Untargeted metabolomics reveals differential metabolic pathways and biomarkers in the acute phase of Kawasaki disease

Abstract

Kawasaki disease (KD) is one of the most common rheumatic diseases in children and manifests with multisystem clinical features. Using untargeted metabolomics, our study investigated alterations in small-molecule metabolites in plasma of children with acute KD. Our study aimed to identify differential metabolic pathways and potential biomarkers. Plasma samples were collected from 30 children diagnosed with KD and 30 age-matched healthy controls (HC) at Jinhua Maternal and Child Health Hospital between January 2025 and December 2025. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was applied to analyse plasma samples. Enriched pathways were identified using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and differential metabolic pathways were determined using MetaboAnalyst 5.0. Differential metabolites were screened using the nonparametric Mann–Whitney U -test and receiver operating characteristic curve area (AUC). The conservative average AUC from nested cross-validation was reported as the primary performance metric. Pearson correlation analysis was conducted to evaluate correlations between metabolites and clinical parameters. In total, 261 differential metabolites were identified between the KD and HC groups, including 87 lipids and lipid-like molecules, 69 organic heterocyclic compounds, 38 benzenoids, 34 organic acids, 14 phenylpropanoids, and 19 other compounds. Pathway analysis of these differential metabolites revealed 30 putatively enriched metabolic pathways for exploratory analysis. Of these pathways, primary bile acid biosynthesis, arginine biosynthesis, histidine metabolism, and phenylalanine-tyrosine-tryptophan biosynthesis were nominally associated with KD. Six metabolites with exploratory discriminatory performance (AUC > 0.8) were further identified: L-tyrosine, L-tryptophan, glutamine, histidine, histamine, and taurocholic acid. A combined model incorporating these metabolites achieved an apparent AUC of 0.984 in the full dataset; nested cross-validation yielded a more conservative AUC of 0.889 (95% CI 0.798–0.968), indicating promising exploratory discriminatory performance. Untargeted metabolomics enables identification of metabolically perturbed pathways during the acute phase of KD. L-tyrosine, L-tryptophan, glutamine, histidine, histamine, and taurocholic acid may serve as candidate biomarkers for acute phase of KD.