How to interpret ANCOM results when the sample numbers are not even

User Support
1

Hello friends,
I am using ANCOM to test the differential abundant genera in my study. I have two groups to compare, one is positive with a specific pathogen, the other is negative.

I understand the volcano plot, which means higher W means it is statistically significant between the two groups. See the plot below, I got several differentially abundant genera.

Screen Shot 2022-05-20 at 12.09.40 PM
Screen Shot 2022-05-20 at 12.09.40 PM970×1031 110 KB

However, I am not sure, which genus is more abundant in the Negative group compared with the Positive group. Because I have different numbers of samples in each group, so I can't just compare the percentile abundances of features by group. For example, if the Negative group has 10,000 reads in total, and one genus has 200 at 100%, while Positive group has 20,000 reads in total and the same genus has 150 at 100%. Then just by looking at 100%, I will conclude that this genus is more abundant in Positive group than in the Negative group, but that's not true.
For example, if you check Barnesiella in the plot below, you will see that in Negative group, the 50 % is 1 but the 100 % is 2335; while in the Positive group, the 50 % is 39 and the 100 % is 753. How do I interpret whether Barnesiella is more abundant in Negative group or not?

I am not sure if I've explained my question clear. Please let me know if you need any other informaiton.

Screen Shot 2022-05-20 at 12.09.54 PM
Screen Shot 2022-05-20 at 12.09.54 PM1910×454 178 KB

4

Hi @JinjiPang,

Welcome to the :qiime2: forum!

This is such a nice volcano plot! :volcano: :nerd_face:

ANCOM takes the centered log ratio for each group (hence the clr title on the x-axis) - which accounts for differing read numbers between your groups. What you could do to get a tabular representation of the volcano plot would be to download the visualization data as a TSV file (click on Export as TSV from the upper left hand corner of your volcano plot), and use that data in the Qurro community plugin, which allows for clr data comparisons. This should provide you with the appropriate comparisons between your positive and negative groups!

Hope this helps! :lizard:

6

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