Abstract
Cognitive ability (or the g-factor) enables individuals to make sense of the world and to navigate it appropriately. Cognitive ability is typically assessed through various cognitive tests of memory, processing speed, abstract reasoning, reading abilities and more. Research has suggested that individuals with good cognitive ability tend to be healthier, have higher income, live longer, and retain more autonomy in old age. Hence, a comprehensive understanding of the drivers of cognitive ability and its age-related decline may lay the foundations for interventions that have the potential to considerably improve quality of life across the general population.Research has devoted much attention to so-called brain networks of interconnected brain regions that may collectively underpin cognitive ability. Brain morphometry in the central executive brain network, for example, was demonstrated in a well-powered study to play a centrally important role in cognitive ability (Madole et al., 2021). However, different brain morphometry studies have delivered conflicting evidence and it remains unresolved whether structural brain networks reliably correlate with cognitive ability. One reason for the differences in results might be the lack of methodological consensus between studies. Methodological choices that dictate study results include covariate adjustment for brain size – which some studies perform, and others do not – as well as brain atlases used to subdivide the participants brain images into regions. Therefore, there is a need for hypothesis-free exploratory work to help make optimal analytical decisions and to perform meaningful hypothesis tests. This exploratory groundwork should help establish more reliable correlations between brain morphometry and cognitive ability.
This thesis has two major aims: The first aim is to perform exploratory studies to deliver evidence in support of study designs optimal for the investigation of the relationship between cognitive ability and brain morphometry. The second aim is to investigate the multidimensional relationships among structural brain networks, cognitive ability, and age-related processes. This thesis aims to present comprehensive models of cognitive ability and its associated biology by systematically evaluating the impact of certain methodological decisions. Genetic analysis techniques are employed to triangulate phenotypic analyses using innovative and biologically-informed technology. Presented studies analysed genetic data and structural MRI data in two large samples: the UK Biobank (N ~ 40,000) and the Adolescent Cognitive Brain Development study (N ~ 10,000).
In Chapter 2 I aim to characterise the impact of brain size covariate adjustment on the relationships between cortical brain volumes and cognitive ability. Results indicated that the relationship between regional volumes and cognitive ability is closely entangled with brain size to the extent that their relationship cannot be reliably modelled when made statistically independent of brain size. This study delivered evidence that instead of assessing region-by-region correlations between brain morphometry and cognitive ability, multivariate study designs may help to account holistically for the complex biological underpinnings of cognitive ability. The study provides empirical and theoretical arguments that brain size adjustment induces collider bias in genome-wide analyses.
In Chapter 3 I derive and validate a novel multivariate framework – Genomic Principal Component Analysis (PCA) – that integrates multiple traits as well as genome-wide information. Genomic PCA takes genome-wide association data as input to extract genetic principal components (PCs) underlying multiple phenotypes (Genomic PCA does not capture ancestral PCs of shared genetic make-up between individuals). I use Genomic PCA in Chapter 4 to model genome-wide PCs underlying multiple brain regions that are part of canonical brain networks. The study demonstrates that genome-wide PCs underlying nine canonical brain networks – unadjusted for brain size – are significantly correlated with cognitive ability and brain ageing. However, this study finds no evidence for localised brain network-specific correlates of cognitive ability, as the central executive network is not any more associated with cognitive ability than other brain networks. The results suggest that genetic correlates of brain morphometry relate to cognitive ability through general brain-wide features shared among multiple regions that are not specific to brain networks.
In Chapter 5 I compare brain atlases that are commonly used to subdivide study participants’ brain images into regions-of-interest. The brain atlas comparison in this study relies on multivariate prediction models of multiple behavioural traits, including cognitive ability. This study finds that using fine-grained brain atlases maximises the relationship between brain morphometry and cognitive ability. Future studies may be able to model more robust brain trait associations by adopting multivariate models of hundreds of thousands of vertex-wise brain measurements. The final chapter discusses the overarching implications, limitations, and future directions of these findings that should motivate multidisciplinary approaches to more comprehensively account for the complex web of biological factors that give humans advanced cognitive ability.
| Date of Award | 1 Sept 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | James Cole (Supervisor), Stuart Ritchie (Supervisor) & Claire Steves (Supervisor) |