Extended Data Fig. 3: Sensitivity analysis of the PPI used for training PDGrapher.
From: Combinatorial prediction of therapeutic perturbations using causally inspired neural networks
Performance of sensitivity analyses evaluated by nDCG (a) and recalls (b-d) for datasets genetic-PPI-breast-MCF7 (left) and chemical-PPI-breast-MDAMB231 (right). The PPI used here is from STRING (string-db.org), which includes a confidence score for each edge. The edges are filtered by the 0.1, 0.2, 0.3, 0.4, and 0.5 quantiles of the confidence scores as cutoffs, resulting in five PPI networks with 625,818, 582,305, 516,683, 443,051, and 296,451 edges, respectively. The results of the percentage of accurately predicted samples are shown in Figure 5b.
