A weighted generative model of the human connectome

Abstract

Probabilistic generative network models have offered an exciting window into the constraints governing the human connectomes organization. In particular, they have highlighted the economic context of network formation and the special roles that physical geometry and self-similarity likely play in determining the connectomes topology. However, a critical limitation of these models is that they do not consider the strength of anatomical connectivity between regions. This significantly limits their scope to answer neurobiological questions. The current work draws inspiration from the principle of redundancy reduction to develop a novel weighted generative network model. This weighted generative network model is a significant advance because it not only incorporates the theoretical advancements of previous models, but also has the ability to capture the dynamic strengthening or weakening of connections over time. Using a state-of-the-art Convex Optimization Modelling for Microstructure-Informed Tractography (COMMIT) approach, in a sample of children and adolescents (n = 88, aged 8 to 18 years), we show that this model can accurately approximate simultaneously the topology and edge-weights of the connectome (specifically, the MRI signal fraction attributed to axonal projections). We achieve this at both sparse and dense connectome densities. Generative model fits are comparable to, and in many cases better than, published findings simulating topology in the absence of weights. Our findings have implications for future research by providing new avenues for exploring normative developmental trends, models of neural computation and wider conceptual implications of the economics of connectomics supporting human functioning.