AUTHOR=Zhou Yimin , Li Xin , Wang Zixiu , Ng Liqi , He Rong , Liu Chaozong , Liu Gang , Fan Xiao , Mu Xiaohong , Zhou Yu 

TITLE=Machine learning-driven prediction model for cuproptosis-related genes in spinal cord injury: construction and experimental validation

JOURNAL=Frontiers in Neurology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1525416

DOI=10.3389/fneur.2025.1525416

ISSN=1664-2295

ABSTRACT=IntroductionSpinal cord injury (SCI) severely affects the central nervous system. Copper homeostasis is closely related to mitochondrial regulation, and cuproptosis is a novel form of cell death associated with mitochondrial metabolism. This study aimed to explore the relationship between SCI and cuproptosis and construct prediction models.MethodsGene expression data of SCI patient samples from the GSE151371 dataset were analyzed. The differential expression and correlation of 13 cuproptosis-related genes (CRGs) between SCI and non-SCI samples were identified, and the ssGSEA algorithm was used for immunological infiltration analysis. Unsupervised clustering was performed based on differentially expressed CRGs, followed by weighted gene co-expression network analysis (WGCNA) and enrichment analysis. Three machine learning models (RF, LASSO, and SVM) were constructed to screen candidate genes, and a Nomogram model was used for verification. Animal experiments were carried out on an SCI rat model, including behavioral scoring, histological staining, electron microscopic observation, and qRT-PCR.ResultsSeven CRGs showed differential expression between SCI and non-SCI samples, and there were significant differences in immune cell infiltration levels. Unsupervised clustering divided 38 SCI samples into two clusters (Cluster C1 and Cluster C2). WGCNA identified key modules related to the clusters, and enrichment analysis showed involvement in pathways such as the Ribosome and HIF-1 signaling pathway. Four candidate genes (SLC31A1, DBT, DLST, LIAS) were obtained from the machine learning models, with SLC31A1 performing best (AUC = 0.958). Animal experiments confirmed a significant decrease in the behavioral scores of rats in the SCI group, pathological changes in tissue sections, and differential expression of candidate genes in the SCI rat model.DiscussionThis study revealed a close association between SCI and cuproptosis. Abnormal expression of the four candidate genes affects mitochondrial function, energy metabolism, oxidative stress, and the immune response, which is detrimental to the recovery of neurological function in SCI. However, this study has some limitations, such as unidentified SRGs, a small sample size. Future research requires more in vitro and in vivo experiments to deeply explore regulatory mechanisms and develop intervention methods.