Table 1 Performance of various strategies to match ligand-binding pockets assessed with the area under the ROC curve (AUC). Three types of algorithms are evaluated, direct methods based on the local alignment between a pair of pockets, indirect techniques employing structure-based virtual screening to detect chemically similar binding sites, and meta-predictors combining a direct and an indirect algorithm. Direct approaches are benchmarked on APoc and TOUGH-M1 datasets, whereas indirect methods and meta-predictors are assessed against the TOUGH-M1 dataset only

From: Comparative assessment of strategies to identify similar ligand-binding pockets in proteins