How to identify statistically significant differential taxa between 6 groups


I am interested in identifying significant differential taxa between 6 groups using qiime 2. What is the appropriate analysis to perform and is there specific way to show the data in a graphical format?

Thank you for your help.


In qiime2, we have a few visualizations available.

With ANCOM, there are volcano plots you can draw with the ANCOM statistics. An example can be found in the moving pictures tutorial :

In gneiss, there is an assortment of tools ranging from heatmaps and boxplots. Examples can be found in this tutorial:


Thank you so much for the response. May you please explain why one should use GNEISS vs ANCOM vs LEFSE. The statistical approaches are somewhat different and I wanted to know what is the most appropriate methodology for my question. I also wanted to know what to do if I wanted to compare most significantly differential taxa in these six groups over 7 different time points. Again, I greatly appreciate your help.



Hi. You also can perform ANCOM first to identify which taxa are the most variable. After it, you can filter your table based on ANCOM output and perform Longitudinal only with those taxa. Longitudinal IMHO best pipeline for comparison between different timepoints in Qiime

1 Like

Unfortunately, there aren’t too many benchmarks on that published. In the supplemental material in the balance paper there was a simple example showing cases where t-test / Mann-Whitney U test can have false positive rates reaching 100%. While there aren’t benchmarks like this published on Lefse, I suspect that Lefse will have these sorts of issues since it relies on univariate tests.

For a comparison between ANCOM and Gneiss, I’d also check out the ANCOM paper in addition to the link above.


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