Discovering Conserved Properties of Brain Organization Through Multimodal Integration and Interspecies Comparison

Abstract

The primate cerebral cortex is broadly organized along hierarchical processing streams underpinned by corresponding variation in the brain's microstructure and interareal connectivity patterns. Fulcher et al. recently demonstrated that a similar organization exists in the mouse cortex by combining independent datasets of cytoarchitecture, gene expression, cell densities, and long-range axonal connectivity. Using the T1w:T2w magnetic resonance imaging map as a common spatial reference for data-driven comparison of cortical gradients between mouse and human, we highlighted a common hierarchical expression pattern of numerous brain-related genes, providing new understanding of how systematic structural variation shapes functional specialization in mammalian brains. Reflecting on these findings, here we discuss how open neuroscience datasets, combined with advanced neuroinformatics approaches, will be crucial in the ongoing search for organization principles of brain structure. We explore the promises and challenges of integrative studies and argue that a tighter collaboration between experimental, statistical, and theoretical neuroscientists is needed to drive progress further.

Keywords: data integration; interspecies comparison; neuroinformatics; spatial embedding.

Figure 1.

A low-dimensional projection of mouse brain cortical areas, represented as a multimodal feature vector. The feature vector combines properties of gene expression (Pvalb, Grin3a, and Grik2), cytoarchitecture type, weighted axonal in-degree, T1w:T2w, estimated hierarchical level, and mean density of parvalbumin-containing cells. Brain areas with a similar set of properties are close in this projection space. Shading has been added manually to highlight the different functional families, labeled according to Harris et al.

Figure 2.

Pvalb expression is strongly correlated with an independent measure of parvalbumin containing cell density across the mouse cortex. We plot Pvalb expression (z-scored expression energy estimated from coronal section data from the Allen Mouse Brain Atlas) as a function of direct quantitative measurement of parvalbumin containing cell density across cortical areas. The 2 measurements are highly correlated, Spearman ρ = 0.95, increasing from prefrontal and anterolateral areas through to somatomotor areas (brain areas are colored as labeled in Figure 1).