Beta diversity signicant differences but not in alpha-diversity

Hi all,
I have performed a nutrition interventional study in mice with five experimental groups. When I analyze beta-diversity (Permanova) I find signicant differences in all analyzed metrics: bray-curtis dissimilarity, Jaccard index and in weighted and unweighted Unifrac with significant differences between all groups and also between the different group pairs. However, when I analyzed alpha-diversity I don't find any difference in any parameter: shannon, observed features, faith PD..I don't know how to interpret my results. Can somebody help me?

Many thanks!!!
Mar

Hi @MarLar,

Welcome to the :qiime2: forum!

As a quick reminder, alpha diversity measures the difference within sample. Two samples can have the exact same value for alpha diversity and share no organsims.

Beta diversity compares the two communities, and asks what protion of features are shared and what portion are unique, for difference measurements of "shared".

A few questions to answer you:

  1. What was your sampling design? Do you have repeated measurements or a single time point?

  2. What was your timeline? Do you think it was long enough to cause a significant gain or loss of species compared to a shift in composition? (You might find Justin Sonenberg's mouse work useful here)

  3. How have you addressed cage effects in your analysis? Is it possible that the dietary effect you're seeing is just because of coprophagy?

  4. Can you find other examples of situations where the beta diversity changes, but not the alpha? (You might look at the Parkinson's literature). What are the justifications/mechanisms there?

Best,
Justine

3 Likes

Hi Justine,
Thank you for your answer!

  1. We have 3 time points for fecal samples (0, 2 weeks and 4 weeks) and one time point for mucosa sample. We think four weeks of diet are enough to find changes in microbiota but not to gain or loss of species (it is not a drastic treatmen).

Every experimental group consisted of 10 mice that are separated in two cages n=5 all of the same experimental group, so coprophagy is not our problem.

The results I am talking about where we found differences in beta-diversity but not in alpha are the results for mucosa-associated microbiota. Moreover, when we applied the ANCOM analysis to see differences in abundance we only found a couple of differentially significant taxa, which is also very strange given the results in beta-diversity.
Thank you very much for your help!!
Mar

Hi @MarLar,

So, given that beta diversity describes the between communtiy difference and alpha decribes the within community difference, what does this mean?

If your mice are co-housed, you need to at least consider the potential effect of coprophagy in modeling. You might consider nested modeling in your adonis, for instance. (Check out the vegan documentation for more details on how to do this.)

ANCOM is pretty conservative and with your group sizes, you may not have that many taxa that are detected as significantly different. I would also double check you W values to see how many were detected as differential abundant. I'm a big proponent of filtering, but YMMV.

Best,
Justine

I have tried with ANCOM-BC and I have found important differences in absolute abundances between the treatments.

Thank you very much for your help!!
Best
Mar

1 Like

I'm glad you found a technique that helps.

Best,
Justine