Chemistry Reference
In-Depth Information
on the fact that the compounds are to be used in the context of fluorescence readouts so
a subset of the filters avoid readout artefacts due to the screening compounds themselves.
The filters were applied as substructures encoded in SMILES notation to automatically
identify undesirable compounds in the database.
3.4.3
Solubility
A solubility filter using an in-house
in silico
model to select compounds predicted to have
an aqueous solubility in excess of 1mM was applied. Aqueous solubility is an important
factor for the success of HTS screening in general and fragment screening in particular.
Due to the low affinity binding of fragments at protein interaction sites, the screening
concentration range is typically at least 20-fold higher than that used for general HTS
screening. Depending on the target, the primary fragment screens at Evotec are usually run
at concentration ranges between 250 M and 1mM final concentration. Dose-response
curves can reach final concentrations up to 2mM. The stock solutions of the fragments are
stored in 100%DMSO. In a proprietary process (NanoStore), the solvent is removed on the
assay plates prior to screening to generate a DMSO-free assay environment.
[
51
]
We validated
the application of NanoStore to fragment screening by high-concentration bioassay early on
in a pilot study of a subset of the Stage 1 fragment library. This fragment set was screened
against PTP1B, for which we achieved superior results using NanoStore in comparison
with presentation of the fragments in DMSO.
[
52
]
The NanoStore method of presentation of
compounds for screening avoids many assay artefacts but still requires fragments to have
reasonable aqueous solubility. Setting the cut-off of the predicted solubility at 1mM gives
a good compromise between reliability of the
in silico
predictions and constraining the
coverage of chemical space.
3.4.4 Physicochemical Parameters, Features and Molecular Weight
Calculated properties were restricted to: Clog
P
≤
3, the number of rotatable bonds
≤
5, the
number of hydrogen bond acceptors (HBA)
4 and the number of hydrogen bond donors
(HBD) between 1 and 3. The topological polar surface area (TPSA) was set to
≤
70 Å
2
.
In addition, a special feature count excludes structures that are too functionalised (feature
rich) or dull (absence of pharmacophoric interaction points). This feature count is the sum
of HBA, HBD and the number of five-and six-membered aromatic rings and selection was
restricted to fragments with a feature count in the range 4-7.
In order to meet our requirements for a relatively large fragment screening library to
maximise the potential of the HTFS concept, it was necessary to relax the upper limit
on MW in order to access sufficiently diverse compounds. A final MW filter was applied
to allow only compounds with MW in the range 150-350 Da. This MW range is slightly
broader than used by others (Table 3.2).
Application of all of the above filters resulted in a list of approximately 32 000 unique
structures that are available from commercial vendors. During the assembly of the Stage 1
and 2 iterations of the fragment library, an overlap check to the current screening collection
was applied trying to avoid overlap in those cases where a target would be addressed by
both full-deck HTS screening and the fragment-based approach. This concern was dropped
later for the Stage 3 iteration, since the exclusion of the chemical space from the full HTS
≤
