Chemistry Reference
In-Depth Information
3.3.8 Avoidance of Unnecessary Restraints
In order to map thoroughly the available chemical space, we avoided the introduction of
assay-related chemical restraints, which are frequently used for other fragment assays,
such as bromine libraries for X-ray assays, fluorine libraries for
19
F-detected NMR assays,
fluoro/chromophore elimination for biochemical assays, lengthy linkers for chip-based
assays, chemical tags for transient immobilisation, methyloxime-based linking or
in situ
click chemistry. It is our view that starting points for small molecule drugs should be as com-
pact and as binding efficient as the target allows. This goal is ideally supported by our NMR
assays, which impose solely the following, nonstringent restraints on the chemical matter:
1.
1 non-water-exchangeable H atom for ligand-detected NMR assays;
2. preference for
≥
1 aromatic ring which ensures stronger chemical shift signals in protein-
detected NMR assays and better spectral resolution in ligand-detected NMR assays.
≥
3.3.9
Size and Diversity
Finally, the fragment collection filtered for the above-mentioned criteria needs to be pri-
oritised for the desirable size and diversity. For several biophysical fragment screening
techniques, particularly crystallography, the library size is limited by throughput and the
cost of protein consumption. Nevertheless, library size and diversity need to be balanced
within an economical range, whichwe approximate to be between 500 and 25 000 fragments
for NMR techniques. We intentionally decided to assemble a library of 20 000 compounds,
which is the largest NMR screening library published so far, because we wish to sample
thoroughly the available fragment space in order to identify the most efficient binders. This
screen-big strategy reduces the impact of false negatives in fragment assays and minimises
the number of singleton hits. A small and highly diverse fragment library may miss out
attractive hits, which show up as hits of analogues in a larger, but less diverse, fragment
collection.
In practice, the size of fragment library screened will vary between protein targets due
to more than 10-fold differences in protein expression yield and substantial cost differ-
ences between bacterial and mammalian expression systems. Therefore, we sorted our
entire collection of 20 000 fragments into a maximum diverse subset of 5000 fragments for
costly proteins and a second subset of the remaining 15 000 fragments by using a chemical
dissimilarity algorithm.
[
50
]
Protein targets, which are even too costly for the 5000 subset, can be prescreened by
virtual screening (VS) or a biochemical assay and theNMRassay restricted to an economical
library size of 100-1000 hit compounds. Taken together, we believe that it pays off to screen
experimentally as many fragments as economical for a target. This notion appears to have
been adopted by other proponents in NMR screening, such as Abbott with an NMR library
of approximately 10 000 compounds (Table 3.2).
3.3.10 Library Extensions, Modifications
Since the most favourable chemical fragment space for any given class of targets varies
over time, we recommend planning ahead for dynamic library extensions, which may
comprise target-focused fragments, new chemistries, new privileged chemical classes or
