Thanks for reaching out. BayesSpace does not use scRNAseq data. In the manuscript, we only used scRNAseq/deconvolution methods for validation of the BayesSpace method.

@edward130603 thank you -- since spatial transcriptomics are not single cell resolution after enhancing the clustering and resolution to subspot level - is there downstream analysis to then deconvolute that sub-spot level of resolution?

We have not specifically implemented that analysis. One approach to do it would involve the following:

First predict subspot level expression using enhanceFeatures(). I suggest just doing the top HVGs rather than all genes to save compute time. Then use a deconvolution method that works on the logcounts. I believe we used CIBERSORT in the manuscript but you can try another more recent method.