I am processing CITE-seq data, in which Cellbender is recommended to remove ambient RNA contamination to reduce the noise of gene and ADT signals. However, at the same time, I noticed that TotalVI also have the ‘denoise’ step. I am not sure whether these two steps will conflict? In addition, in which step should ‘isotype’ be input as control? There is also the question of whether the ‘Hashtag’ demultiplex step should be done before or after backgroud denoising?
It is generally speaking safe to run ambient removal with Cellbender before running VI models. Ambient and background might be two different things though as the distribution of antibody and in cells might be different due to effects like FC receptors.
As to setting isotype: I’m not sure as to what code you are referring?