Valid Visualizations Of Significant Analysis Of Microarray Data Followed By Pathway Enrichment Analysis

Question

[user1409015]

I am using SAM to find differentially expressed genes between three groups of variations upon a disease. I have then used pathway overrepresentation analysis (using a hypergeometric test with correction for multiple testing) to see what biological pathways from a curated database are overrepresented for each of the three groups using the top-ranking genes from SAM (q < 0.05). I want to visualize the network to show the following

• Node Size: number of genes in pathway
• Edge Weight: degee of overlap between pathways
• Node Color: not sure what I could use here. Here are some candidates
  • average d-score of interclass differentially expressed genes in a given pathway
  • average expression of all genes (regardless of whether in result from SAM) in a given pathway
  • q value for overrepresentation analysis (from the corrected hypergeometric test)

Any technical criticisms of the 3 I've proposed here or any other suggestions that prove more useful (i.e. to show that a given variation of the disease is dependent in some way upon some sort of activity (or lack) in a given pathway)?

Answer 1

[Adrian]

You might want to have a look at the Enrichment Map (Merico, Isserlin, Stueker, Emili and Bader) http://baderlab.org/Software/EnrichmentMap which is very similar to what you're suggesting.

That said, I'm not sure I see how the network figure you propose helps compare between the three groups, which sounds like the real question. I'd start with some simple tables / bar graphs.