Chemistry Reference
In-Depth Information
energy with the lowest variance for a given set of forward and backward work values is produced by BAR and may
be one of the efficient nonequilibrium work estimators [33,71].
Another interesting set of methods use weights for the size and position of increments to improve the sampling of
the switches. Particular areas of a switch may be more or less prone to producing large work values. It is possible to
reduce the amount of work through the use of favorable system configurations or very small λ increments [33].
Configuration bias sampling uses the idea of sampling to bias the configurations, which are used to increment λ,
producing switches with a lower work distribution. Due to the possible presence of a nonsymmetric bias in Jarzynski
equality (JAR) calculations, it may become necessary to make a choice between JAR estimates in the forward and
backward directions [33, 36].
The harmonic oscillator (HO) systems for NE sampling considers the systems A and B, with the same number of
oscillating particles N but with differing Hamiltonians, in terms of x
i
the reference coordinate for particle
i
and ω
A
and ω
B
for the force constant of the two systems, which control the size of oscillations the particles will undergo as
well as the size of phase space each system explores. The HO model is very simple and many of its properties can be
calculated analytically, including free energy differences. This is very useful as the results from this protocol can be
compared to the correct answer rather than an answer found using an exhaustive free energy protocol [33].
DOCKING
Today, for drug discovery programs, simulation and computational modeling have become important components,
whereas for modeling of protein-ligand binding, a well established technique is molecular docking, with numerous
applications ranging from industrial drug research to pure fundamental studies. However, today's molecular docking
programs still fail to satisfy the expectations of researchers, despite two decades of significant investments in
developing docking softwares, leading to limited
in silico
ligand screening in the drug discovery process. For many
workers, the search for improving molecular docking is focused on the search for a better representation of
intermolecular interactions energies, scoring functions and docking algorithms that perform global optimizations on
multidimensional potential energy surfaces. Free energy calculations focus on the main forces contributing to
ligand-protein binding (hydrogen bonds, electrostatics, hydrophobic, van der Waals, etc) [77-483].
In general, the problem of finding the low-energy binding modes of a ligand is based on 'the lock and key
mechanism' within the active site of a known receptor and is what defines the molecular docking problem. A vast
literature is devoted to these methods. An important objective is to search large sets of compounds for new lead
structures from protein-ligand docking and drug optimization. However, the design space grows combinatorialy with
the number of degrees of freedom of the interacting molecules and all possible docking configurations cannot be
completed. As an example, for a protein receptor that occupies a volume of 60 Å
3
, with a translational resolution of
1 Å, rotational resolution of 20
0
in each axis, potential ligands and protein receptor with 35 and 3,500 atoms
respectively, it would be necessary to make on the order of ~ 10
14
pairwise nonbonding evaluations, in order to take
into consideration all possible docking configurations. Computational power thus becomes a limiting factor [443].
Numerous protein-ligand docking programs have been developed to efficiently discover lead compounds for a target
protein from large compound databases. Docking programs can be assessed if the pose or binding mode of a crystal
structure can be reproduced or whether the predicted score can be correlated with experimentally measured binding
affinity. In general docking algorithms are highly successful at generating good binding modes whereas scoring
functions are less successful at correctly identifying binding modes. In comparing docking programs, some of which
can generate conformations very close to crystal binding structures, it is generally not known which conformation
created by which program is close to the true structure. It is thus of interest to further improve scoring methods to
rerank conformations generated.
Although docking of small molecules to protein binding sites, i.e key structure-base drug design (SB) methods was
pioneered during the early 1980s, with the treatment of the protein and ligand as rigid bodies, various problems still
continue to afflict docking. Issues, such as flexibility in the receptor and ligand, covalent interactions, handling
