Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

In this thesis, we performed a genome-wide association study of the human brain cortical thickness, which is based on a data-driven, hierarchical global-to-local cortical segmentation. The UKB dataset of nearly 20,000 samples was used as a discovery cohort for training the segmentation and association tests. To gain insights into the genetic architecture of the cortical thickness, the univariant GWAS of the entire G2L segmentation was performed using PLINK by treating the thickness of each segment as an independent trait. And then, the result of GWAS was further analysed with FUMA, LDSC, to map the loci, perform the pathway enrichment and test the genetic correlation with other traits. To evaluate the G2L segmentation, we compare it with the traditional Desikan-Killiany atlas, the Glasser atlas and the Destrieux atlas, the Desikan-Killiany atlas-based GWAS study was also compared. The result shows that the cortical thickness's genetic architecture is highly polygenic, and the scope of effect of the variants also varies, from the global average thickness to the tiny segments. The results of the genetic correlation are also in general agreement with previous morphological studies. Compared to the traditional atlas, G2L segmentation offers a rather different segmentation method, providing a global-to-local perspective that captures more loci that are significantly associated with cortical thickness.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Viruses have received a lot of media attention lately due to the SARS-CoV-2 pandemic, including extensive coverage of viral evolution and spread. In this thesis methods to analyze viral spread have been investigated. Viruses are usually not considered to be living organisms because they do not have their own metabolism. However, they are very good in hijacking the metabolism of other living organisms (plants and animals - with us humans just being another kind of animal). Viruses consist mainly of DNA or RNA often protected by a cover (a capsid). RNA and DNA consist of four main building blocks called bases (these are the well known letters A, T/U, C and G). These bases can change (mutate) into each other, in viruses these mutations happen regularly. To characterize a mutation we specify where in the DNA the mutation happened and what the initial and resulting base was. In order to gain epidemiological and evolutionary insights into viruses, we often want to reconstruct their evolutionary history based on the sequence similarity. This reconstruction is a phylogeny that tells us how the viruses we observed today are related. Since we cannot observe the past directly we can only work our way back in the phylogeny and find the most likely path the virus took to get where it is today. Like this we "reconstruct" the ancestral locations of the virus. Once we have these ancestral locations we can count how often the virus spread from one location to another (by following the lines connecting the nodes in a pedigree - e.g. a line from parent to child). Finally, we can look closer at what the pairs of states with many transitions have in common and draw conclusions on what factors are involved in viral spread. This could be geographical distance, shared borders or mobility fluxes to name a few. In this thesis we made a web application that is accessible online and can also be used via your phone to explore drivers of spatial spread. It is accessible via this link: https://gentle-taiga-91395.herokuapp.com/. The application takes a phylogeny as described above and reconstructs the ancestral states, followed by counting the transition events. Finally various analysis are carried out that inform the user of the application on which factors could be involved in viral spread. These factors can then be included in further analyses to get a clearer picture about the most relevant ones for viral spread.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Antibiotics are essential tools in the modern world; since their discovery and introduction into medical practices they have saved the life of billions, contributing to an 8-year increase of human life expectancy between 1944 and 1972. However, genetically encoded antibiotic resistance is threatening the efficacy of many antibiotics and is currently one of the most pressing problems that humanity faces. Antibiotic resistance can spread among bacterial strains through Horizontal Gene Transfer (HGT) and is maintained in populations through selection. This leads to the existence of multidrug resistant bacterial strains, commonly known as superbugs, for which therapeutic options are limited and that cause increased morbidity and mortality. In Klebsiella pneumoniae, resistance to beta-lactam antibiotics is often mediated by extended-spectrum beta-lactamases (ESBLs), carried on plasmids and mobilised through Mobile Genetic Elements (MGE) such as insertion sequences (IS). Understanding the dynamics and modalities behind these associations and mobilisations is a fundamental step in the efforts against antimicrobial resistance. The following dissertation is an explorative analysis of a large and unbiased dataset of more than 5,000 samples of Klebsiella pneumoniae assemblies retrieved from public databases, to determine the prevalence and diversity of ESBLs and their associations with IS. Particular attention was dedicated to the families of insertion sequences IS6 and IS1380, especially to ISecp1, which belongs to the latter. Input sequences were explored and compared against appropriate database to locate and annotate resistance determinants, insertion sequences and replicons; assembly’s contigs were classified as plasmid-derived or chromosome-derived. Subsequently, data was integrated to provide information regarding the occurrence with which specific resistance genes were found close to specific insertion sequences. Potential recent transposition events from plasmid to chromosome were also investigated for the associations pair ISecp1-blaCTX-M-15, which is of clinical relevance in this context.