Density and/or proportion changes for each cell type across region

[lcolladotor]

While we expect a positive relationship in #125 between the number of white matter spots and the proportion of WM-enriched cell types observed in the snRNA-seq data, that is, a relationship between the spatial area and the cell types, I wonder if there are differences in the (a) density and/or (b) proportion of these cell types between region (ant, mid, post).

That led to @lahuuki and I drawing this

Ultimately, we might need to make boxplots of the mean number of oligos (likely repeat for other cell types) per spot using the spot-deconvolution results by @Nick-Eagles, across the 30 Visium samples split by region (ant, mid, post). We would do this separately for each BayesSpace cluster (we need to choose whether to do this at k = 9 or k = 16; likely k = 9). This will be different from #125 where we were only using the overlapping 19 samples. While I think there's probably no reason to expect any differences a priori in terms of density (mean number of oligos per Visium spot) or proportion (mean proportion of oligos per Visium spot) across regions, maybe there is. Hence the idea to check.

So it would be 3 boxplots (region) per BayesSpace cluster (one page of a PDF or a panel) for each cell type, then also for the mean number and mean proportion. So that's k (BayesSpace clusters) * n_cell_types (broad resolution only) * 2 (mean number, mean proportion) PDF pages or panels.

I imagine that @lahuuki will do this, but maybe @Nick-Eagles could. Hence why I'm assigned it to both of them for now.

[lcolladotor]

Here we'll use the results from #129.

[lcolladotor]

I believe that we can close this one @Nick-Eagles @boyiguo1, right?