Assembling a Fragment Library - Fragment-Based Drug Discovery - page 50

3.3.2 Elimination of Toxic and Reactive Chemistry

Although present in several approved drugs, protein reactive chemical groups are known
to produce screening artefacts and are undesirable starting points for F2L optimization. [ 46 ]
We eliminated potentially protein reactive chemical groups, such as aldehydes, phosphon-
ate esters and activated ketones, in order to avoid identification of covalent, irreversible
fragment binders. Furthermore, we also removed potential toxicophores.

3.3.3 Molecular Weight (MW)

MW is the most obvious, but also the most stringent, filter for fragment-like compounds
within commercial screening compound collections, reducing the number of available com-
pounds immediately by a factor of 10-20. We set the upper MW limit of our NMR fragment
library to 310 Da and targeted an average MWof 250 Da. This also allows for the merging

or linking of two fragments binding at adjacent sites, as demonstrated in the SAR-by-NMR
study of Fesik and co-workers, [ 42 ] to provide linked/merged compounds that conform to
drug-like criteria.
In addition, we set a lower MW limit of 115 Da and accepted less than 0.4% of all
compounds smaller than 150 Da because fragments of significantly lower MW become
similar to solvents, such as DMSO (94 Da). Such small fragments are less likely to
overcome the energetic hurdles of desolvation and of inducing small conformational
changes in amino acid side-chains, that have been demonstrated for many protein-small
molecule complexes. [ 47 ] Moreover, small fragment binders can change their binding geo-
metry during F2L optimization, as was shown experimentally in a recent crystallographic
study. The two 128 and 153 Da constituents of a 343 Da lead compound adopt com-
pletely different binding geometries when co-crystallised one at a time as compared
with their pose in the co-crystal of the lead. [ 48 ] Consequently, utilizing fragments of
too low MW puts binding mode conservation and hence rational fragment elaboration
at risk.

∼

3.3.4 Solubility

Solubility is another critical and stringent filter for fragment libraries. Compared with
HTS libraries, fragments require additional hydrophilicity to be soluble to a minimum
NMR assay concentration of 200-1000 M in aqueous buffers containing 2-10% DMSO.
On the other hand fragments should contain sufficient hydrophobicity to make effective
van der Waal interactions with the target protein and to provide the basis for optimisation
into lead compounds with satisfactory permeability in cellular assays and in vivo efficacy
models. Hydrophobicity is essential for preferentially targeting protein pockets with strong
hydrophobic contributions to binding because protein pockets with primarily hydrophilic
interactions have a lower probability of leading to orally available drugs.
Balancing these opposing requirements, we estimated fragment solubility by utilising
the parameters calculated log P (Clog P ) and calculated log S w (solubility in water, Clog S w )
and set the upper and lower limit to Clog P s of 3.5 and 0.5, respectively. In comparison
with most other fragment libraries for biochemical, X-ray and other biophysical assays,
this Clog P range uniquely shifts the NMR library towards higher hydrophobicity.