I assume you’re using Qurro with Songbird outputs? You may want to check out the section titled “Interpreting ranks” in this paper. In particular, this paragraph might be useful:
The other possibility is to identify candidate differentially abundant microbes. To this end, one can construct rank plots (e.g., Figs. 2b and 3a). The rank plots show the ordering of all of the taxa with respect to how much they are associated with a particular metadata covariate, and specific taxa can be highlighted to show their ranks as positions on the rank plot. From the ranks, one can focus on taxa that have very high ranks or very low ranks, since those are the ones that are increasing/decreasing the most relative to each other, and are likely to be important contributors.
I personally like using autoselection to do this. If you have a certain covariate you’re interested in (e.g.
Oxygen in the Red Sea data), you can set the
Differential field to
Oxygen, and then auto-select the log-ratio of e.g. the top to bottom 10% of features. This should do a decent job of selecting a log-ratio that is associated with samples’
Oxygen values, since you’re just looking at the features “…that have very high ranks or very low ranks.” From these “autoselected” features, you can investigate further to see what types of features are at these extreme positions, and use these results to determine a hypothesis (e.g. maybe a certain type of genus is highly ranked? for the Red Sea demo data, the features are KEGG orthologs, so you may want to check out their
If you have further questions, we have some tutorials showing how to use Qurro here. You may be particularly interested in this tutorial, courtesy of @gibsramen, which uses Qurro to look at ALDEx2 outputs: section 6 covers how to use autoselection and some ways of investigating the results.