Background <p>Systemic lupus erythematosus (SLE) is a chronic autoimmune disease influenced by multiple genetic and environmental factors.. This study used bioinformatics to identify new diagnostic biomarkers and explore the pathogenesis of SLE.</p> Method <p>Three array datasets of peripheral blood mononuclear cells (PBMCs) from SLE patients and control subjects were obtained from GEO, including GSE82221 and GSE11909 (merged for feature discovery) and GSE154851 (independent validation/machine learning training dataset). RRA analysis was used to identify statistically changed DEGs. GO and KEGG analyses explored biological mechanisms. Consistent clustering and machine learning models were employed to identify diagnostic biomarkers, further assessed using ROC and DCA analysis. The proportions of 22 immune cells in SLE patients were calculated via CIBERSORT algorithm, and correlations between biomarkers and immune cells were explored. Data from 76 clinical blood samples were collected, including ALT, AST, CR, BUN, CRP, PLT, WBC, lymphocyte (%), and neutrophil (%). RT-qPCR was used to validate the expression of diagnostic biomarkers.</p> Results <p>A total of 62 DEGs were identified, mainly involved in the cell cycle and TGF-beta signaling pathway. Six candidate diagnostic biomarkers for SLE were found: E2F2, KIAA0319L, TRIM58, MMP8, FKBP5, and TFDP1. E2F2 and TFDP1 showed AUC values &gt; 0.7 in training and validation datasets. CIBERSORT analysis revealed disturbed immune cell types in SLE patients, with plasma cells most relevant to E2F2 and TFDP1 expression. Expression of E2F2 and TFDP1 was upregulated in patient samples. The diagnostic models, including E2F2 and TFDP1, performed better than those without these biomarkers.</p> Conclusion <p>Our study identifies E2F2 and TFDP1 as potential diagnostic biomarkers for SLE patients and uncovers their most relevant immune cells.</p>