Chemistry Reference
In-Depth Information
(a)
(b)
9000
8000
3000
0
1
2
3
4
5
6
RMSD to Target
Figure 5.9
(A) COST versus the RMSD (Å) to the target pose for PDF/
6
. The predicted protein
chemical shifts set to the corresponding BMRB average values. Cross peaks from the experi-
mental 3D X-filtered NOESY spectrumwere used as input for NOE matching. (B) Superposition
of target pose and the minimum cost pose (dark gray) from (A).
NOE intensities. Due to the relatively large uncertainties associatedwith predicted chemical
shifts, there are often several
1
H
13
C groups within the binding pocket that yield predicted
chemical shifts that match the experimental chemical shifts within the defined tolerances.
NOE matching evaluations have been typically carried out with protein chemical shifts
set to those corresponding to BMRB average values. To begin to assess whether predicted
chemical shifts might improve the overall ranking of poses, we performed an initial NOE
matching evaluation of the PDF poses in which the protein chemical shifts were assigned
values predicted with the program SHIFTX.
[
28
]
SHIFTX is a computer program (developed
by Wishart and co-workers) which predicts
1
H,
13
C and
15
N chemical shifts using a hybrid
prediction approach that employs precalculated, empirically derived chemical shift hyper-
surfaces in combination with classical or semiempirical equations (for ring current, electric
field, hydrogen bonds and solvent effect). The hyper-surfaces in SHIFTX are generated
using a database of IUPAC-referenced protein chemical shifts (RefDB)
[
29
]
and correspond-
ing high-resolution (
<
2.1 Å) crystal structures. Although SHIFTX generally improves the
overall prediction of the chemical shifts relative to the BMRB values for residues that have
lowmobility in the target structure, we believe that it can be misleading to attempt to predict
dynamic residues using a single static structure as a template. For example, if an aromatic
side-chain in a binding pocket can adopt multiple conformations and the target pdb struc-
ture we have selected has the side-chain aromatic ring in an incorrect conformation (e.g.
rotated 90º around
χ
2
), then the SHIFTX-predicted chemical shifts of many nearby atoms
would move upfield (downfield), rather than downfield (upfield), due to reorientation of
the local magnetic field predicted for the ring currents. Such a case would adversely affect
the scoring in NOE matching. To minimize this possibility, we used the SHIFTX chemical
shift prediction only for those residues whose buried surface area was greater than 65% of
their accessible surface area;
[
30
]
for all other residues, which were deemed surface exposed
and thus potentially mobile, we used the average BMRB value for the chemical shift. Using
these chemical shifts, predicted as just described, we re-ran the NOE matching protocol on
the poses for the PDF case with experimental data. The results are shown in Figure 5.10.
