# help understanding Unifrac

Hello,
I was wondering if someone can help me understand the proper way to use the Unifrac command in mothur. I understand Unifrac to be a distance measurement for looking at phylogenetic information. It would make sense to me that you could use a tree made with clearcut (from a phylip-formatted distance matrix) and then run the Unifrac commands on it. I am confused about how (as in the 454 SOP) we are putting a dissimilarity tree made using the thetayc calculator (or any other one) into the Unifrac command. Since thetayc weights for relative abundance, how does it make sense to then apply another weighted formula on top of it with Unifrac weighted? Aren’t we sort of…double weighting things?
I have spent some time trying to figure this out and I am getting deep in the weeds. I would appreciate any help!

It can be used in two ways. The first is as you mention, to calculate a beta-diversity metric. The second is as a hypothesis test where the input is a tree and you want to see whether the clustering/branching length in the tree is significantly different between groups.

Ok. But then how does the weighted vs unweighted unifrac add in when you are inputting an already “weighted” distance tree such as thetayc?

The weighted refers to how the branch length is portioned in the analysis. Frankly, I’m not sure why you would use the unifracs as a hypothesis test instead of amova/homova since unifrac requires converting the distance matrix to a tree, which causes you to lose information. amova/homova just use the distance matrix.

Pat

Thanks, I was wondering the same thing. It seems to make much more sense to just use AMOVA/HOMOVA.

Pat, in the 454 SOP for the Phylogeny-based analysis, you used
mothur > unifrac.unweighted(tree=final.phylip.tre, …
if i understand you correctly, it will be better to do the following instead:

mothur > amova(phylip=final.phylip.dist, …)

O.

many thanks

The commands in the SOP describe the many ways you can use mothur to do common analyses. The alpha/beta/phylog. based analyses are not meant to be a cook book for how to do an analysis.

Pat

it is a “learn-as-you-go” process, and the SOP and forum are great start. 