Bioinformatic identification and clinical validation of VIM and OSM as prognostic biomarkers in cerebral infarction: implications for risk stratification.

Reliable molecular biomarkers for predicting cerebral infarction outcomes remain limited, highlighting the need for integrative approaches that bridge bioinformatic discovery with clinical validation. To identify key differentially expressed genes (DEGs) prognostic for cerebral infarction and evaluate their clinical utility for risk stratification through integrated bioinformatic analysis and prospective cohort validation. Functional annotation employed GO enrichment and protein-protein interaction network analysis. A prospective cohort enrolled 151 cerebral infarction patients, with peripheral blood samples subjected to qPCR analysis of candidate genes. Prognostic predictive capacity was assessed via multivariable Cox regression, Kaplan-Meier survival analysis, and ROC curve analysis with clinical follow-up data. Five candidate DEGs (VIM, OSM, PTGS2, SOD2, SAMSN1) were identified, enriched in inflammatory response, nitric oxide metabolism, and lipopolysaccharide response pathways. qPCR confirmed significantly elevated VIM, OSM, and PTGS2 expression in poor prognosis group (p < 0.01). Multivariable Cox regression identified VIM (HR = 4.475), OSM (HR = 2.800), and homocysteine (Hcy; HR = 1.120) as independent prognostic risk factors. Kaplan-Meier analysis demonstrated significantly reduced survival in high-expression groups (all p < 0.01). The combined model integrating VIM, OSM, and Hcy achieved superior predictive performance (AUC = 0.811; sensitivity 72.55%, specificity 78.00%, Youden's index 0.506) compared to VIM alone (AUC = 0.760). VIM and OSM exhibit robust bioinformatic associations and stable expression with independent prognostic value in clinical cohorts.