Question to your LEfSe

This is Prof. T. Shun Sato, at Department of Biostatistics, Kyoto
University School of Public Health.

One of our students attended your short course at the JSM2019, and has
great interest on your LEfSe.

We have read your 2011 Genome Biology paper, and have a couple of questions.

Although there was no mention in your 2011 paper, looking for sample
data, it seems that LEfSe is designed for the analysis of
compositional data (microbiome data are transformed to percentages).
Is that right?

If so, the simulation in your 2011 paper (for example, Figure 5)
seemed that data were not transformed to compositional data. Why did
you use untransformed data in your simulation?

We are very pleased for you to answer the above questions.
(I’m afraid that we made a serious misunderstanding on your 2011 paper.)

Thank you, in advance.

Hi - thanks for the question, and sorry for the late response! LEfSe will automatically normalize data internally, so you can consider the “raw” inputs counts are implicitly normalized to relative abundances.