LDA score treatment comparison: All treatments in one table or separate table for each treatment?


I have an OTU relative abundance table with 6 treatments including the control. When I input the OTU file with all treatments I get an LDA plot with all treatments listed. When I made separate OTU tables with one specific treatment and the control, I get an entirely different LDA score plot with new features listed along with negative LDA scores (the one with all treatments had positive scores). Which way is most appropriate when I want to compare differences between a specific treatment and the control class? What is causing these differences? Is the first plot comparing all classes with one another?

Hi -
According to the LEfSe publication, when there are multiple classes, LEfSe reports the highest score detected among all pairwise class comparisons, which would explain why you’re always seeing positive LDA scores for the first plot. If the primary goal is to compare treatment vs. controls, then I think your second approach might be more appropriate.