Fig. 1
From: Causal models and prediction in cell line perturbation experiments
a Overview of perturbation data with a leave-one-drug-out (LODO) training / test set split. Drug 12 is never used in training. b Causal graphical model for subset of drugs and responses (proteins and phenotypes). Drugs are exogenous variables with known targets, e.g. it is assumed known that drug 12 directly effects protein \(X_3\). Black dashed lines represent hidden confounding. For example unmeasured cell cycle may effect responses \(X_{10}\) and \(X_8\)