Explore BayesSpace k results: are they related to pathology?

[lcolladotor]

One way to do this will be with boxplots for the percent of the spot covered by pTau and Abeta (separately) for each BayesSpace cluster. We'll likely make one page per sample (so PDFs with 10 pages) or one huge page with 10 facets. So in total this involves:

• Exploring BayesSpace k = 4 up to 15 (12 different ks)
• Two variables (could be 4): pTau and Abeta percent (could also be number of pTau/Abeta instead of %)
• 10 samples (or it could be that the signal is strong enough that it's visible across all samples combined)

To do this, we can use spatialLIBD::cluster_import() to read in the BayesSpace k search results from https://github.com/LieberInstitute/Visium_IF_AD/blob/master/code/08_harmony_BayesSpace/02_BayesSpace_k_search.R#L74-L78. You can use a similar code to https://github.com/LieberInstitute/Visium_IF_AD/blob/master/code/08_harmony_BayesSpace/02_BayesSpace_k_search.R#L35-L47 for loading the SpatialExperiment objects.

[lcolladotor]

Once #25 is done, we could check the frequency of the 5 pathological domains in each of the BayesSpace clusters for each k (and sample?).