Integrative Molecular Pattern Learning for Mental Disorders Via Dual-Effect Matrix-Enabled Multiomics Platform

Mental health issues are surging in contemporary society, especially among adolescents and adults, harming well-being and social functioning. Objectively conducted timely screening, key to effective intervention, is vital in preventing severe social consequences. Here, we develop a high-throughput multiomics platform leveraging a well-established dual-effect matrix. By integrating this platform with advanced machine learning algorithms, we successfully acquire and analyze serum metabolomic and peptidomic profiles across four major mental disorders: major depressive disorder, bipolar disorder, schizophrenia, and anxiety disorder in both adults and adolescents. By integrating selected shared and disorder-specific metabolic/peptide features, our multiomics pathway analysis reveals two dysregulated biological pathways. Also, we achieve an overall diagnostic area under the curve (AUC) of 0.998 for distinguishing all patients with the four mental disorders from healthy controls and an average AUC of 0.974 for the specific classification of each individual mental disorder separately from the others. Moreover, we explore the differences in metabolic and peptide features between adults and adolescents and identify age-independent disease-specific features. This work advances both the mechanistic understanding of psychiatric disorders and the development of next-generation objective diagnostic tools.