When using LEfSe without subclasses

[Crispy13]

Dear all.

Hi. I’m working on a dataset including microbiome data and clinical data.
I want to analysis difference of abundance across clinical parameters,(disease, sex, age…)

As far as i know, LEfSe does 2 statistical test, of which one uses class and the other use subclass.
My question is… when a dataset have no subclasses to compare, could i use LEfSe algorithm?

My colleague showed a ancom analysis result on the same dataset and i found the lefse result show me much more otus than the ancom analysis…

Regards,
David.

[jwdebelius (Justine Debelius)]

Hi @Crispy13,

I’ve moved this to “other tools”, since LEfSe isn’t in QIIME, or part of a QIIME plugin AFAIK.

Best,
Justine

[SoilRotifer (Mike Robeson)]

Hi @Crispy13,

My understanding is that defining a "subclass" is optional. That is when you can subdivide your data in such a way that your...

subclasses should be subsidiaries or subdivisions of a single Class

Check out the phrasing of these posts form the LEfSe-users group and QIIME 2 forum:

• LefSe Class and Subclass
• how does subclass influence the LEfSe result? This is where the above quote is from
• Lefse after QIIME2 Note the mention of needing 2-3 rows at the top of your LEfSe formatted input file, that is you may or may not have a subclass for your data.

-Best wishes!

[Crispy13]

Thank you for reply!
So, not using subclasses is not a problem…

If a class is bahaviors(or symptoms) of disease, is it make sense to designate prognosis as a subclass?