Theses and Dissertations Collection

DNA methylation, an epigenetic mechanism, plays an important role in transcription regulation and in complex diseases. Whereas methylation in gene promoters is known to be generally associated with silencing, the relationship between transcription and methylation in other parts of the gene is much less clear. Additionally, substantially different transcription and methylation profiles have been observed among breast cancer subtypes, but it is unclear whether and how these differences are influenced by different relationships between the two processes.

Here, we studied the relationships between transcription and methylation in estrogen receptor-positive (ER+) and negative (ER-) patients, using data from The Cancer Genome Atlas (TCGA) consortium and Genomics Data Common portal (GDC). We formulated trios, each consisting of the Copy Number Alteration (CNA) of a gene, expression (E) of this gene, and methylation (M) of a site located near or in the same gene. Since CNA is prevalent in cancer, it is a highly effective instrumental variable for this causal inference. In each subtype, we further derived principal components from genomewide expression and methylation data and identified those that are significantly associated with each trio as potential confounding variables.

We applied MRGN, a novel causal network inference method that accounts for many confounding variables under the principle of Mendelian randomization, to each of the 310,412 trios in each subtype. We further examined the features of methylation probes in mediation models as compared to the baseline models. Our analysis provides a first comprehensive picture of causal relationships between transcription and methylation in the two subtypes.