How to tell if enrichment is good or bad from GSEA plot

Question

[gundalav]

I have the following four GSEA plots:

My questions are:

1. Which of four of them are considered good? Is it A, B, C? Or just A, B? Or all of them?

1. Some GSEA plot comes with positive enrichment score (e.g. plot C, D) Some of them negative (e.g. plot A, B) How can we use those difference in making decision which of them are good?

Answer 1

[Kevin Blighe]

Here, enrichments A and B look excellent, while C and D are not quite clear.

A and B look excellent because your most important variables (indicated by colour, with black being most important, and light grey being least important) are positioned at either the left- or right-most part of the plot; moreover, the density of important variables is also high at both of these locations. The enrichments in A and B are going to be highly statistically significant.

'Importance', here, is gauged by whatever is your input data, which may be expression level, fold change, negative log10 p-values, or something else. Further, the directionality of this importance is indicated by the coloured horizontal bar - if fold changes are the input, then red equates to positive fold-change, while blue equates to negative fold change. Thus, we also have information on direction of enrichment. In A, we can infer that the PRONEUAL_PHILLIPS signature is activated, while, in B, PROLIFERATIVE_PHILLIPS is deactivated.

C and D don't look too great, in part due to these signatures being small in size and there being no trend or consistency for your variables. C undoubtedly scores better than D, though.

Kevin