This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Do I have to use beta-values/m-values for RRBS/WGBS data ?

I am working on RRBS data. Someone suggested me to use M-values for the calculation of differential methylation. In my understanding beta-values/m-values are used to measure methylation level in array based method. In sequencing data we get single base pair resolution methylation level after methylation call. Hence, no need for calculating beta-values/m-values.

I would like to make sure that I am right. Could you please confirm that beta-value/m-value calculation is not needed for sequencing data?

rrbs wgbs methylation
• 5,747 views • link
updated by Devon Ryan Devon Ryan • written by Tori Tori

1 answer

Every tool will have its own input requirements. Some tools will require the raw counts, other will require methylation percentages in the range 0-100 (or 0-1, it'll depend on the tool). You shouldn't be writing a tool to do this yourself, use something published and give it whatever it wants as input.

• 0 views • link
written by Devon Ryan Devon Ryan

I have used Methylkit, RnBeads and RadMeth for analysing RRBS data, but after diving into greater detail of the result files of these methods DM regions with high p-value/qvalue does not make sense. Those statistically significant (high p-value/qvalue) DM regions do not seem significant in biological sense. Therefore, I would like to use Bismark output methylation call files in some other statistical test.

• 0 views • link
written by Tori Tori

I've had some nice experiences with metilene, give that a try. It may be that you simply have no biologically relevant DMRs.

• 0 views • link
written by Devon Ryan Devon Ryan

Hi Devon, I tried metilene and I get thousands DMR with different length. But how can I get the DMR relation to gene. I tried bedtool, but get nothing.Maybe there is something wrong. what is you method to work this out?

• 0 views • link
written by t2g4free • t2g4free

bedtools nearest should work fine.

• 0 views • link
written by Devon Ryan Devon Ryan

And I see some Question you answered before, you said not to use Fisher test. when there are two sample and no replicate, which way or software is better to get DMR? thx

• 0 views • link
written by t2g4free • t2g4free

What I was likely trying to convey is that such a design is inherently problematic. Your only option is something like a Fisher's test, though.

• 0 views • link
written by Devon Ryan Devon Ryan

Log in to answer this question.

More posts like this