Chemistry Reference
In-Depth Information
to bind to BACE in a similar manner to
6
using crystallography.
[
40
]
We analyzed this com-
pound and analog
8
. In both cases, the binding affinity decreased in a manner that roughly
correlated with the decreasing molecular weight, resulting in overall similar ligand effi-
ciency among all three compounds. [The typical reproducibility of most biological assays
(2-3-fold for IC
50
) translates to a reliability for ligand efficiency values of about
10% of
the stated value. This is adequate for the assessment of overall trends, although individual
small changes should not be over-interpreted.] This suggests that each atom is contrib-
uting a small amount of binding energy and the result is that to achieve high binding, a
fairly high molecular weight would be necessary. Overall, this would present additional
challenges and risk to the successful delivery of a CNS-penetrant drug from this structural
class.
It is noteworthy that the ligand efficiency remains low and does not change very much
among these three examples, despite an almost 10
5
-fold improvement in affinity.As reported
by Hajduk,
[
41
]
it has been found in several cases that the ligand efficiency remains relatively
constant within structural families. Hajduk started with a set of high-affinity drug-like
compounds across several projects and did a retrospective analysis to compare the properties
of those final compounds with iteratively truncated analogs. It was found that the binding
affinity remained remarkably linear with molecular weight throughout the 'deconstruction'
process in each family of ligands. This finding suggests that once a chemical lead has been
identified, it is possible to estimate the molecular weight of a final compound that will have
the desired affinity. Consequently, ligand efficiency should be valuable in helping to rank
and prioritize hits among different structural classes and affinity ranges.
Using NMR screening against BACE at 5mM ligand concentration, we identified numer-
ous hits with affinities ranging between 1 M and 5 mM. Among those hits,
1
had nearly
the weakest affinity (IC
50
2500 M), but the highest ligand efficiency (0.32). Despite the
weak affinity, we chose this as a lead for optimization according to the rationale above.
±
11.6 Optimization of the Isocytosine Series
As described above, screening of compounds that were structurally related to the initial
hit compound
1
led to the identification of
5
. Despite its weak binding affinity, compound
5
could be co-crystallized with BACE. Although the resolution was low, the structural
information was used with molecular modeling to generate the binding model shown in
Figure 11.9. According to this model, the isocytosine nitrogens at positions N1 and C2
interacted directly with the BACE catalytic aspartates and the phenyl ring was directed
toward the largely open S
3
pocket.
To fill the S
2
pocket better and improve the CNS drug-like potential (by removing a hydro-
gen bond donor), we methylated the N3 nitrogen. This afforded compound
9
(Figure 11.10),
which had threefold improved affinity.
[
42
]
Next, we modified the aryl region to expand
further into the S
3
pocket by replacing the phenyl with indole. Addition of this polar
functionality was also desirable in order to improve solubility. This led to compound
10
.
Combining the
N
-methylation and the indole replacement led to
11
, which had a 35-fold
improvement in affinity relative to
5
and slightly improved ligand efficiency. Further survey
of aryl analogs
12
-
14
(Table 11.1) led to the identification of
14
as the first inhibitor in this
