Identification of Senescence-Related Genes for the Prediction of Ulcerative Colitis Based on Interpretable Machine Learning Models.

Abstract

Background: Cellular senescence, a hallmark of aging, significantly contributes to the pathology of ulcerative colitis (UC). Despite this, the role of senescence-related genes in UC remains largely undefined. This study seeks to clarify the impact of cellular senescence on UC by identifying key senescence-related genes and developing diagnostic models with potential clinical utility.

Methods: Clinical data and gene expression profiles were obtained from the Gene Expression Omnibus (GEO) database. Senescence-related differentially expressed genes (sene-DEGs) between patients with UC and healthy controls were identified using various bioinformatics techniques. Functional enrichment and immune infiltration analyses were performed to understand subtype characteristics derived from sene-DEGs through consensus clustering. Machine learning algorithms were employed to select feature genes from sene-DEGs, and their expression was validated across multiple independent datasets and human specimens. A nomogram incorporating these feature genes was created and assessed, with its diagnostic performance evaluated using receiver operating characteristic (ROC) analysis on independent datasets.

Results: Fourteen senescence-related differential genes were identified between patients with UC and healthy controls. These genes enabled the classification of patients with UC into molecular subtypes via unsupervised clustering. ABCB1 and LCN2 emerged as central hub genes through machine learning and feature importance analysis. ROC analysis verified their diagnostic value across various datasets. Validation in independent datasets and human specimens supported the bioinformatics findings. Furthermore, the expression levels of ABCB1 and LCN2 showed significant associations with immune cell profiles. The logistic regression (LR) model based on these genes demonstrated accurate UC prediction, as confirmed by ROC curve analysis. The nomogram model, constructed with feature genes, exhibited outstanding prediction capabilities, supported by DCA, C index, and calibration curve assessments.

Conclusion: This integrated bioinformatics approach identified ABCB1 and LCN2 as significant biomarkers associated with cellular senescence. These findings enhance the understanding of cellular senescence in UC pathogenesis and propose its potential as a valuable diagnostic biomarker.