Repeated exposure measures with microbiome outcome

Downstream analysis and statistics MaAsLin
1

Hello,

I’ve enjoyed using MaAsLin2 in my research with both non-longitudinal studies and repeated microbiome measurements. I’m wondering now if it’s appropriate to use MaAsLin2 in a case where I have unequal repeated dietary exposure measurements (e.g., up to 3 repeated measures per subject at specific time points) and a single microbiome outcome at a later time point. I would be interested in understanding whether we have time-varying associations between the exposure and the microbiome (in theory by evaluating interaction between exposure and time point on the microbiome).

It seems like I could use MaAsLin2 and implement a mixed model to account for the correlation among the multiple exposure measurements by indicating random effects by participant ID and fixed effects for dietary exposure and potential confounders. Based on earlier forum responses it seems like I would need to first create interaction variable (diet * time point) to include in the model.

Would this make sense for this tool, or would you recommend searching for another tool that may fit better?

Thanks.

2

Hi @Lyoon6 - as you alluded to, you need to create an interaction variable (say, exposure*time) to detect time-varying associations between the exposure and the microbiome, and supply that as a fixed effect to MaAsLin 2. For inspiration, you can check out the MaAsLin 2 tutorial which covers an example on creating and testing interaction terms (Section 4.1). Of course, feel free to let us know if you come across any issues with the analysis.

Best,
Himel