I'm planning to run a GSE Analysis on some microarray datasets. Until now what I have done is to pre-processed them using frma algorithm.

The reason I chose frma instead of gcrma or rma is that I wanted to merge these different dataset into one.

Here is how I'm planing to proceed:


dataset1 : 3x Controls , 3x Treatment1
dataset2 : 2x Controls , 2x Treatment2
dataset3 : 2x Controls , 2x Treatment3

After merging

7x Controls , 3x Treatment1 , 2x Treatment2 , 2x Treatment3

Then I will execute the GSEA for controls , Treatmen1 , Treatment2 and Treatment3 separably.

So the question is if it's more reasonable to follow the path that I described or to run the GSEA on the separate datasets (for example for each control ) and then get the intersect of the gene sets of these three repetitions.

  • 1
    $\begingroup$ Is there any reason you don't want to or can't do both? $\endgroup$
    – Mitra
    Commented Nov 22, 2017 at 10:48
  • $\begingroup$ What have you done to control that each dataset is (presumably) done under different conditions and thus have batch effects? Are the controls the same in each dataset? Could you point me what is the frma algorithm ? It is the first time I read about it. $\endgroup$
    – llrs
    Commented Nov 22, 2017 at 10:56
  • $\begingroup$ @Mitra : No there is no such a reason. I'm just asking to see if something is more reasonable or makes more statistical sense than the other. $\endgroup$
    – J. Doe
    Commented Nov 22, 2017 at 11:48
  • $\begingroup$ @Llopis : At the merging stage I will remove the batch effects with comBat method (I could also try others as well). Here is a documentation for frma $\endgroup$
    – J. Doe
    Commented Nov 22, 2017 at 11:50
  • $\begingroup$ Thanks, take into account that this will remove too the effect of each treatment. as it cannot distinguish between the batch effect and the treatment effect. $\endgroup$
    – llrs
    Commented Nov 22, 2017 at 11:52

1 Answer 1


It makes more sense to evaluate by separate the pathway or gene set you want and see if in the three datasets result in a coherent message than to merge these datasets, as you will mix batch effects and treatment effects.

When in several datasets one looks for the same statistic it is usually considered meta analysis, not integration.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.