Abstract
INTRODUCTION: Ulcerative colitis (UC) is characterized by chronic colonic mucosal inflammation, with its pathogenesis involving multidimensional interactions and limitations in clinical treatment. Dietary restriction (DR) is a commonly used approach for UC patients to alleviate symptoms, and exploring the role of DR-related genes in UC could provide new directions for the development of precision therapies. METHODS: Bioinformatics analysis was performed on UC-related datasets (GSE75214, GSE73661) obtained from the GEO database. Candidate genes were acquired by intersecting differentially expressed genes (DEGs) with dietary restriction-related genes (DRRGs). Subsequently, key genes were identified via machine learning algorithms and ROC curve analysis. A deep neural network (DNN) model and a diagnostic nomogram were constructed. In addition, gene set enrichment analysis (GSEA), gene set variation analysis (GSVA), immune infiltration analysis, and single-cell RNA sequencing (scRNA-seq) analysis were conducted. Finally, the expression of key genes was validated through experiments. RESULTS: CPT1A, ANGPTL4, and CLDN1 were identified as the key genes. The deep neural network (DNN) model achieved area under the curve (AUC) values of 0.914 and 0.933 in the two datasets, respectively; the diagnostic nomogram exhibited high predictive performance (AUC > 0.7), and decision curve analysis (DCA) revealed its potential clinical net benefit. Enrichment analyses demonstrated that the key genes were significantly enriched in dietary restriction (DR)-related pathways, including cytokine-receptor interaction, the IL2-STAT5 signaling pathway, and fatty acid metabolism. Thirty-two activated pathways and five inhibited pathways were detected in UC patients (e.g., the oxidative phosphorylation pathway was suppressed). Immune infiltration analysis identified 27 differentially infiltrating immune cell types. CLDN1 was localized to epithelial cells, ANGPTL4 to fibroblasts, and CPT1A to endothelial cells. Macrophages were identified as a signaling hub in UC, showing intensified crosstalk with stromal and vascular cells via pathways such as ACKR1. Experimental validation confirmed that ANGPTL4 and CLDN1 were highly expressed in UC, whereas CPT1A was lowly expressed, a pattern consistent with the expression trends observed in public database analyses. DISCUSSION: These results indicated that CPT1A, ANGPTL4, and CLDN1 are involved in the pathological regulation of UC by DR through modulating the metabolism-immune-barrier axis, providing novel biomarkers and potential intervention targets for the clinical diagnosis and targeted therapy of UC.