# Ordination using unifrac distances?

Hi
I was asked to plot NMDS or PCoA using unifrac distances (as it is done with quiime). But I cannot find an option to do that, only to use dissimilarities (if I am not wrong, this is to do ordination based on distances from binned data??) and Unifrac distances would be to get an ordination based on distances calculated on phylogenetic data? I’m sorry, I am confused and cannot find the way to get the unifrac distances calculated from a phylogenetic tree clustering the OTUs (the representative sequences from each OTU). Is that possible with Mothur??

We do this in the 454 SOP - see the phylogeentic approach towards the end. You’ll generate a tree from your sequences, then run the unifracs to generate a distance matrix and then build the ordination. or you could tell the reviewer to go put their head where the sun doesn’t shine.

Yeah! youÂ´re right!

And sorry, I thought of several ways to do that and completely forgot it was explained in the SOP :oops:

So, quiime defines OTUs based on distances but then also build a tree using a phylogenetic approach for betadiversity? hmmmm

In your opinion, clearcut is better to build the tree than neighbor function in Phylip? I read somewhere that is possible to copy the neighbor command from phylip into mothur and run it within mothur, is that true? Because once I tried clearcut and the tree I got was not good, but may be then I did sth wrong.

Sus

neighbor and clearcut should give a very similar tree. the clearcut neighbor joining is faster than the phylip command. also, we use the “real” algorithm from clearcut, not the heuristic.

Thanks for your answer! one more question… after PCoA or NMDS with unifrac distances after the phylogenetic approach, is it also possible to use corr.axes for OTUs ? how should I do? I am not sure about using the shared file from the OTU approach and mix up both approaches, and cannot find from the phylogeny approach a relabund file. I’m sorry :oops:

You should use a sub.sampled shared file.

Pat