Chemistry Reference
In-Depth Information
with the AutoDockFilter (ADF) program. ADF calculates a pseudodistance (
d
CSP
)
based on the magnitude of the CSPs and compares it to the shortest distance (
d
S
)
between any atom in the residue that incurred the CSP with any atom in the docked
ligand pose. A violation energy is attributed to each protein residue that is further
from the docked ligand pose then predicted by the CSP pseudodistance. The sum of
these violation energies generates an overall NMR energy (
E
NMR
) for the docked
ligand conformer:
k
X
n
i¼
1
ðD
Dist
Þ
d
CSP
d
S
d
CSP
<
d
S
2
E
NMR
¼
D
Dist
¼
:
(9)
0
d
S
d
CSP
The conformer with the lowest NMR energy corresponds to the best protein-
ligand co-structure based on a consistency with the experimental CSPs. The NMR
energy also provides a qualitative way to evaluate the reliability of the co-structure,
with high NMR energies correlating to unreliable co-structures (Fig.
8
).
NMRScore [
159
] is very similar to ADF. NMRScore uses poses generated by
AutoDock and seven other docking programs. CSPs are calculated for each pose
using DivCon, where a CSP RMSD is determined between the calculated and
experimental CSPs. The best pose corresponds to the conformer with the lowest
CSP RMSD. The McCoy and Wyss method [
157
] also uses simulated chemical
shift changes. But, unlike the NMRScore approach, the docked ligand is replaced
by a number of randomly placed amino-acid probes within the ligand binding site.
Proton chemical shifts, primarily from ring-current effects, are calculated for the
protein with and without the docked amino-acid probes. The proton chemical shifts
are calculated using the SHIFTS program [
161
], where CSPs are determined based
on the difference between the two sets of calculated proton chemical shifts.
The best pose for the amino-acid probe is chosen based on a minimal difference
between the experimental and calculated proton CSPs. The ligand is then docked to
the protein by aligning the ligand with the amino-acid probes.
Instead of simulated chemical shifts, the HADDOCK [
44
] and LIGDOCK [
158
]
programs use CSPs to define ambiguous interaction restraints (AIRs) [
162
]. AIRs
are an intermolecular distance restraint between all atoms of the residue with the
CSP and all atoms of the ligand. Importantly, other experimental information (STDs,
mutational data, etc.) can also be used to define AIRS. HADDOCK and LIGDOCK
employ a three-tiered approach to refining the protein-ligand complex. First, the
ligand is docked to a rigid protein structure. Next, the protein-ligand structure is
refined with simulated annealing in torsional space [
163
]. Finally, the structure is
optimized with explicit solvent to remove any remaining structural problems.
HADDOCK and LIGDOCK are particularly beneficial since the protein-ligand
co-structure is directly refined against the experimental CSPs. The methods do
suffer from long computation times and potential difficulties with proper parameter-
ization of the ligand. HADDOCK was initially developed to dock protein-protein
interactions and was later modified to accommodate ligands, whereas LIGDOCK
was specifically designed to generate protein-ligand co-structures.
