Question

How do I discern if a magnitude difference in expression is up-regulation in one group or down-regulation in another?

0

Entering edit mode

philip.freda • 0

@philipfreda-13499

Last seen 5.6 years ago

When exporting TopTags in EdgeR, you of course get LogFC which have a sign associated with them and that refers to which group has higher expression but it doesn't really tell you anything about in which group the gene is significantly up-regulated in.

An example would be a LogFC of 3.5. This would mean it's higher in one of my groups (adults in my dataset) and a reasonable deduction would be that the expression would be up-regulated in adults. However, it can also mean that the gene is downregulated in my other group, larvae. And when I graph out the expression trajectories of all my genes, most of the time, the former explanation is correct, but more than once it has been the latter: downregulation in the other life stage. Considering I have hundreds of genes that I don't have the time to graph, is there a way in EdgeR to actually determine which genes are up-regulated and not just which ones have higher LogFC's in a comparison?

Thanks

edgeR R differential gene expression up-regulation gene list • 1.1k views

ADD COMMENT • link updated 5.6 years ago by Gordon Smyth 50k • written 5.6 years ago by philip.freda • 0

0

Entering edit mode

Do you have more groups than just adult and larva? With only 2 groups, there is no difference between "up-regulated in group A relative to group B" and "down-regulated in group B relative to group A".

ADD REPLY • link 5.6 years ago Ryan C. Thompson ★ 7.9k

0

Entering edit mode

Nope, just two groups. Perhaps what I did not express very well is no expressional activation in one group and down regulation in the other looks like upregulation in the first group when it isn't as all comparisons are adult - larvae. Therefore a positive LogFC just means adults have higher expression not that the expression up-regulated. Is that more clear?

ADD REPLY • link 5.6 years ago philip.freda • 0

0

Entering edit mode

I think there's something missing from your description. You say that all comparisons are adult - larvae, but if there's only two groups, then there can only be one comparison. And for this comparison, if you have a positive log-fold change for adults - larvae, then "adults having higher expression" is the same as "the expression is upregulated in adults compared to larvae".

ADD REPLY • link 5.6 years ago Aaron Lun ★ 28k

0

Entering edit mode

But if larvae down regulate the gene, it will appear positive in the comparison. I've seen it multiple times

ADD REPLY • link 5.6 years ago philip.freda • 0

0

Entering edit mode

even if the expression in adults remains unchanged from constitutive levels, which I have as part of my data set as a control

ADD REPLY • link 5.6 years ago philip.freda • 0

0

Entering edit mode

Please edit your question to give a concrete example with counts/CPMs for each sample; this is going nowhere.

ADD REPLY • link 5.6 years ago Aaron Lun ★ 28k

score 1 · Answer 1 · 2018-09-07

Dear Philip,

On the face of it, your question doesn't really make sense, as Ryan and Aaron have been pointing out. If you have only two groups (adults vs larvae) then being "up-regulated in adults" is by definition the same as being "down-regulated in larvae", so there can be no conflict and your question doesn't have any meaning.

Reading between the lines, I am guessing that you actually have a third control group that you view as representing "constitutive" expression. When you get a positive logFC for adult vs larvae, you want to know how adult and larvae compare to the control. As I think you are saying, it could be that larve = control and adult is higher, or it could be that adult = control and larvae is lower.

Obviously to distinguish these situations you need to make edgeR contrasts for adult vs control and larvae vs control. edgeR can't possibly return information about constitutive expression if you don't make a comparison with the control.