Fig. 4

Core strategies for the protein-ligand interaction hot spots. a Identify common/specific skeleton. The skeletons (blue circles) highly shared by most inhibitors of a target protein is called common skeleton, the other parts are called specific skeletons (red squares). In general, the common skeletons often form consensus interactions with key residues of the target protein and specific skeletons are often the substitution function groups (e.g., R1, R2, R3, X, and Y) occupying specific functional subsites for lead optimization to increase the potency. b Infer consensus features from preliminary QSAR models. Based on the selected features in preliminary QSAR models, we count the selected times (N _i ) for each feature i and calculate the average (μ) and standard deviation (σ) based on all features. The feature i is selected as consensus feature candidate if its N _i  ≥ (μ-σ)