Compressing the Brain: Bridging Scales, Phenotypes, and Development

Compression and dimensionality reduction are tools through which we can recreate complex endpoints from simple underlying principles. But how helpful are these tools for understanding the reorganisation of structural and functional connectivity across childhood and adolescence? In Chapter 1, I introduce key concepts in developmental systems neuroscience, such as equifinality and multi-finality, alongside the need for a trans-diagnostic cross-modal integrated neuroscience with increased focus on individual-level heterogeneity, as opposed to group-level case-control comparisons. To address these theoretical considerations, I explore compression and dimensionality-reduction techniques as a lens through which we may systematically test, under different conditions, the extent to which group-level developmental principles apply to the individual, across development. In Chapter 2, using a population-level modelling approach, I examine the link between generative wiring principles underlying structural connectivity emergence, namely an economic trade-off, in preadolescent children with their genetic propensity for high cognitive ability. The cost penalty within this generative model varies as a function of polygenic scores for general intelligence, and converges on overlapping genetic ontologies, where children with a particularly strong genetic propensity for high cognitive ability exhibit simulated networks with significantly softer wiring constraints, resulting in a more randomised topology, and thus increased stochasticity and simulated global efficiency. In Chapter 3, across typically-developing and neurodivergent developmental cohorts, I establish organisational principles underpinning structural and functional connectivity variability across childhood and adolescence. I used diffusion-map embedding to derive low-dimensional manifolds or axes, constituting gradients of variability. In contrast to the literature, I demonstrate that such gradients are temporally stable and are refined with age but not reordered. To explore interactions between structural and functional manifolds, I propose a novel manifold-based measure of structure-function coupling, extensively benchmarked with prior measures, and demonstrate that such coupling is significantly predicted by measures of cognition, but not psychopathology, in a developmentally-sensitive manner. In Chapter 4 I take a third approach to compression, this time to focus on compressing brain-behaviour associations to establish trans-diagnostic links between the developing brain and developmentally relevant behaviours. To do this I model the predictive link between resting-state functional connectivity and item-level psychopathology, as a latent variable, through a partial least squares analysis in a cohort at-risk of neuro-developmental conditions, spanning childhood and adolescence. Through examining differential participant-level expression of this latent variable, I find that the latent construct transcended traditional diagnostic borders, revealing a neurotypical-neurodivergent continuum. Using meta-analytic functional activation decoding, functional connectivity associated with heightened risk for psychopathology overlaps with regions related to executive functioning, whilst a protective effect against psychopathology is linked to functional activations related to language. I conclude that the predictive link between functional connectivity and psychopathology is constrained by underlying macroscale and microscale organisational hierarchies, and it aligned with a somatosensory-association axis. In Chapter 5, I contextualise my findings and consider future research directions to enhance translation. Together, this thesis employs a multi-modal approach to chart the organisational principles underlying brain development in childhood and adolescence, their phenotypic consequences, and their relationship to underlying group-level macroscale and microscale hierarchical constraints and genetic influences.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.