Answers are grounded in this paper and cite their sources.
Accurate differentiation of bipolar disorder (BD), major depressive disorder (MDD), and schizophrenia (SCZ) is hindered by symptomatic overlap and subjective clinical assessments. This study applied untargeted plasma metabolomics to a large cohort of 1609 adults (BD, n = 352; MDD, n = 506; SCZ, n = 373; healthy controls, n = 378) diagnosed per DSM-IV criteria. Following stringent pre-processing and quality control, we identified metabolic features associated with diagnostic outcomes using mutual information analysis, a non-parametric method capturing nonlinear relationships, and further evaluated their diagnostic potential. After adjustment for sex ang age, a total of 297 metabolites correction showed significant associations with psychiatric diagnosis after FDR (p < 0.01). Adjusted sex and age predictive models—including logistic regression, LASSO, random forest, and XGBoost—trained on random subsets of these metabolites consistently achieved high classification performance, with mean accuracy exceeding 0.88 across all pairwise comparisons after FDR correction. These results reveal distinct yet overlapping metabolic alterations across disorders, suggesting shared pathophysiological mechanisms. Notably, robust and reproducible predictions were attainable even with small feature subsets, underscoring the stability of metabolomic signatures. Collectively, our findings support plasma metabolomics as an objective, biologically grounded tool to complement clinical assessment and enhance diagnostic precision in major psychiatric disorders.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: