Chemistry Reference
In-Depth Information
the exact value is dampened since the magnitude of both effects will respond in an opposite and compensatory
fashion.
What is the best way to map the dielectric boundary between solute and solvent? What is physically the most
relevant surface to define this boundary which could be located at the solvent accessible van der Waals or the
molecular surfaces? Partitioning of the dielectric regions is sensitive to discretization which could be, in principle,
addressed by continuous dielectric functions. Using the van der Waals surface as a dielectric boundary results in
buried regions with high dielectrics which artificially stabilizes the electrostatic solvation energies which are
however, less severe than when the molecular surface is chosen to delineate the dielectric boundary The solvent
accessible surface avoids the problem of interstitial high dielectric cavities. The PB method can also take into
account the screening effects of counterions via the ionic strength. Dielectric constants, atomic charges and ionic
strengths are mapped on grids. Finer grids thus represent more realistically the system, improving the calculations.
Focusing techniques are necessary to retain acceptable boundary conditions with fine enough grids.
Apolar groups in water tend to self-associate in an effort to minimize their exposure to water which needs to be
accounted for in binding free energy calculations. Non polar contributions favoring association include van der
Waals and hydrophobic effects. Free energy of transfer of pure hydrocarbons from their neat liquid state to water is
one of the manifestations of the positive unfavorable hydrophobic effects. It is in general assumed that non-polar
free energies of binding of a ligand to a receptor should contain a term proportional to the amount of SASA buried
upon complexation which is, however, somewhat empirical and phenomenological, although a term treating
hydrophobic effects needs to be included in the nonpolar contribution. In general continuum models emphasize the
important role of non-polar effects as a main driver of association which can overcome the expected decrease in
configurational entropy with binding.
Although the entropic effects can oppose association or not, there are various contributions involved, whereas the
difficulties in dissecting these entropic contributions have made difficult the development and testing of theoretical
entropic contribution models. Nonetheless, the change in solvent entropy upon binding is implicitly included in the
MM-PBSA method. Regarding solute entropy a pragmatic approach assumes that for similar mass ligands binding
to the same protein yields similar entropic contributions which can be ignored for relative ranking of ligand
affinities. This can lead to calculated absolute free energies which are too favorable consistent with the neglect of
relatively large entropies of translation and rotation opposing binding. MD simulations can also be combined with
MM-PBSA whereas the solute entropy is calculated from normal mode analysis for several snapshots and then
averaged over the snapshots (MD/MM-PBSA) [461]. Overall, it seems that theoretical accounts for the various
solute entropic contributions remains challenging.
Continuum solvent models can be used with single point calculations (SP/MM-PBSA) or an ensemble of
configurations. Although the SP model is a simplification since free energies reflect all configuration states
populated by the system at a given temperature, this method can be conveniently applied to structural models
obtained from cruder scoring schemes including large number of docked compounds. A common strategy to
combine conformational sampling with continuum solvent approach is to first generate structural models and
explicit solvent with MD simulation which should produce a thermodynamic relevant ensemble of solute structures.
After removal of the explicit solvent the continuum solvent model can then be applied to snapshots of these solute
structures.
It is also possible to generate conformational ensembles directly with continuum solvation models. In realistic
virtual screening experiments with hundreds of thousands of compounds the docking poses can be initially generated
with fast empirical scoring functions. Efficient energy minimization algorithm can allow tens of thousands of
docking poses to be rescored. On the other hand, solute entropies may be added when a relatively small number of
poses is rescored. In general, the physics-based free energy scores can, in principle, be productively applied to the
reranking of large number of docking poses of diverse ligands in virtual screening.
The LIE approach, similar to continuum solvation models, only needs to consider the free and bound configurations
of the ligand, i.e only the end points needs to be simulated and no unphysical transformation from one ligand to
another is required. The method considers the van der Waals and electrostatics interaction energies between the
