Integrated Chemometric and Machine Learning Analysis Identifies Peripheral Biosignatures Distinguishing Major Depressive Disorder from Bipolar Disorder: A Translational Cross-Sectional Study

Background and Objectives: Major Depressive Disorder (MDD) and Bipolar Disorder (BD) lack objective molecular stratification despite partial clinical overlap, particularly during depressive phases. This cross-sectional study explored whether coordinated peripheral biomarker patterns could be identified using an integrated multivariate analytical framework. Materials and Methods: A total of 151 participants (MDD n = 41; BD n = 40; HC (healthy controls) n = 70) were profiled for 42 blood-derived parameters including composite inflammatory indices, hematological markers, trace elements measured by ICP-MS, and circulating BDNF and NLRP3 quantified by ELISA. Data were analyzed using univariate testing, unsupervised dimensionality reduction (PCA, t-SNE), and supervised classification (PLS-DA with cross-validation and permutation testing). Results: Thirty-seven of 42 parameters showed significant inter-group differences (p < 0.05). Circulating NLRP3 concentrations were markedly reduced in both psychiatric groups compared with HC. Composite inflammatory indices (NLR, SIRI, SII) were elevated in MDD. Zinc levels were modestly reduced, while manganese levels were increased in psychiatric cohorts. BDNF showed lower concentrations in MDD and higher concentrations in BD relative to HC. Cross-validated PLS-DA classification for psychiatric disorder vs. controls yielded an accuracy of 89.4% (AUC-ROC 0.947), with permutation testing indicating performance above chance. However, the sample-to-variable ratio and exploratory design warrant cautious interpretation. Conclusions: Multidomain peripheral biomarker profiling identified coordinated biochemical differences across diagnostic groups. These findings suggest the presence of multidimensional peripheral signatures associated with mood disorders within an exploratory framework.