Unifrac Design File Generation for No Treatments

Commands in mothur
1

I am new to this topic. Please don’t hesitate to tell me what’s what.

Backgroud
I am more-or-less following the MiSeq SOP (http://www.mothur.org/wiki/MiSeq_SOP) for 16S (Illumina V3 chemistry). I have generated my phylogenetic tree with clearcut, and I would like to compare each of my groups with each other group in the sequencing run without treatments labelling those groups. The Mothur wiki page on design files does not seem to cover how to generate a design file with this property (http://www.mothur.org/wiki/Design_File), and perhaps it is not possible.

What I have tried
One idea I had was formatting the design file with a unique treatment label for each group (as below), but I am not sure sure I trust the output of the weighted unifrac given that all the wscores=1. Although the scores were all equal, I noticed that the significance scores were not. I am not sure whether this is valid, or whether it is plausible for the weighted fraction of branch lengths to be entirely unshared.

group treatment_group
group1 G1
group2 G2
group3 G3
group4 G4

What I would like to learn

  1. How to use UniFrac functions to compare each of my sequencing groups
  2. Whether a weighted score of 1 in every group comparison indicates that I did something wrong given that I want to compare each of my sequencing groups to each other without grouping into treatments.
2

I’m not 100% on what you’re trying to do. Just using the groups file or count file with your tree will generate a distance matrix relating all samples to each other. If you want to pool samples within groups and then compare groups (not sure I’d recommend this) then you need to generate a new groups/count file assigning each sequence to the pooled group name. You can do this with the merge.groups command.

As for your scores being 1, is this the p-value or the unifrac distance?

Pat