xiangpin
commented
2 years ago In the first step, the LEfSe uses the no corrected pvalue of kruskal.test. But mp_diff_analysis uses the corrected FDR of kruskal.test in default. In the second step, The LEfSe use 0.05 * 2 pvalue to choose the feature, but the mp_diff_analysis use 0.05 to choose the significant features. In addition, The method of LEfSe might have a problem for the calculation of LDA value, it used the random seed in the function, which might cause the subset dataset will be the same for each subset. This means it is not cross-validation-like.
Hello, I would like to know what is the difference between the lefse in this package and the one on the Galaxy - The Huttenhower Labwebsite. a RareAbundance is introduced here， I compared the two results, this package got fewer results.