Chemistry Reference
In-Depth Information
Fig. 2 (a) An overlay of the 2D
1
H-
15
N HSQC spectra for the protein YndB titrated with
increasing amounts of chalcone. The perturbed residues can be used to identify a consensus
binding site. (b) NMR titration data for YndB bound to chalcone (
blue
), flavanone (
green
), flavone
(
purple
), and flavanol (
orange
). The magnitude of the chemical shift perturbation can be used to
calculate the dissociation constants for each compound. (Reprinted with permission from [
112
],
copyright 2010 by John Wiley and Sons)
[
79
]. Recent advances like the SOFAST-HMQC experiment [
80
,
81
] and the Fast-
HSQC experiment [
82
] have decreased the time and amount of protein necessary
for a target-based screen. Nevertheless, NMR ligand affinity screens are still very
resource intensive, requiring a significant amount of time and material. Also, since
any high-throughput screen produces a significant amount of negative data (most
ligands don't bind or inhibit a protein), a more efficient approach is to screen a
library of compounds with a higher probability of binding the protein target. In
effect, a virtual or
in silico
screen can be used to enrich a library with likely binders.
3 Molecular Docking
An accurate prediction of the interactions between two molecules requires an in-
depth understanding of the energetics that led to a stable biomolecular complex.
Unfortunately, a model that correctly accounts for all the factors involved in a
productive protein-ligand interaction is currently unknown. Further, the problem is
exponentially more complex than just modeling the specifics of a protein-ligand
interaction. A protein contains thousands of atoms that have specific interactions
with each other, with the solvent, and with other ions; in addition to the bound
ligand. Because of this complexity, computational efforts that attempt to model
protein-ligand interactions require significant amounts of processing power and
time. Many efforts that utilize molecular dynamics and distributed computing
[
83
,
84
] are generally limited to a detailed analysis of a
single
system. These
methods are generally not practical for the majority of researchers interested
in conducting a virtual screen of a library containing upwards of millions of
compounds. To make molecular docking computationally feasible and easily
accessible, many simplifications and trade-offs in the process are necessary.