Is it possible to use cell2location with non-spatial data?

[sam-israel]

Is it possible to use cell2location for deconvolution of non-spatial data?
I am interested in estimating the cell-type specific expression of each gene (starting from bulk RNA-Seq).

Besides knowing whether it is possible, I am interested to know if you would recommend it.

[vitkl]

Hi @sam-israel

It is possible to use cell2location for non-spatial data, however, the distributional assumptions of our data are tailored to UMI-based detection RNA protocols. So you would need to compare cell2location and other methods designed for bulk RNA-seq on your data. We have not benchmarked cell2location on bulk RNA-seq.

[sam-israel]

Thank you.
To pursue the same topic further :

Using Cell2location, it is possible to estimate cell-type specific expression of every gene in the data.

(I) Would this functionality also be possible for bulk RNA-Seq ?
(II) Would it be possible & recommended to use those estimates to perform cell-type aware differential expression analysis between two tissues - to see the differential expression by cell type?