For GO/PATHWAY analysis: should i use over-representation analyses or enrichment analysis

Question

[Muha0216]

i have a list of DEGs that i want to input into a good database for GO/PATHWAY analysis. This website is cpdb.molgen.mpg.de). This website is good as it analyses the DEGS and compares it to several databases all at once (KEGG,biocartome etc).

The problem is i do not know if i should do an over-representation or enrichment analyses - if you visit the link i posted above, on the left panel it shows you can either do enrichment or over-representation analysis.

Can anyone suggest me to this?

Answer 1

[Jean-Karim Heriche]

I am not sure what the distinction between the two is but there seems to be some explanation on the website documentation, check the section on Molecular concept-based analysis of gene/metabolite lists. From there, it seems the former is the standard enrichment analysis based on the hypergeometric distribution and the later is based on statistical tests of measured expression values.

Answer 2

[h.mon]

See the Introduction of A Comparison of Gene Set Analysis Methods in Terms of Sensitivity, Prioritization and Specificity. What the ConsensusPathDB site calls "enrichment analysis" is called in the paper "Functional Class Scoring (FCS)".

The paper will help you to decide, but Functional Class Scoring methods are, in general, considered more sensitive and less arbitrary than Over Representation Analyses methods.

Answer 3

[Muha0216]

The enrichment analysis option found on the website requires me to input the gene symbol together with two values. Some wilcoxon based analysis. Dont even know how and why i need these wilcoxon values.

I will appreciate any new advise.

[h.mon]

You don't need "these wilcoxon values". You need to read more:

Wilcoxon enrichment analysis

The Wilcoxon enrichment analysis method carries out a paired Wilcoxon signed-rank test for each NEST / GO category / pathway based on the user-specified measurement values of its members. The measurement values for every gene / protein, uploaded by the user, typically reflect genome-wide gene expression or proteome-wide protein abundance in two different phenotypes. For every uploaded gene or protein, exactly two values must be supplied in the input form (or uploaded file). The Wilcoxon test assignes a P-value to each functional set based on how probable it is that the combined measurement differences of genes in the functional set between the phenotypes have appeared by chance. Q-values are calculated using the same method as in the over-representation analysis approach.

Answer 4

[Muha0216]

I think i sort of get it about difference between Overrepresentation and enrichment analysis With the latter using some weight for each gene.

But which analysis is acceptable for publishing?