I recently got some EPIC DNA methylation data and I was wondering what are some good practices to follow? I am interested in knowing about normalization and differential analysis. Thank you.
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ Hi deepseas, welcome and thank you for posting a question to Bioinformatics StackExchange. I've added a link to the Infinium EPIC web page for people who might be curious about what EPIC DNA methylation data is. Your question is quite broad, and would be substantially improved by adding more of a story behind why you want to do this analysis (e.g. is this a single individual study? do you want to compare with RNASeq data?). Bioinformatics is a very large subject area, and there are many different ways to do [almost] the same thing, so any help with clarifying questions is greatly appreciated. $\endgroup$– gringer ♦Jun 6, 2017 at 21:49
-
1$\begingroup$ Thanks for the response! This data has 16 samples from different individuals with 5 groups (5+6+1+2+2). It will be difficult to get power for groups with 1 and 2. but the main aim is to get an idea about certain known genes of interest. $\endgroup$– deepseasJun 6, 2017 at 23:27
-
1$\begingroup$ It will be difficult to make contrasts like edgeR,limma or deseq2 makes with low number of replicates per condition. 2 might still work but 1 in a single condition will not hold good. Ideally when you work with EPIC arrays or 450k both will work with any standard 450k analysis tool. Try tools like DimMER/Rnbeads. However things hold better with higher replicates per groups. All the best! $\endgroup$– ivivek_ngsJun 8, 2017 at 13:10
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
EPIC data can be processed in the same manner as the previous iteration of methylation array data from Illumina (450k). This means that starting with .idat files, normalization should be performed (for example, via the minfi package). A recent paper from the creators of minfi is particularly helpful because it makes clear that normalized EPIC data from their package can be immediately compared against, for example, level 3 TCGA data.
After that, I suggest using the manifest to attach genomic coordinates to your probes and segregating them into functional regions. By testing differential methylation in only regulatory regions, for example, you can increase the statistical power by reducing the overall number of tests to the ones you expect to yield major differences.
There are existing packages out there for differential methylation analysis, but without knowing your replicate structure or aims, it is difficult to point you in the right direction.
