Prediction of Protein-Protein Binding Interfaces - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 108

in local energy minima (Lyskov and Gray 2008 ; Gray et al. 2003 ) . Results are
ranked on the basis of residue-residue contacts and clashes, as well as the properties
of pairwise residue-environment and residue-residue interactions derived from pre-
viously registered datasets (Lyskov and Gray 2008 ) . In the re fi nement phase the
goal is to find a structure with the lowest possible free energy. Energy calculations
acknowledge van der Waals forces (Gray et al. 2003 ) , orientation-dependent hydrogen
bonding (Kortemme et al. 2003 ), implicit Gaussian solvation (Lazaridis and Karplus
2000 ), side-chain rotamer probabilities (Dunbrack and Cohen 1997 ) and electro-
static potentials (Gray et al. 2003 ) .
Input datasets consist of subunit (monomer) structures expressed in the PDB
format. It is also necessary to determine a suitable starting structure, free from steric
clashes. Based on the results of the CAPRI challenge, approximately 1,000 struc-
tures are generated for each pair of interacting proteins. The program then deter-
mines the relationship between free energy and RMS-D values for the final structure
and for the initial conformation. Success is defined as the ability to locate an area in
the RMS-D/free energy space where both parameters are suitably low. The best
result (lowest values of both parameters) is treated as the correct final structure.
RosettaDock is therefore a multistart, multiscale Monte Carlo-based modeling
package. The end result is highly dependent on the initial structure, which - as
already mentioned - should be provided by the user, preferably on the basis of
experimental data (e.g. site-directed mutagenesis). In addition, RosettaDock is
capable of exploiting structures generated by other global search-oriented software
packages, such as ClusPro ( http://cluspro.bu.edu/login.php ) (Kozakov et al. 2010 ) ,
GRAMM-X ( http://vakser.bioinformatics.ku.edu/resources/gramm/ grammx)
(Tovchigrechko and Vakser 2006 ) , HEX ( http://hex.loria.fr/ ) (Macindoe et al. 2010 ) ,
PatchDock ( http://bioinfo3d.cs.tau.ac.il/PatchDock/ ) (Schneidman-Duhovny et al.
2005 ) and SymmDock ( http://bioinfo3d.cs.tau.ac.il/SymmDock/ ) (Schneidman-
Duhovny et al. 2005 ) .
RosettaDock has been notably successful in blind-prediction studies within the
CAPRI challenge (Gray et al. 2003 ; Lensink et al. 2007 ) .

6.2.4 ZDOCK ( http://zdock.bu.edu/ )

ZDOCK is a rigid-body simulation toolkit. The docking procedure involves geomet-
ric alignment of the surfaces of two molecules, treated as potential intermolecular
complexation sites. The “target” molecule is rigid, while the complementary mol-
ecule is rotated around its surface using a grid with a predefined density. This
systematic search bases on Fast Fourier Transform algorithms (Wiehe et al. 2008 ;
Mintseris et al. 2005, 2007 ; Chen and Weng 2002, 2003 ; Mintseris and Weng 2003 ;
Li et al. 2003 ; Chen et al. 2003a, b, c ; Pierce et al. 2005, 2007 ; Pierce and Weng
2007, 2008 ). ZDOCK applies a scoring function which acknowledges shape com-
plementarity, electrostatics and pairwise atomic potentials determined on the basis
of known protein complexes. A separate scoring package (RDOCK) is provided,