Recipes¶
Three end-to-end walkthroughs for the datasets people actually bring to BIWT.
-
A dissociated dataset with no positions. You keep the composition and choose the population size.
-
Spot-level spatial transcriptomics with real tissue geometry. The recipe where units matter most — pixel coordinates are not microns.
-
Visium spots with per-type probabilities, expanded into individual cells.
Which one is yours?¶
| Your data has | Recipe |
|---|---|
| No coordinates | Non-spatial scRNA-seq |
| Coordinates and one cell-type label per row | Visium spatial |
Coordinates and *_probability columns |
Spot deconvolution |
If your coordinates are a UMAP or t-SNE embedding rather than tissue positions, treat the data as non-spatial — see which source the spatial query found.