Identification of post-translational modification-related biomarkers in ischemic stroke using bioinformatics and machine learning

Ischemic stroke (IS) remains a major clinical challenge due to the difficulty of early diagnosis and incomplete understanding of its pathological mechanisms. Post-translational modifications (PTMs) regulate key cellular processes in IS, but their roles as diagnostic biomarkers and therapeutic targets have not been fully elucidated. This study aimed to identify PTM-related genes (PTMRGs) associated with IS and evaluate their diagnostic and therapeutic potential using comprehensive bioinformatics and machine learning methods. Gene expression data from two GEO cohorts (GSE16561, training group n = 63; GSE58294, testing group n = 92) were analyzed. Differential expression analysis, weighted gene co-expression network analysis (WGCNA), and intersection with known PTMRGs were performed to screen candidate genes. Protein-protein interaction (PPI) networks and machine learning algorithms (Boruta, SVM-RFE, LASSO) were used to prioritize biomarkers. The expression of key genes was validated in clinical samples by RT-qPCR. An artificial neural network (ANN) was constructed to evaluate diagnostic performance. In addition, immune infiltration, gene set enrichment analysis (GSEA), gene-gene interaction (GGI), and molecular docking analyses were conducted to explore biological functions and therapeutic candidates. A total of 1465 upregulated genes and 1782 downregulated genes were identified. WGCNA revealed modules significantly associated with IS, yielding 75 key PTMRGs identified after intersection with DEGs and PTMRGs. Six genes (ATG7, KAT2A, RNF20, UBA1, UBE2I, and USP15) were identified as diagnostic markers with AUC > 0.7. RT-qPCR in 10 IS patients and 10 controls confirmed differential expression, consistent with bioinformatics results. The ANN model showed high diagnostic accuracy (AUC = 0.983 in training, 0.95 in testing). Functional enrichment linked these genes to ubiquitin-mediated proteolysis, DNA repair, and Myc signaling. Immune analysis showed associations with CD8 ⁺ T cells and neutrophils. Molecular docking suggested that ETYA may interact with UBE2I and KAT2A. In conclusion, the six diagnostic genes and the ANN model provide powerful diagnostic tools for IS, and ETYA may have a potential role in the treatment of IS.