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Alzheimer’s Disease is an increasingly prevalent and debilitating reality in today’s aging society. However, despite the resources devoted to its research and clinical trials, candidate therapies aimed at halting or slowing AD progression continue to fail. Although the reasons for this are inherently multi-faceted, a major driver of these frustrations is late initiation of treatment in the time course of AD development. The deceptively long duration of the preclinical phase of AD—the stage of the disease in which AD brain pathology is present but cognitive defects are not—presents a major hurdle in the fostering of a cohesive understanding of the underlying basis and timeline of molecular and pathophysiological changes in AD. Preclinical AD therefore provides a crucial and opportune window for clinical intervention in order to halt or delay cognitive decline. In this study, we carried out a series of RNA-Seq analyses in a deeply phenotyped, cognitively intact Flemish cohort (n=103, age at baseline: 70 (56-80) years old) to further understand the preclinical phase of AD. These individuals were specifically stratified based on high risk of AD development (i.e. APOE e4 and BDNF 66met carriership) to provide insight into how these risk factors might contribute to AD-associated pathological and cognitive outcomes in a preclinical setting. To this end, we integrated neuroimaging data and cognitive decline with transcriptomic information obtained from peripheral whole blood; the longitudinal nature of the data conferred a valuable temporal layer to the study design. Through the combination of differential expression analysis and WGCNA analysis, we garnered insight into individual genes, co-expressed networks of genes, and pathways that may be peripherally de-regulated due to carriership of risk alleles and/or during the onset of AD-related pathology (i.e. amyloid accumulation in the brain) and cognitive decline. Notably, WGCNA preservation analyses between APOEe4 carriers versus non-carriers and between amyloid accumulators versus non-accumulators revealed marked alterations in gene co-expression patterns within the peripheral blood of these subgroups. In numerous analyses, such as in BDNF 66met carriership, APOE e4 carriership, and SUVR analyses, we saw a direct activation of explicitly neurodegenerative-labelled pathways. In addition to these specific pathways, however, other notable pathways—such as immune, respiratory, phagocytotic, and metabolic—were found to be deregulated across analyses. Furthermore, a number of significant DEGs were identified which might aid in the search for candidate biomarkers for AD development or, at the very least, contribute to the explanation of its onset.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The genetic variation between individuals, caused for a large part by common single nucleotide polymorphisms (SNPs) has been found to attribute to complex diseases or traits in general. SNPs are one nucleotide variations found throughout the DNA sequence, and account for the majority of genetic difference between humans. Genome-wide association studies (GWAS) aim to find genotype-phenotype associations through means of a statistical framework. Most commonly, SNPs are tested individually for association, however, such approach fails to detect SNPs with very small effect sizes. SNP-set GWAS aims to combine these small effects by simultaneously testing for association with multiple SNPs close together in the genome. Grouping SNPs leads to fewer tests thus alleviating the multiple testing burden and relaxing the significance threshold. In addition, highly complex phenotypes such as the brain or face are hard to describe univariately, e.g., by a volume measure. Canonical correlation analysis (CCA) offers a bi-multivariate statistical framework where multiple SNPs can be tested for association with such a multivariate phenotype. This work explored the potential and features of bi-multivariate, genome-wide, brain-wide association testing on cortical surface morphology in 19,643 individuals of European ancestry. SNP-set GWAS were conducted based on gene-based, haplotype-based, and window-based grouping of SNPs, and found 120 genes, 124 loci, and 124 loci respectively influencing cortical surface morphology. By comparing the results to a recent GWAS by Naqvi et al. (2021) it was demonstrated that SNP-set GWAS is able to detect both known and novel associations. Window-based GWAS were conducted using a range of window sizes (5 - 200 kb) and showed that larger SNP-sets result in fewer associations. In addition, it was shown that there exists an optimal range of window sizes and thus group sizes for finding the loci that per-SNP GWAS fails to detect. This thesis proposes an adapted implementation of the genomic control factor, which measure the inflation of the mean test statistic due to population stratification or systematic bias, specifically suited for SNP-set association testing with the CCA framework. In addition, it was demonstrated that the genomic control factor can be used on pooled test statistics to investigate whether certain group sizes yield disproportionately inflated test statistics and as such result in bias. It was then demonstrated that a correction per pool can be used to adequately correct for ununiform inflation. To further illustrate the potential of SNP-set GWAS, the coordinates of interogressed Neanderthal haplotype blocks (i.e., genomic regions in present-day humans that originate from interbreeding events with Neanderthals) were used as the grouping structure, and each block was subsequently tested for association with brain shape. In total 6 blocks exceeded the most stringent significance threshold, suggesting that SNPs within these blocks could potentially affect the brains of present-day humans.