best strategy for analyzing functional changes over time

Downstream analysis and statistics MaAsLin
1

Hi everyone,

I am working with longitudinal calf microbiome data and would appreciate some advice regarding the best strategy for analyzing functional changes over time using PICRUSt2 outputs and MaAsLin2 in Galaxy.

My dataset includes:

  • Two groups: Supplement vs Control

  • Three sampling time points: March, April, and May

  • Repeated measurements from the same calves across time

I would like to identify functions/pathways whose behavior over time differs between the Supplement and Control groups.
In other words, I am not only looking for functions that are simply higher or lower in one group, but rather functions whose temporal trajectory changes due to the supplement.

The functional profiles were generated using PICRUSt2 (KO/pathway abundances).

My main question is:
What would be the most statistically appropriate way to model this in MaAsLin2/Galaxy?

Any recommendations or examples of similar longitudinal analyses would be greatly appreciated.

2

Hi,

For this time series set-up, I’d recommend the interaction model: ~ Group * Time + (1|Calf). There are a number of other threads about interaction models in the MaAsLin section of the forum, but in short, you’d get coefficients for Supplement vs. Control, each time point, and how much Supplement vs. Control matters at each time point. It’s technically possible to set this up with MaAsLin 2 in Galaxy, but I’d recommend just using MaAsLin 3 outside of Galaxy. There’s more information on how interaction models work in the tutorial.

Will