Y-shape bray-curtis-emperor

Dear all,

I would appreciate any help with an outcome from Bray Curtis analysis. The emperor looks like this:

I used this command: qiime diversity core-metrics-phylogenetic --i-table ../QIIME2/Table_Dada2.qza --i-phylogeny rooted-tree.qza --p-sampling-depth 1255 --m-metadata-file ../mapping_file.tsv --p-n-jobs-or-threads 28 --output-dir trial_2

I've found that it may be caused by "a lack of overlap between studies", but this is only one study. Unifrac emperor looks fine, Jaccard looks as well as Bray Curtis.

I'm working at Conda.

Thank you all!

Please, check this discussion on similar topic. Hope that it will clarify your question as well.


Thank you for this reply. But my data is from a single run and I treated them all the same way.

Hello again!
I asked for help from other mods and will share with you great answer of @ebolyen (thanks a lot for your help):

This is happening because Bray-Curtis can only make an informed distance if there's ASVs in common. When there aren't any (or very few) this is a pretty typical thing to see as essentially everything is as "distant" as every other thing, so the best the ordination can do is make orthogonal spokes which then get projected down to 3D.

Unifrac doesn't care because even though the ASVs are different, they map to similar locations in the phylogeny, and so from it's perspective, things are pretty normal looking (an ASV that is missing only a single nucleotide relative to another ASV will still have essentially the same branch lengths to the entire tree unless it was a very exciting nucleotide).

That's mean that you don't have shared ASVs between certain groups of samples, or you have to few of them.


It sounds good, thank you a lot!

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.