A graph of gene expression programs identified by acNMF across eight neuroblastoma single-cell RNA-seq datasets, from humans, PDX, GEMM, and cell lines. Nodes represent a single program’s gene weights (i.e., loading scores), and node size is determined by the degree (i.e., number of connections). Edges connect similar nodes by a Jaccard similarity test. The colors highlight related communities of nodes. The black boxes highlight regions of interest, which are shown in detail in panels b – d . Note: unconnected nodes are not included in this figure but can be explored online at http://pscb.stjude.org .
