Chemistry Reference
In-Depth Information
seen in Figure 9.2 to have an AlogP98
[
10
]
distribution lying between the ActiveSight and
Maybridge libraries.
Maybridge's fragment library, referred to as Ro3 since it strictly adheres to the rule
of three,
[
12
]
has been constructed to be highly diverse. The Maybridge compounds are
predominantly heterocycles that have been clustered for diversity while retaining react-
ive or 'linker-friendly' side-chains to promote their use as reagents in future syntheses
of larger drug-like molecules. In addition, calculated LogS
[
13
]
values greater than 3 are
used to filter out potential library members. Still, from the AlogP98 (Figure 9.2) and PSA
(Figure 9.3) distributions, Maybridge's fragments have higher values than the ActiveSight
set, an indicator of better solubility. Using Maybridge's Reactive Intermediates, uniquely
capped fragments such as carboxamides and sulfonamides have also been included as
suitable building blocks for quick follow-up of fragment hits.
It is clear that similar design criteria can yield very different distributions of physical
properties and topologies. While these libraries are predominantly designed for crystal-
lography screening, even greater differences in fragment libraries may be found when
targeting NMR or SPR detection methodologies where solubility constraints may differ.
More importantly from a virtual screening perspective, the physical limitations imposed by
the various empirical approaches need not restrict which fragments are considered. Thus,
all fragments are possible and only in the context of the design and target are constraints
introduced.
9.3 Virtual Screening Using Fragments
Over the past decade, numerous companies have focused on fragment-based screening
using X-ray crystallography,
[
1, 14
]
virtual screening,
[
15
]
SAR by NMR,
[
16
]
high concentration
bioassay,
[
17
]
mass spectrometry,
[
18
]
and the automated ligand identification system (ALIS)
technology developed at NeoGenesis.
[
19
]
The term 'fragment' itself has had various defini-
tions, referring to compounds typically between
150 and 450 Da. In the present discussion,
'fragments' will usually refer to small molecular compounds of less than 300 Da.
Computational chemists play a significant role in the lead optimization stage of drug dis-
covery. In this position, computational chemists can aid in fragment identification through
at least two avenues. First, fragments may be identified through the use of chemoinformatic
tools such as screening and filtering of candidates by calculated physical properties, which
is the more commonly applied involvement of computational design in fragment selection.
Second, modeling may aid in selecting fragments to couple with a scaffold to enhance bind-
ing. A scaffold may be defined as a single structural motif which has demonstrated binding
efficacy to a specific target or provided support for critical pharmacophoric elements. It is
possible that the scaffold itself does not have measurable binding until it is part of a fully
elaborated molecule. The scaffold could be any size fragment with sufficient functionality
and/or proper shape and size to be accommodated by the protein's active site. The scaffold
may be a fragment whose binding poses have been empirically identified through biolo-
gical screening of a fragment library. Typically, chemists test different adornments of the
scaffold to increase binding and specificity by synthetically combining a scaffold with a
fragment. Molecular modeling can assist this lead optimization effort by selecting potential
∼
