SPATIAL EMBEDDING AND HIERARCHICAL FEATURES OF THE WHOLE-BRAIN NETWORK COMPRISING CORTICAL AND SUBCORTICAL STRUCTURES

The primate cerebral cortex is classically modeled as a spatially embedded hierarchical system governed by an Exponential Distance Rule (EDR) (Ercsey-Ravasz et al. 2013). However, this cortico-centric view neglects the relatively weak but nevertheless crucial inputs from subcortical structures.
In order to build a whole brain connectome modeling how the subcortical structures are integrated into the cortical network, we constructed a weighted and directed whole-brain connectome using quantitative retrograde tracing from ~120 macaque cortical injections, allowing us to determine the rules of subcortical-cortical interaction (Markov et al. 2014).
Principal Component Analysis (PCA) confirms that cortical source areas cluster into distinct functional clusters. Procrustes analysis of cortical input and output profiles reveals a high degree of correlation (~0.87) where, the residual asymmetry in the Procrustes space is associated with high multimodality. We observe a significant divergence in the fingerprints of subcortical inputs to cortex. PCA of subcortical inputs recover similar cortical functional clusters as found in the cortico-cortical network. Subcortical projections to cortex are non-EDR and constitute two groups: (i) non-thalamic and thalamic non-relay nuclei tend to project to non-functionally clustered cortical targets; (ii) Thalamic relay nuclei projections to cortex are highly symmetrical, project to functionally clustered targets and specific hierarchical levels.
We identify potential fundamental architectural principles for different networks in terms of spatial embedding, reciprocity, functional clustering and hierarchical targeting. The whole-brain network that we are building supports predictive coding framework where subcortical inputs regulate the precision of cortical inference (Bastos et al. 2012; Hou et al. 2025).