Unveiling the aging-immune axis in irritable bowel syndrome: a multi-omics and machine learning approach to biomarker discovery and validation.

Background: Irritable bowel syndrome (IBS) is a prevalent functional gastrointestinal disorder with an elusive pathophysiology. Although immune dysregulation and mild mucosal inflammation are recognized as important factors in IBS, the specific contribution of immunosenescence remains unclear. Prior MR studies on IBS focused on single molecular traits; however, an integrated multi-omics framework for aging-immune genes has not been applied. Here, we integrate cis-eQTL/pQTL/mQTL with machine learning and Methods: Aging- and immune-related genes were curated from published databases and analyzed using genome-wide cis-expression quantitative trait loci (cis-eQTL), cis-protein quantitative trait loci (cis-pQTL), and cis-methylation quantitative trait loci (cis-mQTL) datasets. All QTL data were derived from blood or plasma samples. Candidate genes were identified through integrative multi-omics analysis, followed by functional annotation, machine learning-based feature selection, immune infiltration profiling, and drug-target prediction. The expression of key biomarkers was validated using reverse transcription quantitative polymerase chain reaction (RT-qPCR) in colonic tissues from a rat model of diarrhea-predominant IBS (IBS-D). Results: Integrative multi-omics analysis initially identified 34 high-confidence candidate genes. Through multi-algorithm machine learning such as least absolute shrinkage and selection operator (LASSO), support vector machine recursive feature elimination (SVM-RFE), and random forest, these candidates were refined to a core panel of five biomarkers: the antioxidant enzyme catalase (CAT), acyl-CoA dehydrogenase very long chain (ACADVL), chemokine (C-C motif) ligand 4 (CCL4), cyclin E1 (CCNE1), and Jagged canonical Notch ligand 1 (JAG1). These biomarkers demonstrated strong diagnostic performance for CAT, CCNE1, and JAG1 in the validation cohort (AUC: 0.935-0.951), while ACADVL and CCL4 showed only moderate diagnostic potential (AUC: 0.645-0.649). The combined AUC range in the training cohort was 0.898-0.959. However, in the IBS-D rat model, only JAG1 was significantly upregulated in colonic tissue, whereas CAT, ACADVL, CCL4, and CCNE1 showed no significant changes. Among the five markers, only JAG1 was significantly upregulated in the IBS-D rat colon, confirming its local gut relevance. CAT and CCNE1 showed strong diagnostic performance (AUC >0.9), whereas ACADVL and CCL4 performed modestly (AUC <0.7). Given the blood-derived QTL data and colonic validation, cross-tissue heterogeneity is a key caveat: JAG1 is robustly validated in colon, while the remaining four markers-especially ACADVL and CCL4-require further evaluation in blood, PBMCs, or other relevant tissues. Functional enrichment analysis revealed their involvement in critical biological processes, including oxidative stress, fatty acid metabolism, immune cell recruitment, cell cycle progression, and epithelial-immune crosstalk via Notch signaling. Immune infiltration analysis uncovered a distinct immune landscape in IBS, with significant correlations between biomarker expression and immune cell populations. Notably, only JAG1 was significantly upregulated in the colonic tissues of the IBS-D rat model, confirming its relevance to gut pathophysiology. No significant changes were detected for CAT, ACADVL, CCL4, or CCNE1 in the same tissue samples, highlighting the tissue-specific nature of these candidate biomarkers. The SMR analysis demonstrated consistent causal directions between genetically predicted expression and disease risk for these five markers, with no discordance relative to their upregulation in IBS patients. All biomarker validation was performed at the transcriptomic level (RT-qPCR); no protein-level assays (e.g., Western blot, IHC, or ELISA) were conducted. Conclusions: This study delineates the genetic architecture linking aging and immunity to IBS through an integrative multi-omics and machine learning approach, providing novel putatively causal evidence and identifying JAG1 as a robustly validated biomarker with diagnostic and therapeutic potential. The remaining candidates warrant further investigation in appropriate biological contexts.