Integrative Clustering of Neurobiological and Symptomatic Dimensions Reveals Major Depressive Disorder Subtypes with Brain-Symptom-Treatment Coherence

BACKGROUND: Major depressive disorder (MDD) is a highly heterogeneous psychiatric condition, characterized by substantial variability in both neurobiological and symptomatic dimensions that complicates effective treatment. Despite extensive efforts to parse MDD heterogeneity, robust subtypes linking brain abnormality to depressive symptoms and treatment response have yet to be fully identified. METHODS: This study proposes an integrative clustering framework. A total of 238 patients with MDD, 104 patients with bipolar disorder (BD), and 122 healthy controls (HC) were included. The amplitude of low-frequency fluctuations (ALFF) derived from resting-state fMRI was used to characterize whole-brain activity levels. Across two independent MDD cohorts, we integrated principal components derived from ALFF with five symptomatic factors extracted from the HAMD-17. K-means clustering was performed within each cohort, followed by cross-cohort cluster matching to identify robust MDD subtypes. The neurobiological and symptomatic profiles of each subtype were characterized separately through inter-group comparisons of ROI-level ALFFs and HAMD factor scores. Antidepressant response was compared across subtypes to elucidate their impact on treatment outcomes. External validation was conducted on an independent MDD cohort, where new samples went through subtype assignment followed by inter-group comparisons. RESULTS: Five MDD subtypes with distinct neurobiological, symptomatic, and treatment profiles were identified based on the proposed framework. Cross-cohort consistency was observed in both neurobiological and symptomatic dimensions. The identified subtypes are as follows: (1) a subtype with decreased activity in the forebrain and increased activity in the posterior regions, accompanied by high anxiety and insomnia, as well as a high SSRI response rate (77\%); (2) a subtype with decreased sensory/somatomotor (SSM) activity, high anxiety and insomnia, and a high HAMD total score; (3) a subtype with decreased activity in the visual network (VN) and posterior default mode network (DMN), accompanied by high somatic symptoms and retardation, as well as a high HAMD total score and a low SSRI response rate (22\%); (4) a subtype with subtle abnormalities and mild depressive symptoms; and (5) a subtype with increased memory retrieval (MR) and DMN activity, accompanied by retardation. Core inter-group patterns in HAMD total score and SSRI response rate was reproduced in the validation cohort. The mean ALFF of the DMN and frontoparietal task control (FPTC) networks showed potential as indicators of HAMD total score, while linear correlations were only observed in four of the subtypes. CONCLUSION: Our findings demonstrate that integrating neurobiological and symptomatic dimensions is a promising approach for identifying MDD subtypes and reveal brain–symptom-treatment coherence. The results enhance our understanding of MDD heterogeneity and provide insights into the development of personalized treatment strategies.