Chemistry Reference
In-Depth Information
split into fragments. The most rigid fragments are often used as the core or anchor
and are docked first into the receptor binding pocket. The remaining fragments are
incrementally added back onto the core fragment, where each addition is systemat-
ically rotated to evaluate the most optimal conformation. Thus, incremental con-
struction drastically reduces the number of possible conformations that need to be
searched in order to identify the optimal pose.
Another systematic approach uses rigid docking in combination with a pre-
defined library of ligand conformations, which is implemented in OMEGA [
94
],
FLOG [
95
], Glide [
43
], and the TrixX Conformer Generator [
96
]. This technique
generates several low energy conformers for a ligand that are clustered by RMSD.
A representative conformer from each cluster is then docked into the receptor.
The approach is very fast because the docking process keeps the ligand rigid,
eliminating the need to spend computation time on searching torsional space.
A tradeoff for this increase in speed is a potential loss in accuracy, since the binding
potential for all possible conformers may not be explored. Conversely, a major
benefit of the technique is the fact that the library of structural conformers only
needs to be generated once. This is a significant savings in time for the pharma-
ceutical industry, where screening libraries may consist of millions of compounds.
Unlike systematic approaches that attempt to sample all possible ligand confor-
mations, stochastic searches explore conformational space by making random
torsional changes to a single ligand or a population of ligands. The structural
changes are then evaluated using a probability function. There are three types of
stochastic searches: Monte Carlo algorithms [
97
], genetic algorithms [
98
], and tabu
search algorithms [
99
]. The most basic stochastic method is the Monte Carlo
algorithm, which utilizes a Boltzmann probability function to determine whether
to accept a particular ligand pose:
exp
ð
E
1
E
0
Þ
P
;
(3)
K
B
T
where
P
is the probability the conformation is accepted,
E
0
and
E
1
are the ligand's
energy before and after the conformational change,
K
B
is the Boltzmann constant,
and
T
is the temperature. The simple scoring function used by the Monte Carlo
algorithms is more effective than molecular dynamics in avoiding local minima and
finding the global minimum. Alternatively, genetic algorithms utilize the theory of
evolution and natural selection to search ligand conformation space. In this case,
the conformations, orientations, and coordinates of a ligand are encoded into
variables representing a “genetic code.” A population of ligands with random
genetic codes is allowed to evolve using mutations, crossovers, and migrations.
The new population is evaluated using a fitness function that eliminates unfavorable
ligand poses. Eventually, a final population converges to ligands with the most
favorable “genes” or conformations (Fig.
3
). Tabu searches, like other stochastic
methods, randomly modify the conformation and coordinates of a ligand, score the
conformer, and then repeat the process for a new conformation. Tabu searches
