What Is The Gene'S Correlation With Phenotype In Detail Using The Gsea

Question

[Zhilong Jia]

how to calculate the gene's correlation with phenotype in GSEA ?

For example: two kinds of samples: Ctrl and Exper, GSEA will give a ranking list of genes that related the phenotype Exper. At the same time, can GSEA also give a ranking list of genes that related the phenotype Ctrl? But Does any difference bewteen them?

what is the relationship between phenotype C and phenotype A or B in the original paper about GSEA (Gene set enrichment analysis A knowledge based approch for interpreting geneome wide expression profiles)?

Thank you.

Answer 1

[Sean Davis]

GSEA uses a measurement of gene expression differences, not absolute gene expression. Therefore, when you talk about a list of genes "related to the phenotype Exper", you are actually talking about the genes that are different between "Ctrl" and "Exper". So, if you have two experimental conditions, treatment and control, only one gene list will be used for GSEA.

[Zhilong Jia]

In the FAQ of GSEA (http://www.broadinstitute.org/cancer/software/gsea/wiki/index.php/FAQ)

Why comparing phenotypes A vs B gives different results from B vs A?

This is because these two comparisons produce different ranked lists of genes. You might expect similar results only if the ranked lists would be perfectly symmetrical, and this usually does not happen.

I'm confused.

[Sean Davis]

The gene lists are the same, but the directionality of the results are opposite. Those genes that are upregulated in B vs A are downregulated in A vs B. GSEA, if applied using a ranking that respects directionality, will give potentially different results, I agree--sorry for the confusion.