持向量机-递归特征消除(support vector machine-recursive feature elimination,SVM-RFE)和最小绝对收缩和选择算法(the least absolute shrinkage and selection operator,LASSO)算法共筛选得到脓毒症相关铁死亡核心基因,进一步利用GSE232753数据集进行验证,最终确定TXN、NR1D2、PARP15为脓毒症相关铁死亡的核心基因,且其基因表达量均显示出统计学显著性差异(P＜0.05),ROC分析显示,3个潜在核心基因曲线下面积(area under curve,AUC)均大于0.9.免疫细胞浸润分析表明多种免疫细胞可能参与脓毒症的发展,以上3个核心基因在不同程度上与免疫细胞有一定的相关性.利用本草图鉴数据库对上述3个核心基因,通过《中华本草》和《中药大辞典》筛选整理得到脓毒症相关铁死亡核心基因靶向中药沙棘、蜂房等22味中药,靶向中药大多药性偏平,药味甘,归经入胃、肺.结论 铁死亡参与脓毒症的机制可能与TXN、NR1D2、PARP15基因的表达水平有关,同时预测了针对脓毒症相关铁死亡核心基因的靶向治疗中药,为脓毒症的治疗提供新的治疗方向及用药参考.
Exploring the mechanism of ferroptosis involved in sepsis and predicting traditional Chinese medicine based on bioinformatics analysis
Objective Based on bioinformatic analysis and machine learning algorithms,to explore the mechanism of ferroptosis involved in sepsis and predict potential traditional Chinese medicine for the treatment of sepsis,and provide a new treatment direction and medication thinking for the clinical treatment of sepsis.Methods Gene expression data of sepsis were downloaded from GEO database,and differentially expressed genes(DEGs)were screened by R language.Targets related to ferroptosis were collected on the FerrDb platform.WCGNA was used to screen key module genes.The intersection genes of DEG,WGCNA and ferroptosis gene data sets were obtained,and then the core genes of sepsis-related ferroptosis were screened by three machine learning algorithms.The core genes of sepsis-related ferroptosis were verified by GSE232753 data.The abundance of 22 immune cells was evaluated by the CIBERSORT algorithm to evaluate the correlation between sepsis and immune cell infiltration.Finally,the core gene was used to predict the potential treatment of traditional Chinese medicine,and the theory of traditional Chinese medicine was expanded according to the nature and taste of the drug.Results The core genes of sepsis-related ferroptosis were screened by RF,SVM and LASSO algorithms,and further verified by GSE232753 data set.Finally,TXN,NR1D2 and PARP15 were determined as the core genes of sepsis-related ferroptosis,and their gene expression levels showed statistically significant differences(P＜0.05).ROC analysis showed that the area under the curve(AUC)of the three potential core genes was greater than 0.9.Immune cell infiltration analysis showed that a variety of immune cells may be involved in the development of sepsis,and the above three core genes had a certain correlation with immune cells to varying degrees.The above three core genes were screened and sorted by the herbal database,such as"Chinese Materia Medica"and"Dictionary of Traditional Chinese Medicine".The core genes of sepsis-related ferroptosis were targeted at 22 traditional Chinese medicines,such as sea buckthorn,beehive,and so on.Most of the targeted traditional Chinese medicines had a relatively mild nature,a sweet taste,and entered the stomach and lungs through the meridians.Conclusion The mechanism by which ferroptosis was involved in sepsis may be related to the expression levels of the TXN,NR1D2,and PARP15 genes.At the same time,it was predicted that traditional Chinese medicine for targeted therapy of sepsis-related ferroptosis core genes could provide new treatment directions and medication references for the treatment of sepsis.
SepsisBioinformaticsFerroptosisMachine learningTraditional Chinese medicine prediction
NETL
NSTL
万方数据
