Some questions about best practices

[Dillon214]

Hello Tangram developers,

I'd like to ask a question that might help me clarify whether Tangram is a good choice for my large visium dataset.

So, it seems to me that Tangram works best when both the single cell RNA dataset and spatial dataset have the same cell types respresented in each. That is to say, that the single cell reference dataset ought to have all the same cell types you would expect in the visium dataset. Is this a hard rule? I have age and condition-matched visium and scRNA datasets, however the scRNA dataset I do have has been filtered to only contain cells of a single type. I wonder if I can still run Tangram to map these cells to the visium data, or whether this would be an invalid approach. I imagine the latter.

Any advice and guidance would be highly appreciated.

-Dillon

[gaddamshreya1]

Hi @Dillon214, I wouldn't its a hard rule but an assumption that Tangram makes for best results. You can still try mapping your single cell data with only one cell type to the visium data. In your case, Tangram will do its best to map that single cell type over the visium spots, however, I think it would require some interpretation based on the biology of the tissue section.

Another way you can do this is to find single cell refernce datasets that are as close as possible to the visium data you have and continue mappaing with Tangram.