Dear all, I have a theoretical question concerning the normalization of cross-platform DNA methylation data. I would like to compare signals from both Illumina450K array and eRRBS/WGBS assays.
The Beta distribution of those signal is quite different, so I'm using the R function
preprocessCore::normalize.quantiles.use.target() to perform a targeted quantile-quantile normalization.
After normalization, my array signal closely resembles the one from NGS based techniques.
However, I am concerned about 2 issues:
- First, am I forcing the introduction of a signal which is not there? Is my procedure sound?
- Second, what is the most legit target choice for the qq-normalization distribution, given that I am dealing with tumor and normal samples? Looking at different samples or sample-groups, they display different signal density, thus changing the result I obtain after normalization.
Thank you for your kind attention, best