Enrichr input genes, up, down, or both at once

Question

[chrisgr]

Hi, I had a question about Enrichr and what genes you are supposed to enter as input.

After doing a DEA between 2 groups, it leaves you with a list of up regulated genes in say, 2 groups. So genes up / down regulated in group 1 and up / down regulated in group 2. This comes in a large list of gene symbols with both upreg in group 1 and group 2 mixed together with P-values and log-fold changes.

If I want to analyse this with enrichr, am I supposed to enter both groups as input, or categorize the input into 2 groups based on upreg in group 1 and upreg in group 2?

Thanks in advance,

Chris

[EagleEye]

One way of doing it would be to perform enrichment analysis for your up and down genes separately. Then check if same/similar pathways picking up as top pathways by both up and down genes. In case if it is too similar you can simply merge up and down and perform enrichment analysis. If it gives different pathways, it is better to keep it separate. I know this is not a perfect solution but It is complicated when you consider the complexity of the biology.

Answer 1

[ATpoint]

That is on you. Usually it is easier to interpret if you separate the lists. But biology is complex and it can that you have some genes being upregulated that activate a pathway, and some downregulated that inhibit a pathway, and then putting them together makes sense as both would contribute to the same message, which is the pathway being more active. No gold standard for this, it's on you to decide. Enrichr just takes genes and checks whether it overlaps more with pathways than expected by its background model. I usually start with the separated lists first, and only if I find nothing then merge them and see whether this creates some hypothesis of what is going on in my transcriptomes.

[Papyrus]

Only to add to the good response by ATpoint , expect to get more results when merging all the genes. But beware this will often not be for a biological reason, rather because you have more power to detect pathways with more input genes (especially large pathways which are often uninformative).

[chrisgr]

Thank you so much for the quick and clear response!

So that means that it is up to me to 'manually' check if the say, 'downregulated gene' in the pathway is also downregulated in my DEA. (Same for the up reg)

It only seems a bit strange to me that adding whether the input genes are up or downregulated is not a 'standard' input option for Enrichr. Because it can easily be matched to the genes in a pathway right? So that Enrichr can directly say; these pathways, considering the up AND down regulated genes of your set is probably active in this group. Because like you say, down regulation can also be important to activate a pathway. : )

edit: Hm, perhaps this dismisses the user-friendliness too much? Just adding a list of genes and getting a "suggestion" is also computationally faster?

edit: After giving it some thought I think I understand that it is not needed to add this input option. Because the results of Enrichr only tell you which pathways are 'different' in activeness between the 2 groups and it is very easy to see in what group the pathway is active by comparing your DEA results

[ATpoint]

Because the results of Enrichr only tell you which pathways are 'different' in activeness between the 2 groups

Yes, I think this is what it really comes down to. It is a suggestion the tools makes, that you then have to follow up to make sense of. After all, validation experiments will tell whether it is true or not. This is aat least how we usually deal with these enrichment results, we take them for validation in case we find them interesting.

[chrisgr]

Okay great! Thanks a lot for the insight, much appreciated!