Integrative machine learning and multi-omics framework identifies shared biomarkers for rheumatoid arthritis and ulcerative colitis

Abstract

Background Rheumatoid arthritis (RA) and ulcerative colitis (UC) are chronic inflammatory diseases with shared immune pathologies but limited common diagnostic biomarkers, which hinders the development of targeted therapies. Methods Public gene expression datasets were analyzed to identify differentially expressed genes (DEGs) common to both RA and UC. Functional enrichment and immune infiltration analyses revealed dysregulated pathways. A comprehensive machine learning framework that incorporated 12 algorithms and cross-validation was applied to screen for robust diagnostic biomarkers. Further, RA- and UC-related molecular subtypes were delineated, and the relationship between these shared biomarkers and immune infiltration characteristics was explored. Key findings were validated using single-cell RNA sequencing (scRNA-seq) of UC tissue to localize gene expression and qRT-PCR in cell models mimicking RA and UC. Results Analysis identified 19 shared DEGs, with functional enrichment analysis highlighting IL-17 signaling. Machine learning prioritized four key biomarkers (DUOX2, IDO1, NPY1R, SELL) with high diagnostic performance. scRNA-seq localized these genes predominantly to a pro-inflammatory “Macrophage-High” subpopulation and revealed VEGF-mediated crosstalk with endothelial cells. qRT-PCR confirmed significant expression changes of IDO1 and NPY1R in both RA-like and UC-like inflammation models. Conclusion This integrative approach identifies DUOX2, IDO1, NPY1R, and SELL as shared RA-UC biomarkers potentially linked to macrophage-driven inflammation and VEGF signaling. These findings provide insights into the common pathogenesis and potential targets for dual-disease diagnostics and therapeutics.