To decipher ferroptosis-related biomarkers of spinal cord injury and to perform the prediction of traditional Chinese medicine based on bioinformatics and machine learning

Xu Shaohua; Yang Xiaoshan; Nong Xiaoxia; Zhang Yong; Zhang Wenjie

doi:10.3969/j.issn.1002-1949.2025.04.004

Abstract：Objective To explore the expression and mechanism of ferroptosis-related genes in spinal cord injury(SCI)based on bioinformatics and machine learning algorithms and to predict potential ferroptosis-related biomarkers(PFRBs)in SCI.To perform drug susceptibility testing and the prediction of traditional Chinese medicine.Methods The GSE24179 dataset was downloaded,the differentially expressed genes(DEGs)were screened by the limma R package,and the protein-protein interaction(PPI)network was constructed for pruning.Ferroptosis-related genes were obtained,and potential biomarkers were screened by support vector machine(SVM)and receiver operating characteristic(ROC)curves for decision-making and evaluation,followed by immune infiltration analysis,drug sensitivity prediction in CellMiner database,and the biomarkers were finally determined by LASSO regression model.Through the TCD database and Coremine database,the traditional Chinese medicine-chemical composition-gene regulation axis was constructed,and the statistical analysis of properties,tastes and meridians was performed and the key traditional Chinese medicines were summarized.Results A total of 128 DEGs were obtained,and their biological functions included lipid metabolism and oxidative stress,which were closely related to ferroptosis.12 ferroptosis-related differentially expressed genes(FRDEGs)were obtained,a total of 10 PFRBs were included through SVM decision-making and ROC curve evaluation,and most genes in PFRBs showed linear correlation,with Arnt-like protein 1(ARNTL)-nuclear receptor coactivator 3(NCOA3)being the most correlated.Immunoinfiltration analysis showed that ARNTL was statistically correlated with NBCs,MBCs,gdTCs and Mono,especially for NBCs.Antimicrobial susceptibility prediction showed that 6-mercaptopurine was sensitive to ARNTL and AKR1C3.The LASSO regression model showed that NCOA3,CYP4F8 and AKR1C3 were the smallest variables to maintain the prediction performance.The prediction results of traditional Chinese medicine showed that the medicinal taste was mostly bitter,sweet and pungent,the medicinal properties were cold and warm,the surface and inner meridians were mostly spleen,stomach,lung and large intestine,and ginger,safflower,hemp seed and dodder seed were the key traditional Chinese medicines.Conclusions ARNT and NCOA3 participate in the microenvironment of SCI by regulating primary B cells through the ferroptosis pathway.The drug sensitivity of 6-mercaptopurine is the strongest,and some drugs/chemical components and traditional Chinese medicine can intervene in SCI,especially pirinixic acid and ginger.The correlation between them needs to be further studied.