Neighbor Embedding (tSNE). Cells were then grouped by a shared nearest neighbor modularity optimization using FindClusters function with four resolution value. To compare transcriptional similarity of clusters, classification tree was constructed based on average expression of the set of variable genes. Internal nodes of the tree were assessed by the out of bag error (OOBE) from random forest classifier. Transcriptionallysimilar clusters were merged by selecting internal nodes with more than 0.2 OOBE. Furthermore, differentially-expressed genes were identified with all pairwise comparisons of clusters. Clusters with less than five differentially-expressed genes (> 1.5 fold and T test p-value < 0.05) were merged. Each cluster was characterized according to unique marker expression and its significant GO terms (Figure S3G and S4A). Clusters with the same annotation were plotted by the same color in Figure 5C.
