**2.3. Popular docking programs**

One of the most important and useful areas of application of molecular modelling is the approach of docking a protein onto a second molecule, typically a small ligand. This is of interest because it models the possible interactions between the protein and the ligand in the formation of a biologically important protein-ligand complex. To perform a computational docking, experimental or model 3D structures of both the protein and ligand molecules are required together with the charge distribution for each molecule.

Using Molecular Modelling to Study Interactions Between Molecules with Biological Activity 191

ΔGbind = ΔGinteract – ΔGsolv,L - ΔGsolv,R (1)

for a receptor (ΔGbind) depends on the interaction free energy of the two molecules relative

where ΔGinteract is the interaction free energy of the complex, ΔGsolv,L is the free energy of desolvating the ligand, and ΔGsolv,R is the free energy of occluding the receptor site from the solvent. Various methods have been proposed to evaluate or estimate these terms. The problem is difficult because the energy of each component on the right hand side of

An accurate way to calculate relative binding energies is with free-energy perturbation techniques, although they are usually restricted to calculating the differential binding of similar compounds and require extensive computation, making it impractical as an initial screen, but quite useful sometimes [Buch et al., 2011; Reddy & Erion, 2007]. Several authors have described force fields that consider the bound and solvated states [see for example Chen et al., 2008; Moon & Howe, 1991], successfully predicting new ligands and also the

When calculating interactions in congeneric series, the cost in electrostatic free energy of desolvating both the enzyme binding site and the burial part of the ligand (ΔGdesolv) is roughly constant within the series. This is particularly true when the calculation is done partitioning the electrostatic free energy contributions into a van der Waals term from the molecular mechanics force field, and an electrostatic contribution computed using a continuum method [Checa et al., 1997]. For that reason, it has been proposed to neglect

The binding energy between ligand and receptor is approximated to the interaction enthalpy calculated by means of empirical energy functions that represent van der Waals repulsion, dispersion interactions by a Lennard-Jones term, and electrostatic interactions in the form of a Coulomb term that uses atom-centred point charges [Ajay & Murcko, 1995]. In most cases, these calculations of molecular mechanics are performed on a structure that is taken to represent the ensemble average of each complex. Entropy contributions are usually ignored although solvation terms are sometimes added to the scoring function by calculating changes in buried nonpolar surface area [Viswanadhan et al., 1999] or differences in the ease of desolvation of both the ligand and the binding site upon complex formation [Checa et al., 1997]. Molecular mechanics-based QSAR studies on ligand-receptor complexes can benefit greatly from proper incorporation of solvation effects into a COMBINE framework based on residue-based interaction energy decomposition [Pérez et

The relevance of solvation in modulating the biological activity of drugs is well known [Orozco & Luque, 2000]. In the last years, theoretical methods have been developed to calculate fragment contributions to the solvation free energy, particularly in the framework of quantum mechanical (QM) continuum solvation methods [Klamt et al., 2009]. Thus, fractional methods based on GB/SA methods have been developed [Cramer & Truhlar,

Equation 1 is large while the difference between them is small.

structures of ligand-receptor complexes [Wilson et al., 1991].

to their free energy in solution:

ΔGdesolv in earlier studies.

al., 1998].

There are several software programs that are available for carrying out docking calculations, only some of them will be considered here. The DOCK program suite [Kuntz, 1992] is one of the best known. First of all, a set of overlapping spheres are used in the program to construct a negative image of a specified site on the protein or another macromolecule, and the negative image is then matched against structures of potential ligands. Matches can be scored in this program by the quality of the geometric fit, as well as by the molecular mechanics interaction energy [Meng et al., 1992] and can lead to protein-binding ligands that have micromolecular levels of binding affinity [Kuntz et al., 1994]. It has also been used for modelling protein-protein docking [Shoichet & Kuntz, 1996].

The program GRID [Goodford, 1985] identifies likely protein binding sites for ligands [Reynolds et al., 1989; Cruciani & Goodford, 1994] using a 3D grid around the protein.

The program AutoDock developed by Morris et al.; [http://www.scripps.edu/pub/olsonweb/doc/autodock/] uses a grid-based scheme for energies of individual atoms, allowing a quick computation of the interaction energy of the protein-ligand complex as the interaction between the ligand and the grid.

GLIDE software [Friesner et al., 2004, 2006; Halgren et al., 2004] also uses a grid-based scheme to represent the shape and properties of the receptor and then uses a systematic search algorithm to produce a set of initial conformations, using a OPLS-AA force field for ligand minimisation in the field of the receptor.

SURFLEX [Jain, 2003, 2007] is a fully automatic flexible molecular docking algorithm that presents results evaluated for reliability and accuracy in comparison with crystallographic experimental results on 81 protein/ligand pairs of substantial structural diversity.

In a recent study, comparison of seven popular docking programs [Plewczynski et al., 2011] clearly showed that the ligand binding conformation could be identified in most cases by using the existing software. Yet, there is still the lack of universal scoring function for all types of molecules and protein families. One can always hope that incremental improvements in current techniques will gradually lead to major advances in this field.
