DNA Methylation: Methods and Analyses

Epigenome-wide Association Studies (EWAS) have been a popular method to investigate the genome over the past decade. From these experiments, more than 75,000 samples have been assayed using the high-throughput, cost-effective HumanMethylation450 microarray (450k) developed by Illumina. With the recent release of the HumanMethylationEPIC microarray, the size of data is expected to increase considerably so advances are needed in the methodologies used to analyse such data. The first part of this thesis focuses on the development of tools that can be used for the analysis of DNA methylation microarray data. Firstly I develop a wide range of tools that can be used to quality control data. These tools focus specifically on data-driven aspects of quality control that are often overlooked and can cause problems during downstream analysis. Comparison of these tools to other popular methods demonstrate that the tools I created are effective in decreasing test statistic inflation while conserving the largest number of samples (Chapter 2). Secondly to accommodate the increase in the size of data, I developed the bigmelon R package which reduces the amount of memory required to perform the analysis typically required of EWAS (Chapter 3). I then demonstrate how both the tools described in Chapters 2 and 3 can be used in EWAS settings. I perform an EWAS between DNA methylation and various blood-lipid traits and statin-use on a dataset comprising of 1,193 samples from the Understanding Society: UK Household Longitudinal study and replicate the findings of many previous EWAS (Chapter 4). Lastly, I demonstrate how the data from tens of thousands of microarrays can be utilised in preliminary analyses that focus on the wide-spread characterisation of the probes on the 450k microarray and how tissue-specific DNA methylation patterns may correlate with tissue-specific gene expression (Chapter 5).