Journal Article

IDDF2025-ABS-0023 Identification of biomarkers for treatment escalation of ulcerative colitis based on untargeted metabolomics and machine learning algorithms: a prospective cohort study

Han, Muzhou, Shi, Haiyun, Li, Peng, Zhu, Siying, Wu, Yongdong, Wang, Hao

2025

Overview

BackgroundBiomarkers to guide clinical decision-making in active ulcerative colitis (UC) patients are urgently needed. This study aims to identify metabolites associated with UC treatment escalation and establish prediction models based on untargeted metabolomics and machine learning algorithms.MethodsIn this prospective cohort study, we enrolled active UC patients and followed up for 8 weeks after collecting blood samples to judge whether they needed treatment escalation during the subsequent course of the disease. Liquid chromatography-mass spectrometry-based untargeted metabolomics analysis was performed on 88 plasma samples (44 active UC patients requiring treatment escalation later on and 44 active UC patients not requiring treatment escalation later on). Univariate and multivariate analyses were applied to identify metabolic biomarkers for UC treatment escalation. Metabolic pathway enrichment analysis was performed to reveal the disturbed metabolic pathways related to UC treatment escalation. Four machine learning algorithms, including support vector machine (SVM), random forest (RF), k-nearest neighbor (KNN), and logistic regression were used to build diagnostic models for UC treatment escalation.ResultsNine significantly differential metabolites were identified as the candidate biomarkers for UC treatment escalation. Of these, levels of 8 metabolites were decreased, including 12-hydroxydodecanoic acid, canthaxanthin, phenylacetaldehyde, 3,5-dihydroxybenzoic acid, benzene, amantadine, azelaic acid, and theophylline. Conversely, the level of 4-aminophenol was significantly increased in the UC treatment escalation group. Pathway analysis revealed that phenylalanine metabolism and ether lipid metabolism are the disturbed metabolic pathways related to treatment escalation. The protein-metabolite interaction network identified 21 proteins associated with 9 treatment escalation-related metabolites. The areas under the receiver operating characteristic curve of the SVM, RF, KNN, and logistic regression models based on metabolic biomarkers were 0.909, 0.999, 0.918, and 0.900, respectively.ConclusionsThe plasma metabolome represents a promising source of biomarkers for the prediction of treatment escalation in active UC. Metabolic biomarkers, combined with machine learning algorithms, could be efficient for risk assessment and early identification of UC treatment escalation.

• Publisher: BMJ Publishing Group LTD
• Subject: Algorithms / Biomarkers / Cohort analysis / Decision making / Inflammatory bowel disease / Learning algorithms / Lipid metabolism / Liquid chromatography / Machine learning / Mass spectroscopy / Metabolic pathways / Metabolism / Metabolites / Metabolomics / p-Aminophenol / Patients / Phenylacetaldehyde / Prediction models / Regression analysis / Risk assessment / Support vector machines / Theophylline / Ulcerative colitis