Hello, Thanks for the awesome tool. We are using edgeR for the RRBS for differential methylation analysis on multi regional tumors (multiple regions collected from each tumour). One of our purpose is to find out if there is any different pattern of methylation between the regions in a tumour in a particular tumour. So in this case I do not have any replicate (I have one library for regionA and one library for regionB in a tumour) I see you have some suggestions to run EdgeR when there is no replicate but I am wondering how it should be performed for the methylation data, not gene expression? For example what would be ideal bcc value?

I managed to create the following list. I thought I can assign group 1 to region A and group 2 to region B and do exactTest(y, dispersion=bcv^2) But I do not think it is correct because I h ave methylated and unmethylated read counts for each region.

I really appreciate if you could help me how I can proceed to find out if there is any difference in methylation between two or more regions in a tumour when there is no replicate (one library per region)? Thanks

Enter the body of text here

Code should be placed in three backticks as shown below

> y
An object of class "DGEList"
$counts
        SRC159-T1-Me SRC159-T1-Un SRC159-T2-Me SRC159-T2-Un
1-10525           55            9           23            9
1-10526           72           21           61            7
1-10542           33           31           20           12
1-10543           44           49           51           17
1-10563           30           34           17           15
9470247 more rows ...

$samples
             group lib.size norm.factors
SRC159-T1-Me     1 57103934            1
SRC159-T1-Un     1 44889934            1
SRC159-T2-Me     2 33037087            1
SRC159-T2-Un     2 22688158            1

$genes
        Chr Locus
1-10525   1 10525
1-10526   1 10526
1-10542   1 10542
1-10543   1 10543
1-10563   1 10563
9470247 more rows ...

sessionInfo( )

The correct way to analyse methylation data is explained in this article https://f1000research.com/articles/6-2055/v2 and in the edgeR User's Guide. You must use glmFit and glmLRT rather than exactTest.

Our advice for analysing data without replicates in the same for RRBS BS-seq as it is for RNA-seq and is explained in the edgeR User's Guide. You need to input a preset value for the dispersion but I cannot tell you what that value should be, it depends on your previous experience with the same sort of data. If you have no idea, then try the values suggested in the User's Guide. Or try treating your two samples as a single group with n=2 and see what dispersion value you get. Then use that for the real analysis.