Multi-Omics to Study Causes and Consequences of Type 2 Diabetes

This thesis discusses the benefits and limitations of using molecular studies to understand Type 2 diabetes (T2D). As a complex disease, T2D is caused by a combination of multiple genetic and environmental factors. Its consequences on the glucose homeostasis often trigger a variety of co-morbidities, including cardiovascular diseases, stroke, and neuropathies, generating a major health burden. In the first chapter, we looked for genes which were differentially expressed with T2D across 49 tissues using data created by the GTEx consortium. In twelve of those tissues, we found between 1 and 5174 genes to be significantly differentially expressed (FDR=0.05). The relevance of the tissue was not a major factor in the number of genes discovered. In fact, in some cases the tissue suggested that the differential expression was due to comorbidities with T2D rather than the disease itself. For example, the most differentially expressed genes were found in the tibial nerve, likely due to comorbidities with neuropathy. We also found a larger number of differentially expressed genes with biological factors such as sex or age (a maximum of 14319 and 12897 respectively, FDR=0.05), though this could be due to the limited number of individuals with T2D in the dataset. In chapter two we looked to discover genes responsible for T2D development using a class of methods called transcriptome wide association studies (TWAS), which combine information from genetic studies into disease with information from reference molecular datasets. One question we wanted to answer was about the choice of the molecular reference: was it better to collect data from a tissue that was directly relevant for the tissue, or could a well powered study in a non-relevant accessible tissue also be informative? For this we used expression data from 49 tissues previously mentioned, a mix of relevant and non relevant tissues, expression data from the directly relevant pancreatic islets produced by the Inspire consortium, and data from whole blood produced by the DIRECT consortium, non relevant but with a sample size more than three times larger than the others. We found that the relevance of the underlying molecular data had no effect on the number of T2D causal genes we discovered, instead observing a strong correlation with sample size. More genes found using DIRECT data were confirmed using multiple instrument Mendelian randomisation than with any other molecular dataset. (31 genes confirmed through MR compared to 1 to 9 for the other molecular datasets). However, important GWAS loci were missed when using non relevant tissues; for example, the well known TCF7L2 locus was only found using pancreatic islets. Finally, our last chapter focused on the observable effects of metformin, a commonly used medication for T2D. We took three approaches to discover genes which interacted with the drug. Firstly, using GWAS summary statistics from a study into metformin efficacy, we applied two methods to find genes associated with drug efficacy: MAGMA, which is based on purely genetic information, and the TWAS methods from the previous chapter. Using MAGMA we discovered 880 associated genes (FDR=0.05) compared to only 38 using TWAS (though TWAS associations contained important information on tissue of action and direction of effect). Finally, we used longitudinal data to look for genes whose expression changed after taking metformin and discovered 348 genes at a less stringent threshold of FDR=0.2. Comparing our differentially expressed genes to TWAS associated genes for efficacy and for causing T2D, we found that metformin tended to produce changes in expression which would promote drug efficacy, a treatment virtuous circle, and counteracted expression patterns predicted to cause the disease, explaining the treatment benefits. However, neither of these properties was statistically significant, due to the small numbers of genes involved. In summary, this thesis has explored the molecular causes and consequences of T2D, as well as the genes and pathways involved in treating it.