University of Washington - Department of Statistics
DNA methylation in mammalian cells plays an important role in gene regulation and is essential for normal development. Patterns of DNA methylation are transmitted through somatic cell divisions generally with a high fidelity rate. Imperfection is due to failure-of-maintenance and denovo methylation events. It is of interest to accurately estimate rates of these events. Single- and double-stranded sequence data are available for this task. We survey existing methods for estimating maintenance and de novo rates, which typically assume independence across sites. We present a Bayesian hierarchical model to make use of the multi-site information in the data. This model accounts for dependence to certain extent and makes parameters more identifiable, thus providing more insights into the biological process. We have applied this model to the data collected in the promoter region of the FMR1 locus that is responsible for fragile X syndrome. In addition, we will discuss the limitations of the model and our plans to incorporate features that have not been incorporated in our current model.