Fig. 6From: Structure-based, deep-learning models for protein-ligand binding affinity predictionInterpretability of representative PLBAP models. A Model-level interpretability of \(T_{ACNN}\) models. B Heatmaps showing the feature importance for a \(T_{IMC-CNN}\) model. C Heatmaps showing the importance of position-related features for a \(T_{Grid-CNN}\) model. D Importance of the node features for a \(T_{Graph-GCN}\) model. E Importance of the voxel channels for a \(T_{Grid-CNN}\) modelBack to article page