Normative modeling of brain morphology reveals neuroanatomical heterogeneity and biological subtypes in major depressive disorder

Abstract

BACKGROUNDMajor Depressive Disorder (MDD) is characterized by high neurobiological heterogeneity, which hinders precise diagnosis and treatment. Traditional group-level neuroimaging analyses fail to capture individual differences, while normative modeling offer a promising approach to quantify individual deviations from healthy brain structure patterns, facilitating the identification of biological subtypes and offering a data-driven framework to dissect this heterogeneity. METHODSUsing 1,190 healthy controls, we constructed normative developmental trajectories of gray matter volume (GMV) across 246 Brainnetome-defined regions using Bayesian linear regression. Deviation maps were derived for 398 MDD patients. k-means clustering was employed to identify GMV-based biotypes. Then, the clinical characteristics and anatomical differences among these subtypes were explored, along with the post-treatment clinical features and treatment responses of participants who completed the 8-week antidepressant treatment within each subtype. RESULTSPatients with MDD exhibited widespread yet individually variable GMV deviations. Clustering analysis revealed two subtypes: Subtype 1 displayed predominantly negative deviations in sensorimotor and occipital cortices, whereas Subtype 2 showed widespread positive deviations in temporal and posterior cingulate regions. Subtype 1 had higher extraversion and symptom-linked deviation patterns; in Subtype 2, deviation burden correlated with generalized anxiety. Longitudinally, Subtype 1s GMV deviation changes predicted symptom improvement, while Subtype 2s deviations correlated with baseline severity. CONCLUSIONSNormative modeling of GMV reveals marked neuroanatomical heterogeneity in MDD and identifies subtypes with distinct clinical and treatment-related characteristics, laying a foundation for precision psychiatry and individualized interventions.