Molecular Mechanisms Explaining Neuroanatomical Subtypes in Major Depressive Disorder: Insights From Cortical Morphometric Inverse Divergence.

Major depressive disorder (MDD) exhibits substantial neurobiological heterogeneity that complicates treatment selection and mechanistic understanding. While conventional group-level analyses identify diverse structural alterations, they obscure clinically relevant individual differences. We employed heterogeneity through discriminant analysis (HYDRA) clustering to decompose morphometric inverse divergence (MIND) network patterns into distinct neuroanatomical subtypes and examined their molecular underpinnings. We analyzed MIND network data from 240 Japanese individuals with MDD and 367 healthy controls using unsupervised clustering. Subtype-specific alterations were mapped onto neurotransmitter receptor density distributions, and transcriptomic data from the Allen Human Brain Atlas were integrated using partial least squares regression. Two neuroanatomically distinct subtypes emerged. Subtype 1 (n = 78) exhibited widespread increases in MIND strength across all Yeo networks, with predominant serotonergic, dopaminergic, and GABAergic associations. Gene expression analysis revealed SST and CUX2 correlations, with enrichment for metal ion homeostasis and circadian rhythm pathways. Subtype 2 (n = 162) showed reduced MIND strength in dorsal attention, somatomotor, frontoparietal, limbic, and default networks, with glutamatergic, cannabinoid, and dopaminergic dysfunction. This subtype demonstrated negative CRH correlations and enrichment for glutamatergic signaling and calcium/cAMP-mediated processes. Our findings demonstrate systematic decomposition of MDD heterogeneity into distinct neuroanatomical subtypes with unique molecular signatures. The identification of subtype-specific neurotransmitter profiles and transcriptomic architectures provides mechanistic insights into MDD heterogeneity, offering potential for biomarker-guided treatment selection and personalized therapeutic approaches.