Including host phylogeny as random effect

Downstream analysis and statistics MaAsLin
1

Hi all,

I’m using Maaslin3 for the first time and I have a few questions about random effects. I have a dataset that includes 3 microbiome sample types per individual, with individuals from different species. I’m interested in the differences in presence/absence and relative abundances between the sample types. I’m currently using the following formula: formula = '~ sample_type + reads + (1|subject_id)’, where reads = sample read depth, and with small random effects.

I have the following questions:

  1. Since there are multiple host species in the dataset, is it sensible to add host species identity and host phylogeny as added random effects to the formula? Or is it sufficient to add only subject ID as a random effect, since subject ID and host species+phylogeny are collinear?
  2. If I do want to add host species and phylogeny as random effects, how can I do this within Maaslin3? I have a species co-variance matrix with phylogentic distances between species, is it possible to add this? Or can I add for example phylogenetic eigenvectors?

Many thanks in advance!

2

Hi,

If you already have subject IDs, you don’t need to include host species or phylogeny as random effects since they’re collinear (the subject ID determines which host species it is). You might still include host species as a fixed effect if you’re interested in differences among the hosts, but that’s up to you.

Will