Characterization of the Metabolites and Construction of a Novel Diagnostic Panel in Calcium Oxalate Urolithiasis by Electrospray Ionization – Mass Spectrometry (ESI-MS) Metabolomics. (2nd June 2022)

• Record Type: Journal Article
• Authors: Wang, Rui; Xie, Linguo; Zhang, Jingdong; Li, Jie; Xie, Haijie; Yang, Yu; Ren, Haotian; Shang, Zhiqun; Liu, Chunyu
• Is Part Of: Analytical letters. Volume 55:Number 12(2022)
• Page Start: 1997
• Page End: 2010
• Publication Date: 2022-06-02
• Subjects: Calcium oxalate -- diagnostic panel -- electrospray ionization – mass spectrometry (ESI-MS) -- metabolites -- urolithiasis; Chemistry, Analytic -- Periodicals; Chemistry, Analytic -- Abstracts; 543
• DOI: 10.1080/00032719.2022.2040523
• Languages: English
• ISSNs: 0003-2719

Abstract

Abstract: The metabolic differences between patients with calcium oxalate (CaOx) urolithiasis and the healthy population remain unclear. Currently there is no diagnostic panel for CaOx urolithiasis. This study aimed to identify the full-scale metabolic changes in the urine of CaOx urolithiasis patients and to establish a novel diagnostic panel as a noninvasive early prevention method. 240 urine samples were collected from 137 patients diagnosed with CaOx urolithiasis and 103 healthy controls. Electrospray ionization – mass spectrometry (ESI-MS) metabolomics and univariate, multivariate, and bioinformatic analyses were combined to identify the differential metabolites and related metabolic pathways between the groups. Integrated machine learning methods were adopted to discover potential biomarkers and establish a diagnostic panel. 79 differential metabolites were identified in the urine of patients compared with healthy controls. Pathways closely associated with these abnormal metabolites included metabolic pathways, amino acid biosynthesis, microbial metabolism in diverse environments, and protein digestion and absorption. Five potential metabolites, n-acetylputrescine, n-alpha-acetyl-L-arginine, ethyl 3-hydroxybutyrate, s-adenosylmethionine, and l-pyroglutamic acid, were selected as biomarkers by machine learning models, which included logistic regression (LR), random forest (RF), and support vector machine (SVM) models. The panel showed an excellent classification to differentiate patients from healthy controls with areas under the curve (AUC) of 0.998 in the training cohort and 0.993 in the validation cohort. The results revealed differential metabolites between the two groups. Furthermore, a panel of potential biomarkers was established for the first time and showed promising diagnostic value.