Enrichment p-values

[hnsa...@gmail.com]

Dear EGAN team, 

Are the p-values calculated after enrichment are corrected? 

Kind regards

Hussein 

 

[ucsf egan]

Hello Hussein,

The p-values are not corrected at first.  However, adjusted p-values for the Hypergeometric enrichment tests can be calculated after performing the tests.  Simply right-click on the bottom table to dynamically calculate permutation-based adjusted p-values for the Gene Set (Association Node) type that you wish.  For example, for Gene Ontology Process, right-click on that row in the bottom table.  Then select the option to calculate adjusted p-values.

The method used is the Westfall-Young minP method, as that is robust to the high degree of correlation between gene set enrichment scores.

If you desire a different adjustment method (although beware of the correlation between gene sets - many methods are not appropriate), the raw values can be exported to a table using the button at the top of the table (TXT).

Best,

Jesse

[hnsaghire]

Hello Jesse, 

Thanks for the explanation. I have another question:

Many of the gene sets are overlapping between different databases like cell cycle can be found in GO and Reactome. But certain databases contain certain gene sets that are not in the others. My question is: it is necessary to limit our analysis to one database or we can use more than one? 

Kind regards

Hussein

[ucsf egan]

Hi Hussain,

Great question - that's exactly the value of EGAN.  Instead of just looking at tables and p-values of gene set enrichment, EGAN shows how different gene sets from different databases overlap and relate with respect to different genes from your significant set.  Knowing how different databases represent biomolecular concepts/pathways with different sets of genes is very important.

Can I ask a favor from you?  Can you post your question to the Google Group?  I'd like it (and the answer) to be available for other people to read publicly.  It's asked quite often.

Best,

Jesse

--
You received this message because you are subscribed to the Google Groups "UCSF EGAN" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ucsf-egan+...@googlegroups.com.
To post to this group, send email to ucsf...@googlegroups.com.
Visit this group at http://groups.google.com/group/ucsf-egan?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Arundhati Bakshi]

Dear Jesse,

What is the enrichment statistic that is given after enrichment calculations in the downloadable node table. Is that the p-value or the q-value or something completely different?

Also, is the calculation correct when we select a particular term to be "visible" but not "selected"?

Thank you,

Runa

[ucsf egan]

Hi Runa,

The visible enrichment score is enrichment of that set/pathway in the genes that are currently visible on the graph.  The selected enrichment score is enrichment of that set/pathway in the genes that are currently selected (they may be visible or they may not be visible).

The original enrichment scores are un-adjusted p-values.  q-values are difficult to produce due to the fact that so many pathways/sets overlap, and are therefore highly correlated.  If you would like to adjust the p-values for multiple testing, follow the procedure listed in the previous messages in this thread.

Hope that helps!

Jesse

--

You received this message because you are subscribed to the Google Groups "UCSF EGAN" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ucsf-egan+...@googlegroups.com.
To post to this group, send email to ucsf...@googlegroups.com.

Visit this group at http://groups.google.com/group/ucsf-egan.

[Arundhati "Runa" Bakshi]

Hi Jesse,

Thanks for the explanation. That helped a lot.

I have another question. I was wondering how important the minP method of correction really is for a gene set that I obtained from a single experiment and I am trying to show multiple enriched pathways/associations in it. The reason I ask is, I tried to follow the steps outlined in the previous message on this thread and the P-values did not change hardly at all for the visible enrichment statistic. Would you suggest trying some of the other methods, like PermuSEED or ParaSEED? What is the best way to decide which of these tests would be best?

Thank you for your time.

Runa

[ucsf egan]

I would try them both, although I've been having problems with them recently.  Please let me know how it goes for you!

Best,

Jesse