Question: Methylation analysis with MEDIPS
0
gravatar for ilyco
6.3 years ago by
ilyco50
United Kingdom
ilyco50 wrote:

Hi,

I have performed a differential methylation analysis using MEDIPS (R Bioconductor package) using data from 2 types of cells to identify regions which are methylated in one but not in the other (or hypermethylated in one and hypomethylated in the other). For each cell type, I had 3 replicates and then I separated the regions obtained according to the log fold change (if it was positive I assigned that to one cell type and if it was negative to the other).

However, after performing the analysis,  I obtained some regions which appear to be methylated in both.  More exactly, I have intersected the regions obtained with gene coordinates and one gene appears in both sets.

Does anybody have any idea why I obtained the same gene after performing differential methylation analysis for genomic regions which are differentially methylated in each of the cell types?

Thank you!

R medips dmr methylation medip-seq • 2.4k views
ADD COMMENTlink modified 6.2 years ago by Devon Ryan97k • written 6.3 years ago by ilyco50
1
gravatar for Devon Ryan
6.2 years ago by
Devon Ryan97k
Freiburg, Germany
Devon Ryan97k wrote:

This isn't all that surprising. If you were to plot the methylation in each sample in these regions, what you'd likely see is changes from near 100% methylation to near 0%. What ends up happening, then, is that you have shifts to the left/right of this dip between the groups. Whether these changes are functional is pretty questionable. In most case, I would guess that they're not, particularly if the shift is fairly small. If the change is quite large then perhaps it's functional (e.g., you have a change in accessible transcription factor binding sites). That's at least one possibility.

ADD COMMENTlink written 6.2 years ago by Devon Ryan97k
0
gravatar for jwade44
6.2 years ago by
jwade4410
United States
jwade4410 wrote:

Hi,

Although what you observed may seem strange, it is a common misconception than differential methylation (DM) analysis will find only find differences where one group is hypomethylated and the other is hypermethylated. The confusion comes from absolute versus relative comparisons; unless there is a priori data measuring the absolute level of methylation, terms like 'hypo' or 'hyper' are always relative comparisons, so you should ask yourself, "relative to what?".

Here's a concrete example. 

<caption>Hypothetical Scenarios</caption>
Gene Reads in Cell Type A Reads in Cell Type B
alpha    80    20
beta    20    80
delta    20    5
gamma    80   320

 

Suppose more reads implies more methylation. A is hypermethylated relative to B for gene alpha and delta, but A is hypomethylated relative to B for gene beta and gamma. Note that in each case the difference is 4-fold in one direction or the other. So based on a DM analysis, all of these genes would be DM between the cell types (suppose they had a low q-values so we could call them DM).

Now what if you believed that values greater than 50 implied something was methylated, so that you could make absolute statements like "alpha in cell type A is methylated". Then you might expect MEDIPS to only report alpha and beta as DM. What do you make of delta and gamma then, you ask (at least I think you do because that is how I interpreted your original question)?

Well, delta is NOT methylated in either cell type, but cell type A does have a greater propensity for methylation than cell type B, so much so that MEDIPS would call it DM.

Similarly, gamma is methylated in both cell types, but much more so in cell type B than cell type A.

I hope this helps a little. 

ADD COMMENTlink modified 6.2 years ago • written 6.2 years ago by jwade4410
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 2.3.0
Traffic: 2099 users visited in the last hour