Re: [GenMAPP] WikiPathways vs KEGG Pathways

[Nathan Salomonis]

Hi Rad,

Thanks for your email, see below:

Although I appreciate the fact of including WikiPathways for ORA, if you analyze Gene Expression data using other approaches, specifically command line tools and bioconductor ones, you will notice some differences in the enrichment results. This may be due to the statistical approach you are using, (z-score, fisher exact test instead of hypergeometric tests), in another hand you seem not to use KEGG for ORA (you do it though through GO-Elite not incorporated in GENMAPP-CS.

I did some GE analysis and I found for an experiement (mouse) that I have some genes enriched in P53 pathway which is not existent in WikiPathways. Although I have some overlap with your package, (Kegg vs WikiPathways) this particular pathway which is a key in my study is not existant.

The permutation based p-values are fairly unique to GO-Elite/GenMAPP and were recommended over other alternatives since this approach makes no assumptions about the underlying distribution of data. However, the latest standalone version also has a Fisher's Exact option. We need to update the GO-Elite webservice and the GO-Elite GenMAPP-CS plugin to accommodate for this (putting in a feature request now). See a more detailed description/comparison of these stats here:

http://www.genmapp.org/go_elite/help.htm#ora

As for KEGG, this is also something we need to update in the Webservice.

There are two issues here (1) analysis of non-GO and non-WikiPathway gene-sets and (2) automatic updates of WikiPathways via the webservice. In GO-Elite version 1.2.5 and above we support KEGG, PathwayCommons, WikiPathways, GO, Disease Ontology, Phenotype Ontology, transcription factor target, miRNA target and other gene-set over-representation analyses. We have been holding off on updating the web service to the latest version of GO-Elite, since we have to make some changes to accommodate for these new options. GenMAPP-CS only runs one gene-set type at a time, which means we could add the option to run KEGG, but no pathway visualization would be available since these pathways are not available at WikiPathways. Some KEGG pathways however have been manually curated over, but automatic translation is usually not pretty in terms of layout with sometimes dozens of gene IDs per KEGG node (which is why automatic KEGG translations are excluded now).

There are a couple p53 related pathways though: WP1742, WP656, WP655. The latter is rat and was manually converted from KEGG. To get into human they have to be translated automatically from the GPML or manually (download and switch all of the IDs) and then add to WikiPathways as a new pathway. I've done this with WP655 (KEGG p53 pathway - see the attached) using a script I wrote. You can cleanup this pathway and then add to WikiPathways. Once added, if you update all WikiPathways in GO-Elite (stand-alone) it will then show up as a WikiPathway option that can be colored. However, for the GO-Elite plugin, we currently don't update frequently so this wouldn't show up as an option in GenMAPP-CS. I will also add this as a feature request for the GO-Elite webservice (e.g., update every 24 hours).

Let me know if I can answer other or related questions.

Best,
Nathan