Question: About methylation data processing
0
gravatar for Asma rabe
3.7 years ago by
Asma rabe290
Japan
Asma rabe290 wrote:

Hi,

I am analyzing methylation data, I have two files with beta values one for disease samples and another for control samples.

I would like to convert both disease and control beta values in to M values and calculate Fold change of disease/control using M values

I checked many articles and found that in some studies only beta values were used and delta B was calculated to assess DMR. 

I tried both ways in the following example and found that different ways could give different assessemnt for a probe or position but i got different results

disease=.73

cont=.57

M_disease=1.434937

M_control=0.4066253

FC=M_disease/M_control=3.528893 which is differentially methylated in case threshold is 2 fold

delta B=.73-.57=.0.16 #NOT differentially methylated in case threshold=.2

Any one has an experience which has more robust output, using either delta B or  FC for instance or using p-value of statistical tests as implemented in many R packages.

Any help is appreciated

 

 

methylation • 1.4k views
ADD COMMENTlink modified 3.7 years ago by Peter Hickey460 • written 3.7 years ago by Asma rabe290
Answer: About methylation data processing
1
gravatar for Peter Hickey
3.7 years ago by
Peter Hickey460
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Peter Hickey460 wrote:

Hi Asma,

In any test of differential methylation, you want to consider the variation in each group, not just the mean difference (delta). I'm assuming here that you have multiple samples in each of the "disease" and "control" conditions. There are well-founded methods for performing differential methylation testing in, for example, the minfi Bioconductor package. You may find reading the included User's Guide helpful. 

As to the choice of M-values or Beta-values, there is some literature discussing the pros and cons of each approach. The paper "Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis" by Pan Du et al. provides a good overview of these tradeoffs. I think the argument can be reasonably summarised by "beta-values are more interpretable but M-values may offer better statistical properties".

ADD COMMENTlink written 3.7 years ago by Peter Hickey460

Thanx Peter. The data I have came from illumina K27 methylation arrays which minfi package does not support. Are there other packages that support illumina K27 arrays? 

  

ADD REPLYlink written 3.7 years ago by Asma rabe290

You might try the lumi package. The vignette discusses analysing Illumina 27k data

ADD REPLYlink written 3.7 years ago by Peter Hickey460

Hi Peter,

I checked a bit older user guide which mentioned that illumina 27k is not supported. 

Thank you very much for the link to latest vignette

 

ADD REPLYlink written 3.7 years ago by Asma rabe290

To clarify, "minfi" does not currently support Illumina 27k but "lumi" does (these are different packages).

ADD REPLYlink written 3.7 years ago by Peter Hickey460
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 175 users visited in the last hour