[Should I used CAGE?]: gene set enrichment analysis on image analysis experimental values

Question

[Aurelie MLB]

Hello,

I have experimental values for a large set of genes that are NOT coming from sequencing. They are image analysis derived GFP intensities. From these values, some of those genes are interesting and some are not so I could if needed define a set of "interesting" genes. But I would prefer to take the actual experimental values into account.

I would like to analyse my interesting genes and understand what pathways and bioprocesses are triggered (compared to the rest of my genes).

I have seen many get set enrichment analyses but they all seem to work on p-values (or other stats) or log fold changes (like GAGE). May I ask some expert views on what tool I should use please? Could I use CAGE for instance and replace the log fold changes as input by the intensities?

Many thanks

Answer 1

[Jean-Karim Heriche]

You have to make a distinction between the methods used to do enrichment analysis (in particular their assumptions) and their implementation. Enrichment analysis is in principle not restricted to a particular experimental type but some implementations may be for convenience. If you can partition your genes to get a contingency table, you can also do standard enrichment analysis using Fisher's exact test. Alternatively, you can run GSEA on a list of ranked genes, independently from the method used to ranked them. As far as I understand it, for GAGE, you need paired control and conditions and that the change you measure is normally-distributed (because GAGE is based on the use of t-tests) and satisfies the assumption that the measured changes for biologically relevant groups of genes are in the same direction.

[Aurelie MLB]

Thanks a lot! This is helpful.