The current diagnostic criteria for systemic juvenile idiopathic arthritis (sJIA) lack specificity. Diagnostic biomarkers are in urgent need to help with the early diagnosis of sJIA. Gene expression data of a JIA cohort study from Gene Expression Omnibus (GEO) database was adopted to get hub genes of sJIA by using integrated bioinformatic analysis including differentially expressed gene (DEG) analysis, weighted coexpression network analysis (WGCNA) and protein‒protein network interaction (PPI) analysis. Least absolute shrinkage and selection operator (LASSO) regression analysis was subsequently applied to identify biomarkers with the highest diagnostic potential for sJIA among these hub genes. A prediction model based on the identified biomarkers was constructed and subsequently validated in three additional independent GEO cohorts. Totally 761 DEGs were obtained by comparing gene expression profiles between sJIA patients and healthy controls. Twenty-two hub genes were identified by integrating WGCNA and PPI network analysis. All hub genes underwent LASSO regression analysis and three genes-ALAS2, S100A9, and S100A12-were eventually identified as the most promising diagnostic biomarkers. A prediction nomogram model based on these three genes was constructed, yielding an area under the curve (AUC) of 0.9337, and was subsequently validated in independent validation datasets, achieving AUC values of 0.9412, 0.9018, and 0.7064. The genes ALAS2, S100A9, and S100A12 showed significant association to sJIA and may serve as candidate diagnostic biomarkers pending further clinical validation.