Chemistry Reference
In-Depth Information
Table 5.1
Statistics on pose generation with
Poser
Target
Box size (Å)
No. of
Poses
Poses
Best
Best
conformers
a
evaluated
b
saved
c
conformer
pose
(millions)
found
d
found
e
FKBP-12
11
×
12
×
16
5
>
629
45 693
0.51
0.64
PDF
12
×
14
×
10
3
>
485
50 792
0.52
0.96
CDK2 (1)
11
×
12
×
16
5
>
632
157 380
0.30
0.45
CDK2 (2)
16
×
15
×
12
13
>
808
10 579
0.49
0.69
Bcl-x
L
19
×
16
×
17
1
>
191
9 556
0.25
0.71
a
Number of conformers generated by Omega and used as input into
Poser
run.
b
Poses were generated with a grid spacing of 1 Å, a rotational sampling of 5
◦
and a radii scaling of 0.9. No steric clashes
between the target and ligand were allowed. For Bcl-x
L
,a10
◦
rotational sampling was used.
c
Total number of poses that fit into the binding site.
d
The RMSD of the Omega conformer for the ligand with the lowest RMSD to the target conformer. Compounds were
aligned for best fit before calculating RMSD.
e
The RMSD of the
Poser
-generated binding pose with the lowest RMSD to the target binding pose. RMSDs were calculated
with ligand molecules in the context of protein, that is, no alignment was performed.
typically led to three or more low-energy conformers of each compound. The conformer
with the lowest RMSD to the target conformer was generally
∼
0.5 Å, reflecting that the
experimentally determined conformation of a compound is often different from the com-
putationally defined local and global energy minima that exist in the absence of the target
protein. This RMSD value sets the limit for what we could expect our pose sampling with
Poser
to achieve. For each of the NOE matching cases with fragments described below,
poses were generated with a 'rotational sampling' of 5º. This led to hundreds of millions of
poses being evaluated for each input conformer of the compound and tens of thousands of
poses being saved for later evaluation by NOEmatching. In general, the best pose generated
using
Poser
was within1Åofthetargetpose and, as Table 5.1 indicates, was often closer.
5.5 Applications to Fragment-like Compounds
Although the boundaries between drug-like, lead-like and fragment-like compounds can be
somewhat fuzzy, fragment-like compounds are generally smaller and less functionalized
than lead-like/drug-like compounds. This distinguishing feature carries with it a significant
consequence; namely, the binding of a fragment to its receptor is often much more difficult
to characterize structurally than that of a lead-like molecule. There are several reasons for
this. First, the binding affinity of fragments tends to be weaker than what one might typic-
ally observe for a more complex molecule, leading to the requirement of higher compound
concentrations to attain receptor saturation. Second, the lack of structural complexity of
fragments provides fewer distinguishing features that can be used to guide structural refine-
ment. Third, binding of a fragment to its receptor may not be limited to a single binding
pose. The reduced potency and structural simplicity of fragments presents challenges for
both X-ray and NMR structural determinations.
In applying NOE matching to fragment pose determination, we were very concerned
that the fragments might not be large enough to contact enough of the binding pocket
(i.e. multiple residue types) to give rise to sufficient information content in the observed
NOE patterns (a requirement of NOE matching) to permit discrimination between true
