Combat for microbiome batch effects

Question

[ymj]

Hi

Does anyone have experience with the Combat method for removal of batch effects in microbiome analysis.

I have applied this on my dataset, since there was a time batch effects. Samples from timepoint 1 were having much lower number of reads, compared to timepoint 7 and 13. I now get a normal-like distribution for the batch-corrected CLR values. However, when I plot the core microbiota and the top15 phylum, this is what I get:

It seems like not what I would expect.

Can anyone help? Or are there other methods than Combat that I can use to remove batch effects?

Thank so much in advance!

Answer 1

[guyshur]

This nice review summarizes the state of this issue: https://academic.oup.com/bib/article-abstract/21/6/1954/5643537?redirectedFrom=fulltext