I have been experiencing some confusing results from LEfSe lately.
In a dataset with only two groups, LEfSe indicates that it detects a certain number of “significant features with given criteria” which does not match the bar plot produced which, in turn, does not match the features highlighted in the table. For example:
After trying a few other scenarios, similar inconsistency is observed. For example:
It seems as if LEfSe only mentions as significant features those with positive LDA scores after selecting on its own which group to consider as a reference. However, in the Result Table tab, random samples get selected as significant ones, regardless of the selected cutoff criteria or LDA score direction.
Currently, I am focused only on the bar plots and the table contents, which seem to be matching. However, given such inconsistency, is it okay to trust that LEfSe analysis was performed correctly and these results are trustworthy?
P.S. Such an issue did not exist a few weeks ago, prior to the updates.