Pan-Autoimmune miRNA-mRNA Dysregulation Network Analysis: Notebook Version

Abstract Background: Autoimmune diseases share overlapping immune dysregulation mechanisms, yet miRNA-mediated post-transcriptional regulation has been studied largely in isolation for individual conditions. Methods: We developed an eight-step computational framework integrating differential expression analysis, miRNA-mRNA network construction, machine learning classification, and SHAP interpretability across four autoimmune diseases: Vitiligo, Systemic Lupus Erythematosus (SLE), Rheumatoid Arthritis (RA), and Type 1 Diabetes (T1D) using four publicly available microarray datasets (NCBI GEO, Affymetrix GPL570). PTPN22 and NLRP1 were selected as anchor genes based on established GWAS evidence across all four conditions. miRNA-mRNA interactions were retrieved from miRTarBase (functional MTI only). Machine learning classifiers (Random Forest, XGBoost, SVM) were trained on disease-specific gene expression signatures with SHAP-based biological interpretability. Results: Differential expression analysis identified 9, 44, and 21 significant genes (adj. p < 0.05, |log₂FC| > 1) in Vitiligo, SLE, and RA respectively at a strict threshold; a relaxed threshold (adj. p < 0.05 only) yielding 2,745, 3,265, and 9,127 genes was used for network construction to ensure sufficient node coverage. T1D yielded no significant DEGs under BH correction due to insufficient sample size (n=22) and was therefore excluded from network analysis. Network construction integrating validated miRNA-mRNA interactions across the three diseases yielded a pan-autoimmune regulatory graph comprising 2,698 nodes and 5,464 edges, with hsa-miR-124-3p identified as the top hub regulator by degree centrality. Critically, 289 miRNAs were shared across all three diseases, constituting a pan-autoimmune regulatory signature. Random Forest classification achieved 98.8% test accuracy; SHAP analysis identified IRF6 and S100A16 as RA-discriminative biomarkers and SERPINB5 as a Vitiligo-specific candidate. PTPN22 was confirmed as a GWAS risk gene across all four diseases; NLRP1 was confirmed in Vitiligo and T1D, consistent with its inflammasome-driven role in skin and pancreatic autoimmunity. hsa-miR-181a-5p was identified as an RNAi therapeutic candidate targeting PTPN22. Conclusion: This framework demonstrates shared miRNA regulatory architecture across autoimmune diseases and provides a reproducible computational foundation for RNAi-based therapeutic target identification, with direct implications for Vitiligo and related autoimmune conditions. Keywords: miRNA-mRNA network, pan-autoimmune, differential expression, machine learning, SHAP, PTPN22, NLRP1, RNAi therapeutics, Vitiligo, SLE, Rheumatoid Arthritis.