Subtyping OCD based on individual symptom networks: subtype-specific neural dynamics and morphometric similarity linked to molecular profiles.

Obsessive-compulsive disorder (OCD) is clinically heterogeneous, posing challenges for diagnosis and treatment. Individual symptom network-based subtyping could provide the framework for characterizing this heterogeneity. This study developed an individualized symptom network (ISN)-based subtyping framework, leveraging a healthy control dataset (n = 3227) to model the normative symptom network. Based on the prominence of obsessive-compulsive (OC) symptoms within ISN (Expected Influence within ISN, EIISN), three OCD subtypes were identified in the discovery cohort (n = 238). Subtype 1 (39.50%) characterized by checking/obsessing, subtype 2 (6.30%) by ordering/washing and subtype 3 (54.20%) by hoarding, validated in another dataset (n = 79), the follow-up data from the discovery cohort (n = 109) and a subsample of the discovery cohort assessed using other OC symptom scale (n = 83). Nuroimaging analysis revealed distinct brain dynamics and morphometric similarity: subtype 1 exhibited more frequent transitions from the activated visual to frontoparietal control networks and higher morphometric similarity within the default mode network (DMN) compared to subtype 3. Subtype 2 exhibited increased regional intrinsic activity in the inferior frontal gyrus (opercular part) and the postcentral gyrus, with the lowest transitions from the activated somatomotor to the ventral attention networks. Subtype 1 also exhibited significantly higher morphometric similarity within the DMN and its connections to the dorsal/ventral attention networks compared to subtypes 2 and 3. Moreover, these neural features are associated with symptoms. Neurotransmitter-related subtype-specific dynamics to GABAa and μ-opioid, while transcriptomic analyses associated subtype-specific morphometric similarity with IL-1β and TNF-α. Together, biologically grounded 3 OCD subtypes based on EIISN exhibit replicability across time, samples and assessment scales and offer an individual framework for decoding OCD heterogeneity.