(
A) Regulon specificity score for each cell type, with the top five highest scoring regulons shown in red. Specificity scores are based on Jensen-Shannon Divergence, a metric for comparing probability distributions (
Suo et al., 2018). pySCENIC provides standard functionality to compute the scores. (
B) UMAP representations of the top five mitral and tufted cell type-specific regulons (transcription factors) organized in rows by the cluster membership. Z-scored expression levels are used to highlight relative differences between cell types. Four TFs, namely Prrx1, Snai2, Bhlhe22, and Tcf7l2, are only very sparsely expressed (confirmed by the raw expression levels, data not shown). Their cell-type specificity has to be interpreted with caution. (
C) UMAP representations of the top three network hub genes (Pbx3, Bclaf1, and Taf1). Pbx3 is predominantly expressed in mitral cell type M1 and tufted cell type ET2. Bclaf1 and Taf1 are more broadly expressed across all M/T cell types. Nuclei are color-coded by the raw expression level of transcripts.