An approach for gene ontology and up- / down-regulated genes (RNAseq)

Question

[david.roux]

Hello, I am working on a "simple" lettuce RNAseq experiment comparing one situation to a control.

I did a topGO enrichment of all my differentially expressed genes.

Then, for some pathways my lab is interested in, I am trying to get back the list of the associated genes and then see how they are regulated (UP or DOWN) in the initial DGE table (fold change).

Clearly, this approach sounds true. (Especially since lettuce is not a genetic model.)

Or am I doing it the wrong way?

Note , I have allready read this post: Question: GO analysis- analyze upregulated and downregulated genes separately? and I agree to :

the fact that a gene is upregulated or downregulated does not mean that the process, function or component that the gene is annotated to is also respectively directionally regulated

Just because some genes are up or down-regulators.

I have also read this post and the associated article of Guini Hong. It seems that one could not be convinced...

I understand that GO enrichment is a hot topic, so it is hard to know If a way of analyis is really the good one...

Thanks in advance for your expertise.

[Jean-Karim Heriche]

I am not sure I understand what your problem is. You're interested in a pathway, you know which genes belong to this pathway and you have data that tell you whether these genes are up and down regulated. There's no problem there. So what is your question ? Is it about how to draw conclusions from this ? This is probably going to be pathway-specific because one can imagine that pathways with both positive and negative regulators would see those regulated in different ways, e.g. positive regulators could be down-regulated when negative regulators are up-regulated.

[Devon Ryan]

Is there a question here? Yes, having a bunch of up-regulated genes in a pathway doesn't mean that the pathway is more active (it could be less active due to increased expression of inhibitory elements). If directionality is important then you need to think of that yourself.

[david.roux]

Sorry for the confusion, simply, my question was: What is the better way ? - Make the GO enrichment on a mixed list of up- and down-regulated genes? - Make the Go enrichment on up-regulated genes and then on down-regulated genes separately? I did the first way with topGO / R. But I know that the both approaches are used. Anyway, thanks all for your answers. I will have a look on the technics you suggested.

Answer 1

[mark.ziemann]

Why not try a technique that can use information from every detected gene, like GSEA, CAMERA, etc? These will identify gene sets (aka pathways) that are up and down regulated. You can do set based GO analysis with DAVID or GoSeq but these are less sensitive and are strongly dependent on the selected significance threshold. If using set based GO analysis, the up and down regulated genes should be in separate lists as this will give you a better idea of the effects of your treatment or contrast. For instance if you stress a plant with heat, it will increase stress and heat shock pathways and reduce cell cycle pathways. Direction is very important for interpretation of the biology. A background list of genes detected in your experiment is mandatory for statistical validity. More information on the statistical aspects of gene set analysis here and here.

Answer 2

[Guangchuang Yu]

If you want to find up/down-regulated pathways, you need to use GSEA (gene set enrichment analysis) instead of ORA (over-representation analysis). You can use clusterProfiler which supports both of these analyses.