Could Masslin2 control more than 3 covariates at same time? and How to set interaction with microbiome relative abundance and one of the variable of interest?

Downstream analysis and statistics MaAsLin
1

Dear Sir or Madam,

When I built a model that with one predictor (A) and three covariates(age, gender, BMI) to predict a phenotype, I am not sure that Masslin2 put them all in one single model in analysis or just separate every time.

The second question is if I would like to make up interaction term that contain microbiome relative abundance and one variable such as fiber intake. In this situation, what should I do in the analysis?

Thanks,

Jun

2

Hi Jun,
I’m not part of the Maaslin2 team, but I am also interested in adding interaction terms. I submitted a PR to enable that; any feedback would be appreciated! add support for interaction terms by nickp60 · Pull Request #9 · biobakery/Maaslin2 · GitHub

Nick

3

Hi @jxu4,

The easiest way to check and see if MaAsLin is completing its expected behavior (e.g. including all covariates in the model), is to check the log file. In the generated output you should see a file called maaslin2.log, towards the top of the file MaAsLin will print the model that it ran. It should be able to adjust for more than three covariates.

As for the second question - as the species/taxonomy is usually the outcome and the phenotype is usually the predictor it would be hard to model an interaction term between them - I believe. I might be confused by your question and if that is the case - please see the interaction section of our tutorial. This is something we are planning on adding additional support for in a future version of MaAsLin - thanks @nickp60 for the PR.

Best,
Kelsey