Identifying neurophenotypes of major depressive disorder through normative model of regional homogeneity.

Major depressive disorder (MDD) is a highly heterogeneous mental illness, marked by clinical variability and distinct neuropathological mechanisms. This study sought to enhance diagnostic precision for MDD by identifying neurophenotypes using a normative model of regional homogeneity (ReHo), offering new avenues for precision medicine. Using resting-state functional magnetic resonance imaging data from 1101 patients with MDD and 1011 healthy controls (HCs) sourced from the REST-meta-MDD project, we developed the first normative model of ReHo. Gaussian process regression was applied to predict a normative lifespan trajectory based on age and sex in HCs, enabling quantification of individual deviations from the model for patients with MDD. Unsupervised clustering algorithms were then employed to classify MDD subtypes, followed by validation analyses to assess clustering stability. Significant deviations from the normative ReHo model were observed in patients with MDD. Two distinct MDD subtypes emerged: Emotional dysregulation subtype, characterized by negative deviations in the frontoparietal control network, ventral attention network, default mode network, and limbic network (Cohen's d = 0.40-1.75, FDR-corrected p < 0.05). This subtype correlated with more severe overall depressive symptom (d = 0.17, p = 0.010), better insight (d = -0.25, p = 0.009), younger age (d = -0.19, p = 0.003), lower medication usage (Cramer's V = 0.09, p = 0.017), a negative correlation between symptom severity and illness duration (r = -0.21, p < 0.001), severe brain dysfunction (Partial η