I have a question please, I am running ANCOM for ITS data comparing a group of patients and control healthy subjects.
I included 2 mock communities, 3 positive control samples, in the experiment. Should I filter those before running ANCOM? how their presence could affect the results?
By leaving those samples you are essentially adding additional groups in the analysis, which isn’t really what you want, so filtering them out would be a good idea.
An alternative which I think works (I just can’t remember from the top of my head), when running the ANCOM test there is a required input parameter which requires you to describe which metadata column to use for assigning groups to yours samples. I think if you were to just simply leave the rows in that column corresponding to your positive controls empty then ANCOM should just ignore those samples in its comparison and only compare the 2 groups of interest. The parameter:
--m-metadata-column COLUMN MetadataColumn[Categorical] The categorical sample metadata column to test for differential abundance across. [required]
So basically I’m just proposing to manipulate the metadata file instead of filtering which is just a little bit less work.