Chemistry Reference
In-Depth Information
compounds as part of the virtual selection, to give a fragment library that has evolved over
several generations. The fingerprints, which are essentially identical in nature to Similog
keys noted above, encode the presence of pharmacophoric triangles comprised of standard
features such as hydrogen bond donors/acceptors, hydrophobes and aromatics. The more
pharmacophoric triangles in a molecule, the longer is the fingerprint, the length of which
is taken to represent the complexity of the molecule. Within the context of the library as a
whole, the fragment fingerprints are compared with fingerprints calculated for a drug-like
and a protein binding set of reference compounds. The comparison is made at increasing
limits in the distance between features in a pharmacophoric triangle, thus functioning as a
rough guide to the diversity of the library with increasing molecular size. It also allows the
selection of compounds that are novel to either of the reference sets. Novelty, as previously
mentioned, is another advantage of fragment-based discovery as it operates in chemical
space not normally identified by HTS. The physicochemical profile of the library is very
similar to those outlined above; most (99%) of the fragments in the
∼
1300 library have
MW
≤
300,
S
log
P
≤
3 and
≤
3 hydrogen bond donors, 90%
≤
3 hydrogen bond acceptors
and
60 Å
2
.
The SeeDs strategy uses NMR experiments to identify fragments that bind competit-
ively to a specific site of the target protein and then X-ray crystallography to determine
the exact pose of a fragment hit. SGX also advocates the combination of methodologies
using a high-concentration biochemical assay in conjunction with X-ray crystallography.
The combination of approaches helps to circumvent the individual shortcomings of each
method if used on an individual basis. For instance, false positives are inevitable when
screening at high concentrations in a biochemical assay. X-ray crystallography therefore
provides validation as to the mechanism of action of the compound in the assay. Knowledge
of the binding mode of fragments is also key to the rapid development of hits that are typ-
ically in the micromolar to millimolar potency range. Structure alone, however, tells us
nothing about the binding affinity, making it difficult to assess and rank the effectiveness
of each fragment hit for further modification. The order in which the methodologies are
applied greatly impacts on the size and nature of the fragment library. Fragment libraries
screened using X-ray crystallographic methods are at the lower limits, whereas libraries
for biochemical screening can be significantly larger in terms of both molecular size and
library numbers. Indeed, at the upper limits of what can be construed as fragment screen-
ing, Plexxikon
[
25
]
use a high-concentration biochemical assay to identify compounds from
a library of
∼
80%
≤
3 rotatable bonds and polar surface area (PSA)
≤
20 000 scaffolds. Scaffolds are noted to be smaller, less potent and less
complex than traditional HTS compounds but with MWs up to 350 Da they are obvi-
ously larger than the aforementioned library compounds. Hits from the high-concentration
biochemical assay are validated using X-ray crystallography, an approach that has found
favour in many companies, especially those familiar with crystallography as a tool in drug
development.
The definitions described above, although more restrictive than drug-like criteria, still
encompass a broad range of molecules, for example optimisation of a 1 mM inhibitor with
MW150 Da is a far better prospect than optimisinga1mMinhibitor withMW300 Da. Can
you quantify the quality of a fragment hit in terms of its potential for transformation into a
drug-likemolecule? Hadjuk
[
26
]
attempted to rationalise the selection of fragments for initiat-
ing discovery programmes through a retrospective analysis of the development of a number
∼
