DEPECHE performance with real datasets with 32 or 35177 variables.

A-B) bi-variate t-distributed stochastic neighbor embedding (tSNE) representation of the 32-variate mass cytometry data. A) distribution of manually defined cell populations over the tSNE field. B) distribution of DEPECHE clusters over the tSNE field. C) Heatmap of the expression pattern at each cluster center. Light colors show that the marker is highly expressed at the cluster center, whereas dark colors shows the opposite. Grey color indicates that the variable in question does not contribute to defining the cluster. For Fig a-c all, 104184 cells have been clustered. D-E) tSNE representation of the 137-variate data subset that could efficiently distinguish the clusters in the 35177-variate single-cell transcriptome dataset. D) distribution of the cell types defined by index-sorting and manual gating on protein expression profiles shown over the tSNE field. E) distribution of DEPECHE clusters over the tSNE field. F) Violin plots illustrating the overlap between the original analysis by Björklund et al and the DEPECHE analysis. For each subplot, the left and right side illustrate the distribution of the transcripts defining the clusters, and all other transcripts, respectively. The y-axis shows the log10 of the p-values in the original analysis adjusted for multiple comparisons. For Fig D-F, all 648 cells have been clustered.