Chemistry Reference
In-Depth Information
without loss of drug-like character.
[
4
]
A recent retrospective analysis indicates that the suc-
cess of fragment optimization and progression can be accurately predicted.
[
2
]
The main
factor for successful fragment progression is the availability of structural data to guide the
optimization process.
Optimization of the potency and specificity of lead-like and drug-like compounds using
experimental structural information on protein-ligand complexes is awell-established com-
ponent of the drug discovery process.
[
5
]
As noted above, the availability of experimental
structural information on fragments bound to their target is crucial to driving the pro-
gression from relatively simple, minimally functionalized weak hits to potent, elaborated
leads.
[
2
]
There are primarily three options for generating molecular structure-based inform-
ation: molecular modeling, X-ray crystallography and NMR spectroscopy. While various
computational tools can often generate predictive binding models of larger inhibitors, bind-
ing mode prediction for fragments using computational methods remains problematic.
Although fragments generally have fewer internal degrees of freedom than larger com-
pounds, their location and orientation within a binding pocket can be more difficult to
predict, since alternative binding poses are more likely to yield similar scores or predicted
binding energies, i.e. multiple minima problems are more common. Whereas progress has
been made in the computational prediction of fragment binding poses (e.g. see ref. 6),
experimental binding pose characterization remains critically important.
[
7, 8
]
When feas-
ible, X-ray crystallography can be used simultaneously to screen a fragment library for
hits and to provide high-resolution structural information on the binding poses of the
hits obtained.
[
9 12
]
A challenge in the crystallographic determination of protein-fragment
complexes is the need for high-resolution structures, typically
<
2.0 Å, to place correctly
pseudo-symmetry-related groups with similar electron densities, e.g. a methyl versus an
amino group. Traditional NMR methods, based on observed and assigned protein-ligand
NOEs, have been used to characterize the structures of fragments bound to proteins
[
1, 13, 14
]
and to guide the design of potent 'linked compounds', as exemplified in the original report
of the 'SAR by NMR' method.
1
The main challenges for the NMR-based approaches to
structure determination are feasibility and time. Regarding the latter, both the acquisi-
tion and analysis of the multiple data sets necessary to define the bound structure of the
ligand can greatly constrain the number of protein-ligand complexes that can be charac-
terized, limiting the information that can be generated for the numerous hits identified in a
fragment-based screen.
In an effort to accelerate the structure determination of protein-ligand complexes by
NMR, we have developed an approach, termed 'NOE matching',
[
15
]
that allows one to
determine the binding pose of a ligand without first having to determine resonance assign-
ments for the target protein. Briefly, NOE matching is a pattern-matching procedure that
derives 'cost' values for trial binding poses based on how well each trial pose predicts an
experimental 3D
13
C-edited,
13
C/
15
N-filtered HSQC-NOESY spectrum.
[
16
]
The key advant-
age of the NOE matching approach is that it can be applied without NMR assignments for
the protein in the protein-ligand complex. We have previously shown that NOE match-
ing can identify the correct binding pose with reasonable accuracy. In this chapter, we
briefly summarize the NOE matching protocol and its previous application to lead-like
compounds bound to small proteins.
[
15
]
We then describe the application of NOE matching
to fragments and discuss some of the unique challenges encountered with these small com-
pounds. Finally, because the majority of pharmaceutically relevant targets are relatively
