Page 107 - 76_02

VOL. 76 (2), 223-239, 2010 PREDICTION OF LIGAND BINDING ENERGY...

All the grid maps used to represent the protein in the rigid docking
were calculated by AutoGrid.

Finally, the empirical free energy function and the Lamarckian
genetic algorithm were used, applying a standard protocol with an
initial population of 150 randomly placed individuals, a maximum
number of 2.4 × 107 energy evaluations, a mutation rate of 0.02, a
crossover rate of 0.80, and an elitism value of 1.

The results were clustered according to a RMSD criterion and
were classified taking the predicted energy of binding into account.
In all cases the most favorable conformations were selected
according to the following criteria: best energy and best super-
imposition with the crystallographic ligand. Suitable conformations
were used in molecular dynamic studies for assessing the stability
and the energy of the complex.

2.3. Molecular dynamics

The complexes ligand-enzyme were evaluated in molecular
dynamics simulation. All the topological parameters for the enzyme
were created by GROMACS programs and the parameters of ligands
were built by the Dundee PRODRG Beta Server (8). The complexes
were solvated in a box of SPC/E water, neutralized with sodium ions
and then energy minimized by a two-steps protocol as described
before for the enzyme.

Following the minimization, a simulation of 300 ps at 298K and
1 atm with pressure coupling using Parrinello-Rahman method (9)
was performed using the leapfrog algorithm (isobaric-isotherm, NPT
assemble and periodic boundary conditions) with constraints in all
bonds using LINCS algorithm (10). Particle-Mesh-Ewald (PME)
summation was applied dealing with long-range electrostatics (11)
and a 10 amstrong cut-off for van der Waals interactions was used.
Energy and coordinates were recorded each picosecond to estimate
the energy contribution and the possible interactions between ligand
and the protein.

227