Optimal learning of Markov k-tree topology

Single-cell systems pharmacology identifies development-driven drug response and combination therapy in B cell acute lymphoblastic leukemia

“…To identify targetable signaling circuits in asparaginase-resistant B-ALL cells, we extracted the pre-pro-B cell gene-gene interactome inferred by SJARACNe. We first filtered pre-pro-B cell-type-specific driver genes for the druggable genes in DGIdb, 59 and then performed a maximum spanning k-tree-based analysis 60 of the subnetwork of the top 80 druggable pre-pro-B cell-type-specific drivers. We found BCL2 to be a key driver gene that was upregulated at this stage along with known pre-pro-B markers, including FLT3 , CTSA , and CTSB ( Figure 5A ).…”

“…Pre-pro-B genes with the top upregulated protein activities were selected to further filter against the targetable genes in DGIdb to form a SJARACNe subnetwork of 80 druggable pre-pro-B cell-type-specific drivers. Due to the high connectivity of the subnetwork, we formatted the problem of finding the key driver genes as a graph sparsification problem, where a minimum spanning tree 60 was constructed to optimize the total edge betweenness of the subnetwork. For graphical representation, the spanning tree was visualized using Cytoscape 85 with edge color indicating the Spearman correlation of a pair of connecting genes, and the node size indicating the betweenness of a node in the subnetwork.…”