Molecular Signatures of Early-Onset Bipolar Disorder and Schizophrenia: Transcriptomic and Machine-Learning Insights into Calcium and cAMP Signaling, Including Sex-Specific Patterns.

Early age of onset is a major predictor of poor disease course in Bipolar Disorder (BD) and Schizophrenia (SCZ), often associated with greater symptom severity, cognitive decline, and worse outcomes. However, the biological mechanisms that shape age- and sex-specific vulnerability remain unclear, limiting progress toward early identification and intervention. To address this gap, we conducted an integrative transcriptomic study of 369 postmortem dorsolateral prefrontal cortex samples from the CommonMind Consortium. Differential gene expression, Weighted Gene Co-Expression Network Analysis, and gene set enrichment analysis were applied to identify pathways associated with age of onset, complemented by sex-stratified models and cellular deconvolution. To assess predictive signals, we applied a rigorous two-stage machine-learning framework using nested cross-validation, with Lasso feature selection followed by L2-regularized logistic classification. Performance was evaluated solely on held-out test folds. Genes and modules linked to earlier onset showed consistent enrichment for calcium signaling, with downregulation of CACNA1C and multiple adenylate-cyclase-related transcripts, while female-specific analyses revealed selective dysregulation of cyclase-associated pathways. Network analysis identified a calcium-enriched module associated with onset and sex, and diagnosis-specific modeling highlighted MAP2K7 in early-onset BD. The predictive model achieved an AUC of 0.63, and the top 50 machine-learning features were significantly enriched in calcium signaling pathway. These findings converge on calcium-cAMP signaling networks as key drivers of early psychiatric vulnerability and suggest biomarkers for precision-targeted interventions.