How to interepret DNA methylation by DSS package?

Question

[ahmad mousavi]

Hi everyone,

I have done RRBS analysis by DSS package, it gave differentially methylated regions (DMR) with some strange value, based on manual and author's information. I have got significant result. I have two questions:

1- I have got ~50.000 DMR (siginicant) in each comparison, is it normal? 2- I can't understand DMR output, meanMethy1, meanMethy2, diff.Methy are these numbers depends on nCG and Length of regions?

chr   start    end   length    nCG     meanMethy1   meanMethy2  diff.Methy        
chr4    119521077   119521411   335 38  0.343009905 0.920667978 -0.577658072

Can anyone explain meanMethy1, meanMethy2, diff.Methy and why diff.Methy value is huge?

Answer 1

[Devon Ryan]

meanMethyl1: The average methylation in the first group in the given region

meanMethyl2: The average methylation in the second group in the given region.

diff.Methyl: The difference between the values above (group 2 - group 1).

As one can see, group 1 has moderate methylation and group 2 has high methylation, the difference being ~0.6.

[ahmad mousavi]

Thanks dear Devon, I couldn't find good manual for that. My problem is how to infer high/low methylated values, it might be false idea to compare this result with RNA-SEQ but how average methylation is calculated ? is this mean from for example meanMethyl1: 34 % of 38 CGs were methylated? And when we could consider diff.Methyl as high methylated? Thanks.

[Devon Ryan]

A change of almost 60% is quite large, generally anything over 10-20% could have a biological effect. BTW, DSS produces an FDR column, you should be looking at that as well. Yes, the diff.Methyl is the average methylation change over those 38 CpGs.

Do read through the DSS vignette, it's a decently documented package.

[reksio342]

Hi! Your answers have been very useful, thank you! I am still wondering though, so if diff.methy is the average methylation change over the CpGs does that mean a positive value is hyper methylated and negative hypomethylated?

Because it looks like diff.methyl is doing (group1 - group2), so I thought it was the other way round (a +ve value = hypo and -ve hyper)