Abstract
DNA methylation (DNAm) is an important epigenetic modification involved in nu- merous cellular processes and implicated in disease. Genetic variants that influence DNAm levels at specific CpG sites, or methylation quantitative trait loci (meQTLs), play a significant role in shaping the epigenetic landscape. The identification of meQTLs holds the promise of advancing our knowledge of gene regulation pathways and disease susceptibility. This project aimed to study meQTLs through genome-wide association analyses and to explore the interplay among DNAm, gene expression, and human phenotypes. Additionally, I assessed the utility of meQTLs to impute missing DNAm data.In the initial research chapters, genome-wide association analyses were carried out between genotypes and DNAm levels at more than 700,000 CpGs, to detect meQTLs in blood samples from three UK-based population cohorts. The identified meQTLs had an impact on DNAm levels at 34.1% of genome-wide CpGs, with 98% of meQTLs exerting local effects. After fine-mapping analysis, I found that 1,325 unique meQTLs co-localised with signals from genetic association studies (GWASs) for 34 phenotypes. Furthermore, I developed an online database and visualisation tool to facilitate meQTL exploration and its genomic context, and improve accessibility of our results to the scientific community.
The subsequent research chapter investigated cross-tissue meQTL associations in TwinsUK samples from three different tissues – blood, adipose and skin tissue. I employed a joint model to detect local meQTL effects at nearly 400,000 CpGs, and quantified the number of associations with single-tissue and cross-tissue effects. Nearly half of the tested CpGs exhibited meQTL associations in at least one tissue, with 90% featuring cross-tissue meQTLs.
The final research chapter outlines the development of a DNAm prediction method using meQTL results. I focused on a model with seven pilot CpGs, and systemati- cally compared the performance of different linear-based models, and assessed the significance of co-methylated CpGs and meQTLs in enhancing predictive accuracy. I determined that co-methylation alone lacked the consistency required for viable prediction. Moreover, meQTLs did not substantially enhance prediction accuracy, leading to the conclusion that the developed predictor does not represent a marked improvement over existing models.
These findings collectively show the substantial genetic influence on blood DNA methylation levels and underscore the prevalence of cross-tissue meQTL effects. This research contributes to the broader understanding of the mechanisms underlying DNAm variability.
| Date of Award | 1 Jun 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Jordana Bell (Supervisor) & Mario Falchi (Supervisor) |