Background <p>The incidence of adolescent depression continues to rise, yet objective diagnostic biomarkers are lacking, with current clinical diagnosis primarily relying on subjective scales. Metabolomics offers a powerful tool for systematically revealing metabolic disturbances associated with the disease and discovering potential biomarkers.</p> Methods <p>This study enrolled 85 adolescents with depression and 46 healthy controls. Peripheral plasma samples were collected for untargeted metabolomics analysis. Differential metabolites were screened via differential analysis, and three machine learning algorithms—LASSO regression, random forest, and support vector machine—were employed for cross-validation to identify core feature metabolites. A logistic regression diagnostic model was constructed based on the selected metabolites. Its diagnostic efficacy and stability were evaluated using the area under the receiver operating characteristic curve, calibration curve, decision curve analysis, 5‑fold cross‑validation, and an independent validation set.</p> Results <p>A total of 21 differential metabolites were identified. 3 core metabolites were consistently selected by the three machine learning methods: Tyrosine, 3-Hydroxy-N,N,N-trimethyl-1-propanaminium chloride and Myristoylglycine. These metabolites showed significant content differences between the two groups, and their levels correlated well with Hamilton Depression Rating Scale scores. A logistic regression model built with three of these metabolites demonstrated excellent diagnostic performance in the training set, with an AUC of 0.944. The average AUC remained 0.936 after 5-fold cross-validation, and the independent validation set was 0.968.</p> Conclusion <p>This study identified a panel of core metabolites in the peripheral blood of adolescents with depression, involving amino acid, lipid, and energy metabolism pathways. The diagnostic model based on these metabolites shows high discriminatory power, provides new insights into the metabolic mechanisms of adolescent depression, and demonstrates potential as an objective auxiliary diagnostic tool. Future external validation in multi-center, large-sample cohorts is needed to advance its clinical translation.</p> Clinical trial number <p>Our study is a clinical observational study, so clinical trial number: not applicable.</p>