Modular organization and selective motifs in the insula provide structural priors for efficient learning

Abstract

The insular cortex (IC) integrates diverse sensory and interoceptive signals to support emotional and cognitive functions, exhibiting topologically functional distinctions. Although long-range pathways of the IC have been extensively mapped, the contribution of intra-IC architecture to network-level information processing remains poorly understood. Here we reconstructed single-cell projectomes of 2,267 IC neurons, generating a comprehensive cellular-resolution map of intra-IC connectivity. Graph-theoretic analyses revealed a hierarchical, topographically organized modular network with a hub-and-spoke organization and selective motif enrichment. Embedding these anatomical priors into recurrent neural network models demonstrated that intrinsic IC topology confers a distinct computational advantage: across multiple cognitive tasks, IC-initialized networks learned faster and exhibited better robustness to perturbations than those initialized with shuffled, randomized, or neighboring somatosensory cortical architectures. Together, these findings identify the IC as a structurally specialized cortical hub whose internal wiring supports efficient learning and provides design principles for brain-inspired networks.