FDR adjustment in MaAslin2-update

Downstream analysis and statistics MaAsLin
1

Hi there,
I am using MaAslin2 for the microbiota features wide association analysis. But I found that MaAslin2 performs the FDR adjustment for all the included covariates rather than just focusing on the one variable of interest.
In this way, is the default FDR adjustment in MaAslin2 somewhat over-adjustment?
I attached the codes I used below to show my observations.
Looking forward to hearing from your guys’ reply.

Thanks.

Kai
FDR adjustment in MaAslin2.txt (2.5 KB)

2

Hi @kevinluolk - you are right that the default FDR adjustment can be conservative especially when you have more than a handful of covariates to adjust for. Note that, we do not have a separate user-defined parameter to subset the particular metadata or contrast of interest (e.g., as done in limma and ANCOM), but you can subset the relevant variable(s) of interest from the results table and re-adjust the corresponding p-values although I agree an option to do this automatically will be helpful for users which I have now noted down to include in a future iteration.

Thanks for flagging,
Himel

3

Hi Himel,

Thank you so much for the reply. Looking forward to seeing the update of MaAslin in the next version, if possible.

Kai