GSEA genesets postiively enriched in positive phenotype are necessarily negatively enriched in the negative phenotype?

Question

[liu.huand]

Hi, all, I did GSEA enrichment for my RNA-seq profile. But I'm still confused on understanding the results. I compared the profile (whole RNA-seq results without setting any cut-offs) in my positive samples (positive phenotype) and negative samples (negative phenotypes). I wonder, in GSEA results, genesets postiively enriched in positive phenotype are necessarily negatively enriched in my negative phenotype? Actually, what I got is I found my desired genesets highly positively enriched in my positive phenotype but not highly (negatively) enriched in my negative phenotype..Not sure whether that are acceptable..

many thanks, Huan

[WouterDeCoster]

What is a negative phenotype?

[fanli.gcb]

I believe the OP is referring to the default 'positive' and 'negative' phenotypes that are inferred when GSEAPreRanked is provided signed ranks (e.g. fold change values from RNA-seq).

Answer 1

[fanli.gcb]

If you followed GSEA guidelines for RNA-seq data, your 'negative' and 'positive' phenotypes reflect negative and positive fold changes in your comparison. So even if a particular gene set is enriched in one direction (e.g. among genes that are upregulated in your data), there is no requirement for genes from this set would be depleted among downregulated genes. That being said, one could argue that if said pathway were operating as a consistent unit, you might expect things to move in the same direction but that's not a given.

[liu.huand]

Thanks for the replies. I will take this. One more question about interpretation of GSEA results. For examples, I found a gene set called "positive regulator of development" was enriched in my positive phenotype, is it legitimate to say "Compared with negative phenotype, genes in positive phenotype is more positively enriched in 'positive regulators of development' (p<0.01, FDR<0.10), indicating treatment (negative) impaired development" ? Sorry about this wordy question, but I'm just afraid of being overstating my results...

Many thanks,

Huan

[fanli.gcb]

I don't think you can draw a direct comparison between your phenotypes. I would also avoid using positive and negative phenotype as these are just the default names provided by GSEA. Rather, use your actual RNA-seq comparison (for example, disease vs. healthy). For example,

'Positive regulators of development' (FDR<0.10) genes were enriched among those found to be upregulated in the disease group.

[liu.huand]

Yes, I will use my group name for sure.

Now everything makes sense to me.

Many thanks!

Answer 2

[alserg]

AFAIK, there is no positive enrichement in positive phenotype: it's either positive enrichment, or enrichment in positive phenotype. Each pathway can be either positevely enriched (with positive ES score) or negatively enriched.