Question on computing niche cell type composition

[ankitbioinfo]

Dear developers,

I have a question regarding computing niche cell type composition from following tutorial.
https://scenvi.readthedocs.io/en/latest/tutorial/MOp_MERFISH_tutorial.html

I have two spatial dataset where one has corresponding scRNAseq but another one does not.
Can I run envi for both dataset to get similar plot like in step 153? If yes then what modification need to done.
Also why FDL_COVET and DC_COVET need to transfer from both spatial dataset to scRNAseq at step 21-24.
Can I compute cell type composition without COVET part?

Thanks and apologies for asking layman questions.

[DoronHav]

Hi!

Thank you for your interest in our work! The plot in step 153 shows the neighborhood prediction for the single-cell data, so you would only need to apply in on your scRNA-seq dataset which can be integrated via ENVI using the corresponding spatial data. For the other spatial data, you can compute the niche composition directly from its spatial coordinates (see the function 'niche_cell_type' in our utils.py file for an example of how we did it).

FDL_COVET and DC_COVET are calculated from both COVET representations of the scRNA-seq and spatial data. The spatial data COVET is calculated from its coordinates, while scRNA-seq COVET is predicted by ENVI. Employing DC and FDL serves as an easy way to visualize and capture trends in the COVET representations (such as cortical depth).

For the spatial data, you can compute cell type composition without COVET (again, see utils.py for how we did it), but the scRNA-seq relies on the COVET prediction to infer its cell type composition.

Hope this answered all your questions!