Chemistry Reference
In-Depth Information
the test cases is look at the COST of the poses ranked from lowest COST to highest COST,
as has been done in Figure 5.12. In this figure are indicated both the RMSD of the lowest
COST pose to the target pose and the RMSD and rank of the pose in the ensemble with the
lowest RMSD to the target pose.
(a)
(b)
5000
7000
6000
4000
5000
3000
4000
3000
2000
0.69Å (14)
0.35Å (2)
2000
1000
1000
0.48Å
0.74Å
0
0
0
2000
4000
6000
8000
10
000
0
2000
4000
6000
8000
10
000
POSE #
POSE #
(c)
(d)
6000
8500
5000
7500
4000
6500
5500
3000
0.96Å (12)
4500
2000
0.71Å
1.12Å
3500
1000
0
200
400
POSE #
600
800
1000
0
2000
4000
6000
8000
10
000
POSE #
Figure 5.12
The poses for the NOE matching runs are ordered based on COST and plotted
from lowest COST to highest COST. In each panel, the RMSD of the lowest COST pose to
the target pose is listed. Also indicated are the RMSD and rank of the pose in the ensemble
with the lowest RMSD to the target pose. (A) CDK2/
4
complex using simulated NOE data;
(B) FKBP-12/
5
complex using simulated NOE data; (C) PDF/
6
using experimental NOEs and
predicted assignments set to SHIFTX values; (D) Bcl-x
L
/
7
using experimental NOEs.
Figure 5.12 clearly indicates that in all cases, for both simulated and experimental data,
the correct pose is within the set of poses having the lowest COST, as expected. One only
needs to ensure that a correct pose is a member of the ensemble of trial poses; a robust way
to achieve this is to do extensive, systematic sampling, as we have done with
Poser
. (As
discussed later in this chapter, highly flexible binding sites present the additional challenge
of adequately sampling the conformational space of the protein prior to using
Poser
.) In the
absence of very unusual protein chemical shifts, the correct pose (or a pose very similar to
it) will always be scored with low COST; therefore, one need not consider any of the poses
determined to be of higher COST by NOE matching, facilitating the analysis of the results.
Typically, we have found that any pose with a COSTmore than 50% greater than the lowest
COST pose can be discarded from subsequent analysis.As will be seen below, this becomes
important when evaluating hundreds of thousands or even millions of trial poses.
Even after discarding higher COSTposes, the number of poses to consider can still be very
large. The challenge becomes how to assess whether these poses are similar or distinct. To
