MaAsLin2 with metagenome inference?

Downstream analysis and statistics MaAsLin
1

Hello,

I’m working on a project that uses MaAsLin2 for multivariable modeling of feature (taxa)-metadata association. My ASVs are agglomerated at the genus level, and I have about ~200 of them. I’ve run several different models (adjusting for different sets of potential confounders) and my results are null: no features associated with my exposure of interest.

I have also run my data through the QIIME2 PICRUSt2 plugin and have a feature table of pathways as well as mapfiles. Can I run a similar MaAsLin2 pipeline for these pathway features, in a similar manner to the standard taxa pipeline?

  • Edit: I just realized that the documentation specifies that we can use genes and/or pathways with MaAsLin2 as well. - Nevermind!

Thanks.
L

2

Hi @Lyoon6 - as you have already figured out, you can use functional profiles such as genes or pathways in MaAsLin 2 similar to taxonomic profiles.

All the best,
Himel

3

Thanks, @himel.mallick .

I do have a follow-up question: Are there any reasons to be concerned about the default normalization/transformations that MaAsLin2 uses with respect to the pathway data? For instance, if I were to directly input my MetaCyc abundance data into MaAslin2, are the default TSS-normalization, Log-transformations, and minimum prevalence appropriate? I want to make sure that I am preprocessing appropriately.

Thanks again.

4

Hi @Lyoon6 - the default setting should be fine for pathways although filtering should always be done in a data-dependent manner. In general, the goal of filtering is to ensure that only the most informative features are retained for final analysis and there is no single best way to achieve that.

Best,
Himel

1 Like