For future consideration - including coexpression heatmap for genes in reference dataset

Including a heatmap of the % cells coexpressing a given gene could help interpret the mixing matrix errors. For example, we have seen that unmixing struggles when highly expressed genes are coexpressed. If this is the case, the coexpression heatmap would look similar to the mixing matrix heatmap.

including coexpression heatmap for unmixed spot counts cellxgene table in current data could also be valuable as a point of comparison. The similarity in cellxgene expression patterns to reference dataset expression patterns is one of the most important validations of spot counting that we can do. Of course, many upstream steps play into this, but if we get similar results for that comparison, its a good sign that things are working well.