--p-metric and beta diversity

[atgustin (andrew gustin)]

Hi there,

I recently started exploring the longitudinal analysis tutorial, and like previous users, I am unclear on how to identify which --p-metric to use for beta diversity measurements. (--p-metric for beta diversity) The answer feels a bit murky - is there not a standard value to input for --p-metric when the artifact is the weighted_unifrac_pcoa_results.qza?

I used metadata tabulate and generated the table, but that was uninformative in terms of the correct input. I know this issue was closed, but the conversation did not provide a clear answer.

Thanks for your time and assistance,

Andrew

[Nicholas_Bokulich]

Hi @atgustin,

@atgustin

like previous users, I am unclear on how to identify which --p-metric to use for beta diversity measurements.

The question in the post you linked to was answered in that post, and by the note in the tutorial: if in doubt, use metadata tabulate to look at your metadata or metadata-transformable artifact to see what column names are eligible.

I realize that's not a very satisfying solution, but it is workable — the issue is that sometimes the column names are predictable (e.g., calculating alpha diversity with metric="shannon" on your data gives you a SampleData[AlphaDiversity] artifact with the column name "shannon") but often they are not, and this is out of our control (e.g., calculating alpha diversity with metric="pielou_e" on your data gives you a SampleData[AlphaDiversity] artifact with the column name "evenness" I believe).

@atgustin

is there not a standard value to input for --p-metric when the artifact is the weighted_unifrac_pcoa_results.qza?

Well, yes and no — PCoA results are going to contain many different columns of coordinates. These should have consistent names ("Axis1", "Axis2", "Axis3", etc), but different PCoA results can contain different numbers of axes, so all of the axis names that are present in file 1 may not be present in file 2 if that file has fewer axes. Presumably you would only be interested in the first few PCs, but maybe not...

But more to the point — there are many different types of metadata-transformable artifacts, with all sorts of column names, and we can't list out every single artifact that is transformable as metadata, largely because 3rd-party plugin developers can create their own artifact types and transformers.

It would be very useful to list common types — e.g., if many folks are using PC results, it would be useful to give an example of this on the tutorial — and anyone is welcome to submit a pull request on github to add new examples to the tutorials if you want to help out

@atgustin

I used metadata tabulate and generated the table, but that was uninformative in terms of the correct input

Could you please provide more specifics please?

Sorry there is not a more satisfying answer... it's merely because there are so many different things that can be used as input (sample metadata, metadata-transformable artifacts, feature tables...) so we can't possible keep track of all possibilities!