Data was used from within the UK Biobank. Usable magnetic resonance imaging (MRI) data of the brain were available from 37,330 individuals, each coupled with genetic information and extensive electronic healthcare records. From the imaging data, measures relating to four regions have been extracted, the regions included: hypothalamus (HT), pituitary gland (PG), olfactory bulbs (OB), and hypothalamus grey matter (HTGM). This work was done in collaboration with Professor Stephen Smith’s group at the University of Oxford. For each of these regions, numerous quantitative image-derived phenotypes were extracted relating to both volume and intensity. The extraction of each of these phenotypes was done at two different confidence thresholds, 0 and 0.3. Going forward with analysis, I utilised the more stringent 0.3 threshold to ensure confidence in the values. Further I decided to focus on volumetric measures. This was to ensure biological interpretability and streamline analysis to enable adequate bandwidth for analytical follow-up.
Distribution of these values was assessed across sex and age groups within the population.
Firstly genome-wide association analyses (GWAS) were run for each of the four volumetric phenotypes relating to the HT, PG, OB, and HTGM. GWAS were run on the most densely populated well-mixed population which here largely aligns with those that self-identify as White-European. Age, sex, assessment centre, genetic principal components and a set of technical measures relating to the MRI acquisition (table position, head movement etc.) were included as covariates in an additive linear model. Following selection for independent loci using conditional joint analysis, >50 significantly associated loci were discovered to be associated with one or more of the phenotypes. All GWAS were deemed well controlled for residual population structure and inflation upon assessment of linkage-disequilibrium scores. Phenome-wide association studies were run across each of the genetic discovery sets using ICD-10 codes from within the UK Biobank, discovering three statistically significant associations between disease codes and SNPs associated with HTGM volume. Sexual dimorphism was seen in the effect sizes and a statistically significant sex difference was seen in the effect size of two genetic variants, one associated with HT volume, and the other associated with PG volume. Heritability of each trait was also calculated and interestingly the heritability varied across the sex-stratified populations. Exome-wide association analysis was further run, with a single gene found to be statistically significantly associated with hypothalamus volume.
To fully take advantage of the dimensionality of the data available in the original source images from which the image-derived phenotypes were extracted, an extension of the genetic analysis was completed. We took the set of genetic variants discovered during the four GWAS and built linear models to interrogate their associations with structural volume differences at voxel-level. The volume differences between individuals were captured with Jacobian determinant maps, which were estimated between each individual and a reference average brain. Each voxel in these maps represents an expansion or contraction at a spatial location relative to a reference average brain. Different statistical methods were trialed that encoded genotypes as continuous and discrete variables. This analysis allowed us to visualise associations between the discovered genetic variants and the morphology of the brain at a high spatial resolution, indicating effects at additional brain regions not captured by the previously extracted IDPs. Interpretation of this was aided by collaborators at the university of Oxford and the University of Heidelberg.
