Using unifrac distances in AMOVA

Theory behind mothur
1

This question is really 2 part:

  1. In the Schloss SOP you generate Yu-Clayton distances from a shared file and test the significance of the clustering with AMOVA. Is there any fundamental reason this couldn’t be performed with unifrac distances? (as in, it’s quite easy to do in mothur, but is there a reason I SHOULDN’T do it?)

  2. Using the unifrac p-value alone to look for statistically significant changes in the population membership/structure, what’s the best way to account for any replication in the data? For example, If I had 10 samples, which represent two communities with 5 technical/biological replications, when I use the significance test (random=T) mothur will give me pairwise significance values. The only way I can see to test between the communities would be to merge my groups file according to a design file and then use that. Alternatively, I guess I could use the distances from the first unifrac as input for a new tree and then test that according to a design file, but that feels dodgy (a unifrac of a unifrac).

Any advice would be appreciated!

2

  1. You should be fine.

  2. the design file will allow you to incorporate the replication.