Network controllability mediates the relationship between rigid structure and flexible dynamics

Shi Gu, Panagiotis Fotiadis, Linden Parkes, Cedric Huchuan Xia, Ruben C. Gur, Raquel E. Gur, David R. Roalf, Theodore D. Satterthwaite, Dani S. BassettNetwork 2021 · DOI

TL;DR

This study bridges the gap between the brain's fixed structural wiring and its dynamic functional activity by showing how a region's structural "controllability" (its strategic position in the network) mediates its ability to flexibly change its functional connections over time. This offers a mechanistic framework for understanding how rigid anatomy supports diverse cognitive functions.

What Problem Does This Paper Address?

The human brain is a complex system where rigid anatomical connections (white matter tracts) give rise to incredibly diverse and flexible functional dynamics (how brain regions communicate). Understanding precisely how this fixed structure supports such dynamic behavior has been a major challenge. Existing research has struggled to create theoretical frameworks that link the static structural network to the ever-changing patterns of neural activity. This paper aims to bridge this gap by investigating how network control theory can explain the relationship between structural properties and functional flexibility.

Methods

The researchers employed a network neuroscience approach, using graph theory to model brain connectivity. They utilized diffusion tensor imaging (DTI) to map structural connectivity and resting-state functional magnetic resonance imaging (fMRI) to capture functional connectivity. Key metrics were calculated:

  • Structural Participation Coefficient: Measures how a region's structural connections are distributed across different functional modules (groups of brain regions).
  • Boundary Controllability: A metric from network control theory that quantifies a region's ability to influence the overall brain network's state, particularly for regions located at the boundaries of modules.
  • Functional Participation Coefficient: Similar to the structural version, but for functional connections.
  • Modular Flexibility: Assesses how frequently a brain region changes its functional module assignment over time, indicating its dynamic adaptability.

They used statistical analyses, including bivariate correlations, multiple linear regression, and serial mediation models, to test their hypotheses about how these metrics relate to each other.

Datasets

The study used data from the Philadelphia Neurodevelopmental Cohort (PNC), comprising 823 healthy individuals aged 8 to 23 years. This large developmental cohort allowed for the examination of age-related changes and individual differences.

Key Insights and Ideas

  • Structure Dictates Dynamics: The study strongly supports the idea that the brain's structural architecture fundamentally constrains and enables its functional dynamics.
  • Strategic Location Matters: A region's "boundary controllability" – its position at the intersection of different functional modules – is crucial for its ability to flexibly shift its functional allegiance. Regions that can "reach" across modules structurally are better positioned to do so functionally.
  • Mediation is Key: The relationship between a region's structural connectivity profile (participation coefficient) and its functional flexibility is not direct but is mediated by its boundary controllability and its functional participation coefficient. This suggests a step-wise transfer of influence from structure to dynamic function.
  • Developmental Trends: Both structural and functional flexibility metrics generally decrease with age, with structural markers showing larger effect sizes. This suggests a shift towards more specialized and less flexible network organization as individuals mature.

Key Results

  • A region's structural participation coefficient was positively correlated with its boundary controllability (r = 0.59, p < 10⁻⁵).
  • A region's functional participation coefficient was strongly positively correlated with its modular flexibility (r = 0.78, p < 10⁻⁵).
  • Crucially, boundary controllability and functional participation coefficients serially mediated the relationship between structural participation coefficients and modular flexibility. This means that a region's structural connectivity profile influences its boundary controllability, which in turn influences its functional participation coefficient, ultimately shaping its modular flexibility.
  • Across subjects, structural participation coefficient was positively correlated with boundary controllability and functional participation coefficient, but not directly with modular flexibility. However, the serial mediation effect was still observed.
  • All structural and functional metrics (structural participation coefficient, boundary controllability, functional participation coefficient, and modular flexibility) declined linearly with age.

Limitations and Future Work

The authors acknowledge that their study was cross-sectional, limiting their ability to definitively establish temporal precedence. While they used a method to account for temporal directionality in their mediation analyses, future longitudinal studies are needed to fully understand how these properties change over time. They also suggest incorporating clinical and neurocognitive data to explore how these structural-functional relationships influence behavior and how disruptions might relate to neurological and psychiatric disorders. Additionally, exploring the impact of different parameter choices for modular flexibility calculations and investigating individual differences in these parameters are areas for future research.

Summary updated

Abstract

Precisely how the anatomical structure of the brain supports a wide range of complex functions remains a question of marked importance in both basic and clinical neuroscience. Progress has been hampered by the lack of theoretical frameworks explaining how a structural network of relatively rigid interareal connections can produce a diverse repertoire of functional neural dynamics. Here, we address this gap by positing that the brain's structural network architecture determines the set of accessible functional connectivity patterns according to predictions of network control theory. In a large developmental cohort of 823 youths aged 8 to 23 years, we found that the flexibility of a brain region's functional connectivity was positively correlated with the proportion of its structural links extending to different cognitive systems. Notably, this relationship was mediated by nodes' boundary controllability, suggesting that a region's strategic location on the boundaries of modules may underpin the capacity to integrate information across different cognitive processes. Broadly, our study provides a mechanistic framework that illustrates how temporal flexibility observed in functional networks may be mediated by the controllability of the underlying structural connectivity.
BibTeX
@inproceedings{gu2021network,
  author = {Shi Gu and Panagiotis Fotiadis and Linden Parkes and Cedric Huchuan Xia and Ruben C. Gur and Raquel E. Gur and David R. Roalf and Theodore D. Satterthwaite and Dani S. Bassett},
  title = {{Network controllability mediates the relationship between rigid structure and flexible dynamics}},
  booktitle = {Network},
  year = {2021},
  doi = {10.1162/netn\_a\_00225},
}
Figure