A theoretical feature of a large-scale CMap is the ability to determine mechanism of action (MOA) of a small molecule, based simply on similarity to profiles of genetic perturbagens or compounds of known function. We first determined whether known MOAs could be recovered by the CMap. This is challenging, however, because the definitive list of protein targets (and their associated pathways) of small molecule drugs is unknown. Nevertheless, we used multiple resources to associate 1,902 compounds to protein targets and associated pathway members profiled in the CMap. This led to 58,820 expected relationships that could plausibly be recovered in the CMap (see STAR Methods) (Corsello et al., 2017). We then sought to recover those relationships from among the approximately 160 million pairwise relationships (connections) that could be assessed across CMap.
