Figure 8

Results of B-cubed scoring evaluation on benchmark data set. We determined rigid cluster decompositions for each PDB file in the benchmark data set, using the 7 methods listed in Table 3. The data set comes from publications by two different software for performing rigid cluster decompositions. The MSU-FIRST portion of the data set consists of 4 proteins used to validate the MSU-FIRST software [1, 5]. The RigidFinder portion of the data set is categorized, from top to bottom, into large (greater than 500 residues), medium (between 200 and 500 residues) and small (fewer than 200 residues) proteins. Decomposition methods 1 through 7 are summarized in Table 3. For each method, the maximum B-cubed score and corresponding hydrogen bond and hydrophobic energy cutoffs (where relevant) are shown. See Figure 9 for barplots comparing each method against methods 2 and 3.