Chemistry Reference
In-Depth Information
enhances the value of the bioinformatics analysis. Protein structures often provide
insights into the molecular basis of the protein's biological function and its rela-
tionship to a particular disease. A protein structure also provides detailed infor-
mation on the sequence and structural characteristics that govern ligand binding
interactions. Building a drug discovery effort based on structural information
promises to help in the identification of novel therapeutic targets, in the discovery
of new lead compounds, and in the optimization of drug-like properties to improve
efficacy and safety. Currently, the drug discovery process within the pharmaceuti-
cal industry employs high-throughput screening (HTS) as the primary method for
identifying lead compounds. However, the high false positive rate [
9
-
12
] combined
with a significant cost in time and money has encouraged the development of
alternative methods to drive the drug discovery process [
13
,
14
].
Nuclear magnetic resonance (NMR) spectroscopy is uniquely qualified to assist
in making the drug discovery process more efficient [
15
,
16
]. NMR is useful for
several reasons: (1) it directly detects the interaction between the ligand and protein
using a variety of techniques, (2) samples are typically analyzed under native
conditions, (3) hundreds of samples can be analyzed per day, and (4) information
on the binding site and binding affinity can be readily obtained. These features
allow NMR to be an effective tool at multiple steps in the drug discovery pathway,
which includes verifying HTS and virtual screening hits [
15
,
17
-
19
], screening
fragment-based libraries [
15
,
20
-
22
], optimizing lead compounds [
15
,
17
,
23
,
24
],
evaluating ADME-toxicology [
25
-
27
], and identifying and validating therapeutic
targets [
28
,
29
]. Nevertheless, there are still intrinsic costs to maintaining an NMR
instrument, screening a compound library, and producing significant quantities of a
protein. One way to significantly reduce experimental costs is to utilize
in silico
methodologies to supplement the lead identification and optimization steps of the
drug discovery process [
30
].
Molecular docking is a computational tool that predicts the binding site location
and conformation of a compound when bound to a protein [
30
-
32
]. This approach
has been found to be fairly successful in redocking compounds into previously
solved protein-ligand co-structures [
33
], where more than 70% of the redocked
ligands reside within 2
˚
root mean squared deviation (RMSD) of the actual ligand
pose. During the prediction of protein-ligand co-structures, molecular docking
programs calculate a binding score that allows for the selection of the best
ligand pose. The binding score is typically based on a combination of geometric
and energetic functions (bond lengths, dihedral angles, van der Waals forces,
Lennard-Jones and electrostatic interactions, etc.) in conjunction with empirical
functions unique to each specific docking program [
34
-
39
]. A large variety of
docking programs are available that include AutoDock [
40
], DOCK [
41
], FlexX
[
42
], Glide [
43
], HADDOCK [
44
], and LUDI [
45
,
46
].
Binding energies are also routinely used to rank different ligands from a com-
pound library after being docked to a protein target. The virtual or
in silico
screening of a library composed of thousands of theoretical compounds can be
accomplished in a day with minimal cost [
47
-
49
]. Thus, a virtual screen can
significantly accelerate the hit
identification and optimization process while
