Pairwise comparisons and multiple comparisons corrections

Downstream analysis and statistics LEfSe
1

Hello all,
I am in the process of performing analysis of microbiota data from a study where we compare three different groups (lets call them A, B, and controls). I have managed to perform the LEfSe analysis and form Cladograms for all three groups together, and observed some differences. However, while the overall bacterial communities that describe the individual groups are interestingly, the comparisons I am mostly interested in are comparing:
A vs control
B vs control
A vs B

I know from PCoA plots and other investigations that A and B are highly similar, but different from the controls. Thus I wish to perform the individual comparisons separately. My question is thus:
Can I perform the pairwise comparisons directly using LEfSe, or should I generate different input data.
Do you have any suggestions for correcting for multiple comparisons? E.g. BH?

I am currently working with the Galaxy module, and are starting to familiarize myself with the python package.

Thanks for the help
Caspar

2

Hi @Caspar_Bundgaard,

Your approach should work and yes BH correction would be fine.

If however you are more interested in the A vs. control and the B vs. control comparisons, I would should switching to MaAsLin and setting control as your reference level. That would do both of those tests and it would already MHC correct the results. Further you would be able to adjust for any other variables that you are interested in the model as well. Information on MaAsLin is located here.

I hope this helps!
Best,
Kelsey