Dear all, 

I am performing a gene set analysis on RNA Seq data, using camera on Reactome database. 

camera returns a lot of significant pathways (more than 100) and I suspect that some of them are redundant, like small pathways included into larger ones and other imbrications.

I guess I have several solutions : 

- Sending this large lists of pathways to the biologists, which is complicated for them to study

- Keeping the most significant pathways, or pathways with largest absolute FC : is it relevant to do that? Or I will miss interesting small pathways? 

- Using a method that removes redundancy among pathways by computing an overlap measure but I did not find any satisfying method for the moment. I know that GAGE package has an esset.grp function that does a similar job but I do not find any equivalent for camera. Maybe I am missing it? 

Here is my question(s) : is there a usual solution among the 3 I suggested above? Do you know tools that would remove redundancy into Reactome pathways? 

Thank you very much for your help. 

Emeline 

My personal solution is the first -- to send all the significant pathways to my collaborators. In my experience, biologists are familiar with the idea that pathways will overlap and don't have any trouble with it. If a smaller pathway is part of a larger pathway, then we would often want them both to be significant, otherwise the results would be contradictory. camera is not biased towards larger or smaller pathways.

I haven't implemented an equivalent of the esset.grp function in limma because I distrust such an automatic approach. I would rather see all the results and use scientific understanding to interpret.