Biotypes of deeply phenotyped depressed patients reflect signatures of adverse childhood experience and depressive cognitive biases

Background Negative Cognitive Styles (NCS) are key features of depression contributing to severe clinical outcomes by sustaining negative affect. However, depression is clinically heterogeneous, reflecting complex neurobiological and environmental interactions. Characterizing heterogeneity using multimodal data could help identify mechanisms mapping onto different phenotypes and discover high-impact biomarkers. Methods Using a stability-based relative clustering validation pipeline, 344 depressed patients (135 major depressive disorder, 209 bipolar disorder) were stratified based on multimodal neuroimaging data, including grey matter volumes, cortical thickness, white matter diffusivity indices, resting-state functional connectivity (FC), and spontaneous neural activity. Clusters were derived from each modality separately and in combination, and profiled for Adverse Childhood Experiences (ACEs) and NCS. Results All neuroimaging modalities stratified depressed patients into two clusters, with the FC-based model achieving the highest accuracy (85 %). Compared with healthy controls (HC, N = 138), these clusters exhibited opposite patterns of global functional hyper-integration (Cluster 1) and hyper-segregation (Cluster 2). Unique multivariate neurobiological-ACEs relationships characterized the FC-based clusters. While ACEs negatively affected FC in Cluster 1 and HC, Cluster 2 showed opposite effects specific to ACE subtypes, with sexual abuse positively influencing FC. Mediation analysis showed that FC strength mediated the relationship between ACEs and NCS of overgeneralize only in Cluster 2. ACE-related functional alterations pinpoint brain regions involved in socioemotional and cognitive development, sensitive to maturational neural changes. Conclusions These findings provide evidence of clinically meaningful depression “biotypes” shaped by ACEs and associated with NCS, underlining the feasibility of computational psychiatry tools to uncover data-driven patterns for precision psychiatry.