Does MiloR take into account that the two compared conditions have different number of cells?

[crodriguezres]

Does MiloR take into account that the two compared conditions have different number of cells?

[DarioS]

I am just a user of the Milo package, but I think that Milo will not correctly model such a data set.

Under the assumption that the majority of neighbourhoods are not differentially abundant ...

[MikeDMorgan]

Hi @crodriguezres - the short answer is yes, up to a point. The TMM normalisation factors used in the NB-GLM and NB-GLMM seek to account for imbalances in the number of cells across samples. However, if all of condition A have ~1000 cells and all of B have ~100000 cells, then there is no guarantee that the results will not be driven by this difference (one order of magnitude might be OK, but you would need to evaluate this in your analysis).

If there are big differences in the numbers of cells, but not 100% confounded with your experimental design as you describe, the issue should be less severe. If you are concerned, then inspect the MA plots and look for a large deviation of from the expected median LFC of 0. plotNhoodMA can assist here.