Settings when analyzing functional abundances


I am using MaAsLin2 for a differential abundance analysis at genus level using settings that were recommended in a recent benchmal study Nearing et al. (see below example). I was wondering which settings are recommended if one wants to use MaAsLin2 if one analyses functional abundances. Should I treat it exactly the same or are different settings recommended here?

fit_data <- Maaslin2(
  output = "data/maaslin/test",
  transform = "AST",
  fixed_effects = c("treatment", "age"),
  #random_effects = "subject_id", 
  reference = "0",  
  normalization = "TSS",
  standardize = FALSE,
  min_prevalence = 0 # prev filtering already done before

By “functional abundances” do you mean something like UniRef90’s? In that case you may be better served by anpan.

1 Like

I mean output of e.g. picrust2, which seems to be also count data. On the homepage of the Huttenhower lab they mention that either taxonomic or functional abundances can be used but I was wondering if the setting should differ here.

Ah okay. The settings you have shown will likely work fine, but of course you’ll want to confirm that by inspecting the plots. In the Maaslin2 manuscript they also mention performing some upstream filtering of the input data for prevalence, you may want to look at those settings as well.

1 Like