Selecting features for DEICODE biplot (Emperor vs R)

Hello :qiime2: community,

While making a biplot from Deicode results in R basing on this tuto (How to make pcoa biplot in R using q2-deicode ordination), I noticed that different features where being selected to plot as arrows in the graph in R and Emperor.

In the R code, features to plot are selected with the following cmd, mutate(a=sqrt(PC1^2+PC2^2)) %>% top_n(8, a) according to the tuto; after some attempts I added the PC3 in the cmd like mutate(a=sqrt(PC1^2+PC2^2+PC3^2)) %>% top_n(8, a) and the selected features between Emperor and R were the same.

Questions: is it correct to add the PC3 in the R calculation even if that "dimension" isn't going to be plotted? Which way is the best to select the features?.
Also, in the Emperor plot there is a Confidence value related to the arrows, I haven't been able to find it in the ordination file, where does it come from? should be taken into account while plotting in R?.

Luis Alfonso.


mutate(a=sqrt(PC1^2+PC2^2)) %>% top_n(8, a) is Pythagorean's theorem solving for the hypotenuse (i.e. distance from the origin) and then taking the 8 features that have the largest effect on PC1 and PC2. Perhaps the distance of PC3 is negligible and thus didn't change your value for a.

Not sure about the confidence value and how it is calculated; hopefully someone else could help you out there.

1 Like