As might be expected, connections with support across multiple cell types tended to target core cellular processes (e.g., ribosomal function, proteasome complex), whereas compounds with reproducibly cell-type-selective patterns of connectivity tended to target more specialized mechanisms. For example, connectivity between multiple glucocorticoid receptor agonists was strongest in those cell types in which the glucocorticoid receptor was expressed (Figure S2D, upper panel). Connectivity between multiple PPARG agonists was greatest in HT29 and PC3, the two core cell lines with the highest baseline expression of PPARG (Figure S2D, lower panel). Similarly, the connection between androgen receptor (AR) knockdown and the AR antagonist nilutamide was strongest in the AR-expressing cell line VCAP (Figure S2E). We also note that the naturally occurring genetic diversity can be informative. For example, connections between genetic perturbation of the MAP kinase pathway and small molecule inhibitors of RAF or MEK kinases were strongest in the cell lines that harbor BRAF V600E kinase-activating mutations (Figure S2E).
