Individualized structural brain network and transcriptomic characterization of heterogeneity in type 2 diabetes mellitus

Type 2 diabetes mellitus (T2DM) is a heterogeneous metabolic disorder that also affects the brain, yet prior neuroimaging studies have predominantly focused on group-level comparisons, obscuring interindividual variability in brain structural organization. To capture this heterogeneity, we constructed individualized differential structural covariance networks for 156 individuals with T2DM using 148 healthy controls as a normative reference. Unsupervised clustering based on these individualized networks revealed two neuroanatomical subtypes with distinct network configurations. Most subtype-differentiating edges showed higher values in one subtype, whereas a smaller set of edges, primarily involving the basal ganglia and cerebellum, showed higher values in the other. Integration with transcriptomic data from the Allen Human Brain Atlas identified genes spatially associated with subtype-related differences. These genes were enriched in processes including receptor tyrosine kinase signaling and glycerolipid metabolism, with preferential expression in the striatum and cerebellum. Together, these findings characterize interindividual variation in brain structural covariance in T2DM and identify candidate molecular processes associated with the observed subtype differences.