Fig. 4

IFP, GRIM and ROCS performance in pose prediction. The proportion of correct predictions is based on the RMSD between the predicted and native poses of fragment, considering five threshold values. Proportions are calculated by considering a single pose within the ensemble generated for a complex, as follows: First denotes the top scored pose; Best denotes the closest to the native pose; D, F and DF denote the poses selected by comparison to, respectively, reference drug-like ligands, reference fragments and both. a All protein sites. b The 35 protein sites with frequent scoring issues (scenario 3)