Tools for conducting a Cross validation study design

General Discussion
1

I am currently trying to conduct a study using a cross over model and was wondering if any Qiime-2 plugins would be useful or do I need to use another program such as R. For example, I know that the q-2 plugin can be used for longitudinal analysis and paired sample analyses, but I believe my study design does not fit either of those models. I’ve included a little information on the study to help as well.

The current study involves 10 animals being fed 10 different diets for a week at a time. On a given week, the animals will be assigned one of the 10 diets and then, following a washout period (end of the week), and randomly assigned a new diet. By the end of the study, each animal will have received each of the treatment diets.

From the design, I believe a cross-over model would be best to use and any recommendations on plugins would be highly appreciated!

Best,
Tyler

1 Like
2

Hello Tyler,

Welcome to the forums! :wave:

This sounds like a cool study!
:elephant: :green_salad: -> :stopwatch: -> :elephant: :burrito:
:elephant: :pie: -> :stopwatch: -> :elephant: :green_salad:

I think a lot of folks use a combination of tools when working with microbiome data, so Qiime2 + R could work, but so could Qiime2 + Python, or even Qiime2 + MATLAB if that's what you and your team likes using.

You may have found this already, but this made me think of the Q2 longitudinal tutorial. It's got great examples of how to handle samples from "the same individual subject observed at two different time points", which sounds exactly like your diet-swapping-animals.

Another good example is the PD-mouse tutorial, which also tracks specific animals across treatment groups.

If I may ask, what's your animal model? (:mouse2:?)
What's your data types? (16S amplicons, WGS, something else :test_tube:?)

Colin

2 Likes
4

Hey Colin!

I will have to check out the longitudinal tutorial for sure, thank you for that insight!

As for the animal model, it uses pigs and has 16S data, sorry I should have included that in the first post.

1 Like
7

Hey Tyler,

:pig: + :poop: + 16S amplicons is a classic!

Let us know how well that plugin works for you! Feel free to reply or open a new issue if you have other questions.

Colin

P.S. One of my colleagues reminded me that qiime longitudinal pairwise* compares only two timepoints, so if each animal switches diets multiple times resulting in a series of timepoints, ya might have to get creative

8

Colin,

What exactly do you mean by get creative? I have tried to use the pairwise distances but have had little luck. Do you have any further suggestions?

9

Hello again,

The qiime longitudinal pairwise-distances plugin will let you investigate a cross-validation study or one leg of a crossover study, like this:
Timepoint1: :green_salad: -> Timepoint2: :pizza:

It does not provide an elegant way to track many timepoints: like this
Timepoint1: :green_salad: -> Timepoint2: :pizza: -> Timepoint3: :green_salad:

Can you post what you tried, what you got, and what sort of output you were hoping for?

Colin