Within-individual DE/DA with MrVI

Help scvi-tools
1

Hi all,

I am running MrVI on data with repeated measures per donor across timepoints. I want to test differential abundance across timepoints while accounting for donor IDs.

Current setup_anndata config:

MRVI.setup_anndata(
    adata,
    layer      = "counts",
    sample_key = "final_label", # combination of donor x timepoint
    batch_key  = "batch",
)

Current DA call:

da_res = model.differential_abundance(sample_cov_keys=["batch", "donor", "timepoint"\])

My question is: can I interpret donor_log_probs like it has been conditioned on batch and donor IDs?

Thanks in advance!

Best,
Mikhael

2

It is not.

However, you might artificially achieve that if you select the batch key of MRVI to be a combination of batch and donor.

Either way, you can also check it yourself by running the DA with “compute_log_enrichment”: True and see if their patterns correlates (donor and batch)