Maaslin2 after Deseq2 normalization

Dear all,
I have some gene data that was normalized in the same way as Deseq2 does it. Now I would like to test it with Maaslin2 and my question is if Deseq2 normalization is suitable for Maaslin2 and if yes, should I just run it as it is or some additional normalization within Maaslin2 is recommended.


Hi @timyerg - ideally you should specify no normalization when running Maaslin2 if you have already normalized the data.


1 Like

Thank you for fast reply!

Yes, I disabled normalization in the Maaslin2 since data already normalized.
As I understood, your answer also means that Deseq2-like normalization is valid for Maaslin2 since no comments against it were maid :hugs:, so I am on the safe side :grinning:.


Correct, However, if you are supplying the normalized counts to MaAsLin2 you need to either log-transform the data to use the default LM or use one of the alternative non-LM models available in MaAsLin2.

Thanks a bunch,