Microbiome Analysis: How Can I Find Significant Differences Between Cases Using Metaphlan Output?

Question

[Fedor Gusev]

I have some data from process with metaphlan from one control and many cases. The data looks something like this:

Bacteria|Verrucomicrobia 100 30 40 60 20
Bacteria|Elusimicrobia 20 30 40 20 40
etc.

First column stands for my patient, others for controls. Threre are tools for find out what is different in between patient and controls, like LEfSe and Metastats. But, unfortunately, they seem to be only working when you have at least two samples in each group.

How do I proceed in this situation?

I can, of course, take my only sample and create a artificial simular one, taking into account divergence among controls, But that just does not seem right...

[Cliff Beall]

I hate to tell you, but I think you really need more samples.

Answer 1

[Istvan Albert]

Maybe you could try a simple contingency table based statistical test, like chi-square test. You could take the total number of sequenced reads at root level to be the expected frequencies, then for each row you have the observed frequencies.

[Josh Herr]

I second Istvan's recommendation. This has been helpful for me when using R. Also, if you haven't seen this check it out.

[Fedor Gusev]

This is what I used at the end. But I contacted the authors of LEfSe and they fixed a bug, so now I am using that. Thanks a lot anyways!