Abstract: BACKGROUND: Ferroptosis is closely associated with the pathogenesis of ischemic stroke, and targeting ferroptosis is a promising regimen for the treatment of ischemic stroke, but the specific regulatory targets are unclear.
OBJECTIVE: To screen ferroptosis-related characterized genes in ischemic stroke by bioinformatics and machine learning methods and validate them by cellular experiments to investigate the role of ferroptosis in ischemic stroke. 
METHODS: Eligible ischemic stroke-related datasets and ferroptosis expression datasets were selected based on GEO database and FerrDb database, and ferroptosis-related differential genes were screened by t-test. GO functional enrichment analysis with KEGG signaling pathway enrichment analysis was performed for ferroptosis-related differential genes. Characterized genes for ferroptosis in ischemic stroke were screened by PPI network analysis and machine learning. The reliability and biological functions of the characterized genes were explored using ROC analysis and GSEA analysis, followed by cell experiment. HT22 cells were divided into control and ischemic stroke groups. No intervention was made in the control group, and 0.1 mM H2O2 was added to the ischemic stroke group for 24 hours to simulate cellular oxidative stress injury and ferroptosis. The ferroptosis and the expression of characterized genes were verified by real-time fluorescence quantitative polymerase chain reaction (RT-PCR) and western blot assay.
RESULTS AND CONCLUSION: (1) Forty-five ferroptosis-associated differential genes were obtained, and GO and KEGG enrichment analyses revealed that the differential genes were closely associated with oxidative stress, autophagy, ferroptosis, adipocytokine signaling pathway, and mitochondrial metabolism. (2) A total of one ferroptosis characterized gene, nuclear factor erythroid 2-related factor 2 (NFE2L2), was identified by the MCODE plugin and cytoHubba plugin in the PPI network with the LASSO algorithm and SVM-RFE algorithm in machine learning. (3) Receiver operating characteristic curve analysis of NFE2L2 revealed that the diagnostic prediction models constructed in the training and validation sets had good accuracy and specificity. GSEA analysis of NFE2L2 revealed that the characterized gene was involved in the regulation of ischemic stroke pathogenesis through immunity, inflammatory response, amino acid metabolism, and neurofactor regulation. (4) RT-PCR and western blot analyses showed that the acyl coenzyme A synthetase long chain family, member 4 (ACSL4) mRNA and protein expression levels were significantly higher in the ischemic stroke group compared with the control group (P < 0.05), and the glutathione peroxidase 4 (GPX4) mRNA and protein expression levels were significantly lower in the ischemic stroke group (P < 0.05). Compared with the control group, the mRNA and protein expression levels of the characterized gene NFE2L2 were significantly higher in the ischemic stroke group (P < 0.05). (5) It suggests that ischemic stroke is closely related to ferroptosis, and targeting the characterized gene NFE2L2 may provide certain ideas and directions for the study and treatment of ischemic stroke. 

Key words: ischemic stroke, ferroptosis, bioinformatics, HT22 cell, machine learning, characteristic gene, cell experiment, NFE2L2, ASCL4, GPX4