Using Maaslin2 with centered log ratio transformation

Hi, I have a question about data transformation prior to input into Maaslin2.

I am thinking of transforming a metagenomics abundance dataset using centered log ratio, then input into Maaslin2 with parameters transform = "none", normalization ="None". Other parameters remain default.

Is this allowed?

Reason for doing so is because I wanted to use Maaslin2’s capability to control for covariates but I could not find any other tool that is compatible with both covariate adjustment and centered log ratio transformation.

Hi @jjsimz,

I believe CLR is an option in MaAsLin 2, see here under the normalization section. Therefore, if you have already CLR normalized your data, yes your suggestions above should work just fine under the hood.

This post, also includes a bit more information on the different methods that MaAsLin has available.

I hope this helps!
Best,
Kelsey

Hi @Kelsey_Thompson,
I somehow thought that CLR wasn’t supported by Maaslin2, but it seems that I was mistaken. Many thanks for the correction!

1 Like