Results of scoring decoys from the dataset of Baker and coworkers [15].

[a]“Unbound perturbation” dataset. 54 targets were scored with 1000 decoys each. Scoring criteria were applied according to Baker and coworkers: “N10Å” is the percentage of complexes that have at least three top five decoys with rmsd <10 Å; “Best rmsd” is the percentage of complexes that have at least one top five decoy with rmsd <10 Å. Results from this work and the study by Baker and coworkers are shown.
[b]“Unbound docking” dataset. 54 targets were scored with 200 decoys each. To identify the top 10 solutions, the best scored decoys from the top 10 clusters were considered (see Materials & Methods). Scoring criteria were applied according to Baker and coworkers: “R5Å” (“R10Å”) is the percentage of complexes that have at least one solution <5 Å (<10 Å) in the top 10 decoys. “Best rmsd” is the percentage of complexes that have at least one solution with rmsd <10 Å in the top 10 decoys. Numbers not in parentheses refer to the 30 targets for which at least two “good” decoys are available; numbers in brackets refer to the other 24 targets. Results from this work and the study by Baker and coworkers are shown.
[c]Not determined.
[d]For reasons of comparison with the paper of Baker and coworkers both values are given although they are redundant.