Metabolomic differences in irritable bowel syndrome with and without anxiety: a cross-sectional study with exploratory machine learning analysis

Anxiety is a common comorbidity in Irritable Bowel Syndrome (IBS), yet its impact on the metabolic characteristics of IBS remains unclear. This study explored metabolic alterations in IBS patients with and without anxiety through metabolomics and machine learning. Fecal samples from 42 IBS patients fulfilling Rome IV criteria and 20 healthy controls (HC) were analyzed using liquid chromatography–mass spectrometry (LC–MS). IBS patients were classified as IBS with anxiety (IBS-A, n = 27) or IBS without anxiety (IBS-NA, n = 15) according to the Hospital Anxiety and Depression Scale anxiety score. Metabolomic profiles were assessed using principal component analysis, false discovery rate (FDR)-controlled differential metabolite analysis, heatmap visualization, and exploratory supervised analyses. Support vector machine, random forest, and LASSO logistic regression models were evaluated using nested stratified cross-validation. IBS-A patients showed higher IBS Severity Scoring System and Hospital Anxiety and Depression Scale scores than IBS-NA patients and HC. FDR-controlled analysis identified 253 differential metabolic features between IBS patients and HC, including 50 upregulated and 203 downregulated features in IBS. Principal component analysis showed partial separation among IBS-A, IBS-NA, and HC, and the top 50 FDR-significant features showed distinct abundance patterns between IBS and HC. Human-restricted KEGG pathway enrichment did not identify any pathway that remained significant after FDR correction. Under nested cross-validation, the best-performing random forest model showed moderate internal discrimination for IBS versus HC (mean AUROC = 0.81), whereas discrimination between IBS-A and IBS-NA was weak and unstable (mean AUROC = 0.65). This study identified fecal metabolomic differences associated with IBS and explored anxiety-related metabolic patterns. These findings should be interpreted as exploratory and hypothesis-generating.