Major depressive disorder on a neuromorphic continuum

The heterogeneity of major depressive disorder (MDD) has hindered clinical translation and neuromarker identification. Biotyping facilitates solving the problems of heterogeneity by dissecting MDD patients into discrete sub-groups. However, interindividual variations suggest that depression may be conceptualized as a continuum rather than as a category. We use a Bayesian model to decompose structural MRI features of MDD patients from a multisite cross-sectional cohort into three latent disease factors (spatial pattern) and continuum factor compositions (individual expression). The disease factors are associated with distinct neurotransmitter receptors/transporters obtained from open PET sources. Increases in cortical thickness in sensory and decreases in orbitofrontal cortices (Factor 1) associate with norepinephrine and 5-HT2A density; decreases in the cingulo-opercular network and subcortex (Factor 2) associate with norepinephrine and 5-HTT density; and increases in social and affective brain systems (Factor 3) relate to 5-HTT density. Disease factor patterns can also be used to predict depressive symptom improvement in patients from the longitudinal cohort. Moreover, individual factor expressions in MDD are stable over time in a longitudinal cohort, with differentially expressed disease controls from a transdiagnostic cohort. Collectively, our data-driven disease factors reveal that patients with MDD organize along continuous dimensions that affect distinct sets of regions.