Transcriptomic and machine learning analyses identify entosis-related key genes and potential regulatory networks in ischemic stroke

Abstract Background Ischemic stroke (IS) remains a major cause of mortality and long-term disability worldwide, while effective therapeutic targets remain limited. Entosis is a non-apoptotic form of cell-in-cell death, but its involvement in IS is unknown. This study sought to identify entosis-related key genes and delineate their potential regulatory networks in IS. Methods Bulk-transcriptome profiles were obtained for the training dataset GSE58294 and the validation dataset GSE122709, and entosis-related genes were compiled from previously published studies. Weighted gene co-expression network analysis and differential expression analysis were integrated with machine-learning selection and cross-dataset expression verification to determine key genes. Subsequent analyses included functional enrichment, immune infiltration, construction of molecular regulatory networks, drug prediction, and molecular docking. Results Two key genes, NOL9 and WDR3, were identified and were significantly upregulated in both transcriptomic datasets. The nomogram incorporating these genes demonstrated favorable apparent performance in the training cohort. Eight immune cell subtypes differed significantly in the infiltration analysis. M0 macrophages showed inverse correlations with NOL9 (correlation (cor) = -0.60, P < 0.001) and WDR3 (cor = -0.57, P < 0.001), whereas naive B cells correlated positively with NOL9 (cor = 0.46, P < 0.001) and WDR3 (cor = 0.35, P < 0.001). Genes associated with NOL9 were enriched in ribosome-biogenesis pathways, while those associated with WDR3 were enriched in spliceosome-related pathways. Construction of a competing endogenous RNA network yielded six microRNAs predicted to target NOL9 and WDR3; hsa-miR-499a-5p was additionally connected with three long non-coding RNAs (XIST, MALAT1, NEAT1). The analysis also predicted 398 transcription factors for NOL9 (e.g., SPI1, CTCF) and 331 for WDR3 (e.g., BRD4, CBFB). Digoxigenin emerged as an exploratory compound, with predicted binding energies of -9.82 kcal/mol for NOL9 and − 10.49 kcal/mol for WDR3. In RT-qPCR experiments, NOL9 and WDR3 were significantly downregulated in IS patients relative to controls; because this direction contrasted with the transcriptomic datasets, additional validation is required. Conclusions The analyses nominate NOL9 and WDR3 as exploratory entosis-related biomarker candidates for IS. Further studies are needed to resolve their expression direction and to establish their biological functions and therapeutic relevance.