Chemistry Reference
In-Depth Information
singletons. Superimposing the protein backbone atoms involved in secondary structure and
then calculating the RMSD for resultant positions of
7
in each of the structures, the pose
with the lowest COST has an RMSD of 2.98 Å with respect to the target pose. For the
ensemble of
Poser
trial poses, RMSDs to the target pose ranged from 1.08 to 7.93 Å. As
mentioned above, several residues in the binding pocket of apo Bcl-x
L
to adopt different
conformations upon
7
binding (Figure 5.17B).
The test cases above demonstrate the importance of coordinate selection for use during
the NOE matching runs. None of the test cases using alternative Bcl-x
L
protein coordinates
have satisfactorily reproduced the experimentally determined target pose for Bcl-x
L
/
7.
This
is because of significant backbone and side-chain rearrangement in the binding pocket. NOE
matching is predicated on the fact that somewhere in the ensemble of poses to be evaluated
is a pose that adequately reflects the 'correct' pose. We are evaluating how often significant
movement of side-chains occurs and how to identify in advance when such movements
will complicate NOE matching analysis.
It is interesting to note the significant difference in the number of poses accepted by
Poser
for the three test cases using different Bcl-x
L
starting coordinates. Whereas the
apo NMR structure accepted only 184 poses, the apo X-ray structure accepted
>
57 000
poses and the NMR structure taken from the BAK complex accepted
>
127 000 poses.
This is because the binding site in the BAK peptide-bound Bcl-x
L
structure is more open
than in the other structures. The more poses that can be sampled, the better the chance
one has of finding a pose close to the correct pose. Moreover, nine poses of
7
gener-
ated from the BAK-Bcl-x
L
structure had COSTs lower than any COSTs from poses using
the other structures of Bcl-x
L
. The best scoring pose obtained using the BAK-Bcl-x
L
structure has a COST of 1246, whereas the best scoring pose obtained using the 1YSG
structure had a COST of 1383. These results suggest that, of the four protein conform-
ations used, the BAK-Bcl-x
L
conformation may be the most similar to the true protein
conformation in the Bcl-x
L
/
7
complex. (Rigorous proof of this suggestion could only
be obtained by from an X-ray or high-resolution full NMR structure of the Bcl-x
L
/
7
complex.) These results indicate that it is best to use all available experimental protein
conformations for NOE matching. Additional, computationally derived protein conform-
ations may also be useful, provided that one can be confident that these conformations
are realistic. Because of the importance of sampling the correct protein conformation
in addition to the correct ligand conformation, location and orientation, we are in the
process of evaluating the use of protein ensembles as input target structures for NOE
matching.
The use of RMSD to evaluate the success of NOE matching has been used, in part, for
convenience. An alternative criterion for pose evaluation is to gauge how well the lowest
COST poses explain any known SAR data and to gauge how predictive they are. In other
words, if the lowest COST poses display the correct interactions with the binding site and if
they
correctly predict potential new interactions
, then they should explain any previously
known SAR data and should greatly facilitate structure-based lead optimization. After all,
the primary goal of this work is to be able to identify rapidly poses that provide information
to the chemists to guide the next round of synthesis. In all of the test cases for Bcl-x
L
/
7
, NOE
matchingwas successful at one level. It was able to distinguish between the two predominant
poses for the complex: the low COST poses shown in Figures 5.11B, 5.14B, 5.15B, 5.16B
and 5.17B and a second predominant pose in which
7
is flipped 180° in the binding pocket.
