dmrff: identifying differentially methylated regions efficiently with power and control Matthew Suderman, James R Staley, Robert French, Ryan Arathimos, Andrew Simpkin, Kate Tilling bioRxiv 508556; doi: https://doi.org/10.1101/508556
Background. An epigenome-wide association study (EWAS) tests associations between epigenetic marks such as DNA methylation and a given phenotype or exposure. A popular strategy for increasing power is to test associations of epigenetic measurements taken across genomic regions rather at individual loci. This strategy has seen some success because patterns of epigenetic marks at neighboring loci tend to be under similar regulatory control. Unfortunately, the most commonly used implementations either fail to control false positive rates (e.g. comb-p) or suffer from low power (e.g. bumphunter).
dmrff is a DMR-finding tool that:
- controls false positive rates
- is more powerful than EWAS
- is fast
- can be applied to the summary statistics of any EWAS
- can be used in the context of meta-analysis
A simple example showing how to apply dmrff to a publicly available dataset can be found in the docs directory.
A more extended example shows how to use dmrff to meta-analyze multiple datasets.
A simulated example illustrates how dependencies between CpG sites affect DMR statistics.
A wiki is being developed and includes answers to some frequently asked questions.