Is MaAsLin appropriate for a crossover study design?

Downstream analysis and statistics MaAsLin
1

Hi all,

I’m wondering if you have any thoughts on whether maaslin2 is appropriate for a 3x3 (or k x k) cross-over study design. While technically the design has repeated measurements of the microbiome where a mixed model might be appropriate, it doesn’t seem like we can account for sequence or period with this model. Would you agree? I wonder if there is a method that is more suited to the crossover design than a straightforward longitudinal design.

Interested in your thoughts.

Thanks!

2

Hi @Lyoon6 - you might be able to run MaAsLin2’s mixed-effects model if you can prepare your crossover data in the desired format (see e.g. https://stat.ethz.ch/pipermail/r-sig-mixed-models/2013q2/020443.html). As you have speculated, this is not straightforward using the default MaAsLin2 pipeline, but it seems to be possible based on the description here: https://stats.stackexchange.com/questions/373191/how-should-i-specify-random-effects-in-a-crossover-study-with-lots-of-repeated-m. Hope this helps!

3

Thanks! I’ll look into these possibility. I’m new to the microbiome field and was hoping these was straightforward method for a crossover design with microbiome data but I guess it’s not done often enough to warrant a separate package.