Chemistry Reference
In-Depth Information
optimization process. In the case of the smaller, weaker hit, binding is likely due to fewer,
but stronger, individual interactions. As a result, optimization may effectively be achieved
by maintaining these and adding (or improving) additional interactions.
[
22
]
Addition of
such interaction points (and concomitant increases in molecular weight) can be better
accommodated while retaining drug-like properties if the initial hit has lower molecular
weight.Additional information on fragment-based methods can be found among the several
recent and excellent reviews.
[
18, 22 26
]
Among the principal disadvantages of fragment-based lead generation methods are the
following. First, by starting from weaker affinity hits (100 M to 5 mM), considerable
structural modifications are likely to be needed during optimization to reach typical drug
affinity (
<
100 nM). Such optimization can be greatly accelerated with structural knowledge
(e.g. NMR or crystallography). A second disadvantage is that typical biological screens
may not work reliably at very high ligand concentrations and thus cannot detect weak
binders. This frequently necessitates the development of alternative analytical methods. As
discussed below, we employed three screening and analysis methods: NMR spectroscopy,
surface plasmon resonance (SPR) spectroscopy and crystallography.
11.3
Identification of Fragment-based Leads by NMR Screening
Whereas typical HTS hits have affinity
<
1 M, the smaller and simpler hits emerging
from fragment screening typically have affinities in the range 100 M-5 mM. Because
of the high ligand concentration, we could not use conventional enzymatic assays for
screening. NMR spectroscopy is particularly well suited for the detection of such starting
points because of its sensitivity, throughput and experimental robustness.
[
27
]
We designed
a screening set of approximately 5000 compounds (150-350 Da ) using a combination
of diversity-based and target-based considerations. The diversity-based subset included a
range of compound classes containing features that are associated with good CNS drug-
like features. The target-based subset was influenced by knowledge of the BACE binding
site. For example, the BACE active site contains two aspartic acids and a nearby lipo-
philic region.
[
17
]
Thus our screening library was biased toward lipophilic, weakly basic
amines and lipophilic alcohols since both are precedented as binders to related aspartic
acid proteases.
[
15
]
Ligand binding can be detected by a number of NMR spectroscopic methods.
[
28, 29
]
We employed
[
30
]
the water-LOGSY
[
31, 32
]
approach; this is a 1D experiment that detects
magnetization changes to a small molecule if it binds to its macromolecular target. Mul-
tiple compounds can be screened at the same time, the experiment uses relatively small
amounts of enzyme and signal assignment is not necessary. At high ligand concentrations
the false positive rate can be substantial due to non-selective binding or ligand aggrega-
tion. Therefore, it is very important to have a means to confirm binding selectivity. Unlike
2D methods, it is not possible to assess directly the ligand binding mode because the bio-
logical target is essentially invisible. However, if a high-affinity ligand (for example a
peptide) is available it can be used for mechanistic validation. Once binding is detected
for a compound in the screening set, the high-affinity ligand can be added. Displacement
of the screening compound by that ligand can be observed in the NMR experiment. This
