[BioC] ANOVA for gene ontology analysis

[Gordon K Smyth]

Dear January,

I don't know entirely what you have in mind from your brief description, 
but it does sound like the method you are proposing would entirely ignore 
inter-gene correlations, which can be catastrophic, see:

  http://nar.oxfordjournals.org/content/early/2012/05/24/nar.gks461

Have you consider the roast method in the limma package:

  http://bioinformatics.oxfordjournals.org/content/26/17/2176

which has the flavor of an anova approach but fully acounts for inter-gene 
correlation.

Best wishes
Gordon


------------ original message ------------
[BioC] ANOVA for gene ontology analysis
January Weiner january.weiner at gmail.com
Wed Oct 17 13:36:36 CEST 2012

Hello,

I would like to test the significance of GO terms using ANOVA. Correct
me if I'm wrong, but I think that it makes sense to make a three
factorial ANOVA (where genes are one factor,  the experimental group
is the other, and the specific sample is the third) on a group of
selected genes (for example, group of genes belonging to a given GO
annotation group).

Clearly, this approach might have less power in case of small sample
sizes, but could be better than enrichment tests for larger sample
sizes.

Is there an R package that can use the existing annotation
environments for running ANOVA?

Kind regards,
January

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}