Fig. 7From: Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitorsRelationships between the interval distribution of AUC values in the test sets and the corresponding interval of different compound quantities in the training sets of GCN (A), GAT (B), MPNN (C), Attentive FP (D), Chemprop (E), and FP-GNN (F) modelsBack to article page