Interpretation of several pairwise overlaps

Question

[olechnwin]

The following were taken from an example in GeneOverlap library where it did 12 pairwise gene overlap tests using Fisher test.

re-producible code:

library(GeneOverlap)
data(GeneOverlap)
gom.obj <- newGOM(hESC.ChIPSeq.list, hESC.RNASeq.list, 
                  gs.RNASeq)
drawHeatmap(gom.obj)

My question is whether you can interpret odds ratio as strength of association between different the pairwise overlap. For example: Odds ratio for overlap between H3K36me3 and Exp High is the highest compared to all other pairwise overlap (darkest green compared to others). Does that mean association between H3K36me3 and Exp High is higher than all others? Does that mean overlap between H3K36me3 and Exp High is more likely than other overlaps (i.e. overlap between H3K4me3 and Exp High, H3K9me3 and Exp medium, etc)?

Thanks in advance for your help!

Answer 1

[predeus]

How was this calculated? E.g. did you make a window of constant size, or did you simply take individual intervals and looked for overlap? And what was the condition for two intervals to be considered overlapping?

[olechnwin]

These are just gene names overlap.

For example:

gom.obj["H3K9me3", "Exp Medium"]
GeneOverlap object:
listA size=297
listB size=6011
Intersection size=142
Overlapping p-value=7.1e-13

the background gene size is the same for all the pairwise overlap tests.

[predeus]

I would look carefully into how you associate a particular feature with a gene name

e.g. some genes are very long

but, to answer your question, I think you are correct to assume that odds ratio would show the strength of association.