Figure 3From: Improving protein-ligand binding site prediction accuracy by classification of inner pocket points using local featuresRescoring Fpocket predictions on CHEN11 dataset. Success rates of Fpocket compared with results rescored by PRANK on CHEN11 dataset considering Top-n, Top-(n+2) and all pockets (total coverage). Identification success is measured by DCA criterion for the range of integer cutoff distances. Displayed results for rescored pockets are averaged from ten independent 5-fold cross-validation runs.Back to article page