Modeling spatial, developmental, physiological, and topological constraints on human brain connectivity

Abstract

The complex connectivity of nervous systems is thought to have been shaped by competitive selection pressures to minimize wiring costs and support adaptive function. Accordingly, recent modeling work indicates that stochastic processes, shaped by putative trade-offs between the cost and value of each connection, can successfully reproduce many topological properties of macroscale human connectomes measured with diffusion magnetic resonance imaging. Here, we derive a new formalism with the aim to more accurately capture the competing pressures of wiring cost minimization and topological complexity. We further show that model performance can be improved by accounting for developmental changes in brain geometry and associated wiring costs, and by using inter-regional transcriptional or microstructural similarity rather than topological wiring-rules. However, all models struggled to capture topologies spatial embedding. Our findings highlight an important role for genetics in shaping macroscale brain connectivity and indicate that stochastic models offer an incomplete account of connectome organization.

Homophilic wiring principles underpin neuronal network topologyin vitro

“…A remaining challenge in the field is to be able to directly parse the extent to how stochasticity versus specific economic trade-offs may, together or independently, influence network outcomes under different conditions. Future methodological work should explore how various multiplicative [45][46][47] or additive 49 GNMs may be used to understand these shaping factors.…”

“…To do this, we tested a number of candidate wiring rules that may best explain the selforganization of cellular-level functional connectivity graphs over time. This was done using a generative network model, which was previously used to probe whole-brain network organization [45][46][47][48][49] . Generative network models develop in silico networks according to an economic trade-off, in which new connections are iteratively formed depending on both the modeled costs and values (Figure 2c).…”

“…Previous studies employing generative models of human macroscopic structural brain organization have shown that generative rules based on homophilic attachment mechanisms can achieve very good model fits [45][46][47]49 (although see 50 ). But what do homophily-based wiring rules essentially entail?…”

“…In the current study, we translate prior GNM research at the level of inter-regional connectivity [45][46][47][48][49][50][51][52] to the microscale (functional connectivity between single neurons) to test whether generative wiring principles are recapitulated across spatial scales. Using HD-MEAs, we acquired electrophysiological network recordings of the spontaneous extracellular neuronal activity from populations of developing primary cells (PCs) derived from dissociated embryonic rodent cortices, three different lines of human induced pluripotent stem cell (iPSC)-derived neurons and more mature sliced human embryonic stem cell (hESC)-derived cerebral organoids (hCOs).…”

“…That is, nodes within simulated and observed networks could have different local relationships to one another, because node-wise local organizational properties are not captured per se by the existing energy equation. Local organizational properties have previously been investigated in terms of how well generative models can recapitulate the locations of organizational features, such as hub-nodes, in the C. elegans connectome73 or MRI-inferred human brain networks46,49,50 .For example, consider the topological relationship between central and peripheral nodes commonly found in the classical representation of a brain network. Nodes which score highly in centrality measures (e.g.…”