Construction of docking decoys:
It is important to generate sets of possible protein-protein candidate solutions including at least one near-native solution for testing protein-protein docking scoring functions. The protein-protein docking decoys are built by Gramm-X without refinement stage. The L_rmsd(RMSD of the backbone atoms of the ligand), I_rmsd(RMSD of the backbone atoms of the interface residues), fnat( number of native(correct) residue-residue contacts in the predicted complex divided by the number of contacts in the native complex) and fnon-nat(number of non-native (incorrect) residue-residue contacts in the predicted complex divided by the total number of contacts in the complex) of 50,000 matches from scan stage are computed according to the evaluation protocol(Raul Mendez et al, Proteins, 2005). A solution with L_rmsd<5.0Å is defined as near-native one. Top 100 nonnear-native structures with the highest surface complementarity scores are kept. Each decoy set consists of 100 nonnear-native matches and several near-native matches. The unbound structures are from benchmark set (Version 2).
|
