Chemistry Reference
In-Depth Information
a rapidly increasing number of studies investigating what makes a good starting point in
fragment-based drug discovery and how to formulate libraries to maximize success in the
screening process.
The exact nature of a fragment library is very much dependent on the screening pro-
tocol; however, the methods employed to construct fragment libraries borrow heavily from
experiences in drug-like classification. A simple analogy is that of the Astex 'rule-of-
three'
[
14
]
as compared with Lipinski
et al
.'s 'rule-of-five'
[
15
]
. Here the same physicochemical
descriptors are used but compounds are limited to amolecular weight (MW) of 300Da, three
or less hydrogen bond donors and acceptors and a calculated log
P
of
3. TheAstex library
is comprised of several hundred small organic ring systems, is a mixture of target-specific
and general-purpose fragments and is used to probe a target site using high-throughput
X-ray crystallographic screening.
The application of a filter such as the 'rule-of-three' and subsequent identification of
target-specific compounds are in general the penultimate steps in the virtual selection of
fragments from a significantly larger chemical space. Additional steps common to most,
if not all, selection criteria for fragment libraries, particularly if starting from commercial
vendor space, include the removal of undesirable chemical functionality, elimination of
poorly soluble compounds, a selection based on synthetic tractability and consideration
of scaffold diversity. It is also probably safe to say that the final step in most virtual
screening campaigns involves scientists eyeballing compounds, predominantly to make
the final selection but also to ensure the baby is not being thrown out with the bath water.
The virtual screening process is commutative with respect to the final result, excepting
the manual selection; however, the order of operation will impact on efficiency. 1D and
2D filters, such as the exclusion of undesirable chemical functionality using substructure
searching, can eliminate a high percentage of compounds,
[
16
]
thus reducing the resources
needed for computationally more expensive procedures such as pharmacophore searching
and high-throughput docking.
SGX
[
17
]
outline a series of such criteria in the selection of a
≤
1000 member diverse
fragment library that includes filters for MW,
C
log
P
, compound complexity, exclusion of
undesirable chemical functionality, solubility, ring system diversity, synthetic accessibility
and, interestingly, a selection based on bromination. All the compounds in the library have
≤
∼
1 ring and at least two synthetic handles. Most of the compounds
obey the 'rule-of-three' and, importantly for screening using X-ray crystallography, 60%
have high solubility. Whereas Astex have target-specific fragments, SGX make no such
distinction on the basis that hit rates are in general higher in fragment screening and a
library of 1000 diverse fragments is deemed large enough to yield sufficient chemical
matter to initiate a discovery programme. The size of a fragment library, the complexity of
the molecules within the library and the optimisability are all characteristics important to
fragment screening and are discussed in more detail below.
Approximately half of the compounds in the SGX library contain bromine, a feature
included to enhance synthetic elaboration but also aid in the detection and validation of
crystallographic screening data. Compounds in the original 'SHAPES' library developed
at Vertex,
[
18
]
in addition to many of the aforementioned filters, had to yield simple
1
H
NMR spectra and contain at least two protons within5Åofoneanother, both aids to the
screening of mixtures using nuclear magnetic resonance (NMR) techniques. The library
stemmed from previous work investigating the properties of known drugs,
[
19, 20
]
where
16 heavy atoms,
≥
