Integrated Urinary Metabolomics and Artificial Intelligence for the Identification of Diagnostic, Prognostic, and Progression Biomarkers in Chronic Kidney Disease

Background: Chronic kidney disease (CKD) affects approximately 10% of the global population, and it is a critical challenge for public health worldwide. The integration of metabolomics with machine learning (ML) has emerged as a promising approach to enhancing disease diagnosis and prognosis. Objective: This study characterized the urinary metabolomic profiles of individuals with intermediate-to-advanced CKD and developed predictive models for disease diagnosis and prognosis. Methods: Urine samples from 129 participants, including healthy individuals and patients with CKD, were analyzed using liquid chromatography tandem mass spectrometry (LC-MS/MS). Multivariate analysis was performed to identify differences between groups. ML algorithms were the best for developing predictive models for the diagnosis, prognosis, and progression of CKD. Results: The results show the identities of 1499 metabolites. They also indicate that ML models effectively differentiated between healthy individuals and CKD patients, demonstrating high accuracy and specificity. The RF model was effective for diagnostic purposes, while the SVM showed superior performance in prognostic and disease progression predictions. The differentially expressed metabolites found, predominantly lipids, may serve as potential biomarkers for the detection and progression of CKD. Conclusions: The combination of urinary metabolomics with ML algorithms represents a promising, non-invasive approach for CKD diagnosis and follow-up, offering an alternative that may improve both the precision and personalization of disease management.