Ways To Do Gene Ontology Analysis For Microarray Data

Question

[gundalav]

I have collections of probe/genes from 3 different conditions (e.g. stimulated by chemical1 2 3). They are already selected based on some fold change threshold. What I want to do now is to look at GO that are preferentially occur in each of the above conditions.

I face two following ways to do it, which I'm not sure which one is the correct one to do:

1. Obtain GO for each of the conditions together with the p-value of the GO term. Then perform clustering (hierarchical/knn) on these p-values.

2. Perform the clustering for the foldchange or the expression values. Then for each of the clusters determine the GO.

Your advices would be truly appreciated.

g.v.

Answer 1

[Devon Ryan]

I would suggest option 3, performing the DE tests such that you answer the actual biological question you're trying to ask (e.g., chemical 1 vs. 2 or 1 vs. 2 and 3 as a group) and then calculate GO enrichment on that. You can then simply avoid clustering altogether and have more confident results.