Figure 14

Case Study of Pyruvate Phosphate Dikinase. In this case study, we demonstrate how B-cubed scoring may be used to determine the parameter settings for rigidity analysis. (a) shows the RigidFinder decompositions of PPDK, which was validated against literature-annotated functional domains [6]. The KINARI v1.0 decompositions of the open (2r82) and closed (1kc7), shown in (b) and (c), have B-cubed scores of 0.66 and 0.45 respectively. By varying the H-bond energy cutoff (method 2 in Table 3), decompositions with higher B-cubed scores for 1kc7 could be generated. The KINARI decomposition for 2r82 was optimal at cutoff energy 0, meaning all H-bonds were included. For 1kc7, the maximum score, 0.66, was attained at -1.5 kcal/mol when excluding weak H-bonds from the modeling (d). By using a bar to model weak H-bonds, a slightly better score (0.67, cutoff -2.75 kcal/mol) was achieved (e). The B-cubed score plots for the two conformations, using method 7, are shown in (f). As the cutoff is varied, the precision and recall are monotonically increasing and decreasing, shown in (g). An optimal B-cubed score is achieved when the F1-values combining the precision and recall is optimized.