Stratifying Alpha and Beta Group Significance


I was wondering if it is possible to run the qiime diversity alpha-group-signifance and beta-group-significance with only a portion of your data?
As an example I have both nasal and gut microbiome samples and I want to compare the difference between treatment and control groups. Is there a way to run the command (or another command) comparing treatment group, but stratified by origin site?
I’ve checked the --help features of both commands, as well as searching the qiime forum / google with no luck.

Thanks for any help for this Qiime-newbie!


Hello Kathryn,

Welcome to the Qiime 2 forums! :qiime2:

Yes. The main method is to use feature-table filter-samples to make two new tables that just have nasal samples and gut samples as shown here, then use the normal group-signifance plugins on each of these two tables.

Or try using qiime diversity adonis to compliment the results of beta-group-significance.

The adonis test lets you you can partition out sources of variance in order, so that you don't have to break them into seperate tables. For example, you could use the formula
--p-formula "tissue + treatment"
to show the variation explained by tissue, then show the variation explained by treatment after controlling for tissue.

Let me know if that's helped answer your question.



Enabling diversity analyses to use metadata-based subsetting instead of filtering the actual feature-table would be quite useful indeed! At the moment, @colinbrislawn suggestion is the most correct approach, but an alternative that I have been using myself recently is to simply add extra columns to my metadata file that includes the specific groups that I want and leaves the others empty. This works fine in most diversity actions (though perhaps not all, haven’t checked them all) and saves having to make a whole bunch of new feature-tables if you have a lot of different groups you want to compare.


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