Based on the workflow published here: https://pmc.ncbi.nlm.nih.gov/articles/PMC5747346/

Curious if I have a treatment varying by dose if that be modelled in this framework? I'm thinking at first to model a simple linear response (I understand this may not be entirely accurate).

edgeR works with any linear model, so handles any combination of factors and covariates. If you include dose as a covariate in the linear model, then edgeR will be (in effect) modeling the proportion of methylated reads as a logit response curve of the dose.