Introduction to Fragment-based Drug Discovery - Fragment-Based Drug Discovery - page 7

protein site due to free energy considerations in the shift to a single entity. Indeed, initial
attempts to exploit fragment-based drug design centred on the strategy of linking fragments;
the ideas behind the strategy have since been explored further. Murray and Verdonk [ 29 ] ana-
lysed the change in the free energy of binding for a molecule formed through linking two
separate fragments. Essentially the total free energy of binding for a fragment is divided
into two components, the free energy associated with the loss of rigid body entropy and the
remaining free energy contribution, termed the intrinsic free energy, which incorporates
factors such as the protein-ligand interactions and intramolecular conformational restric-
tion of the fragment. The free energy for the fragment-linked molecule is then a summation
of both fragment intrinsic free energies, additional free energy terms associated with the
linking group and only a single rigid body free energy contribution. The magnitude of
the rigid body term is estimated to be 15-20 kJ mol − 1 and independent of the size of the
molecules under consideration. The larger fragment linked molecule only then incurs the
same entropic penalty as for a smaller fragment, thus leading to a saving in the total entropic
penalty as comparedwith the individual fragments. Fragment linking has shown limited suc-
cess as it is very much dependent on the ability to link chemically the individual fragments
without significantly perturbing the effectiveness of the fragment receptor interactions. As
discussed previously, this is by no means easy and careful consideration must be taken in
the design of the fragments and the ability to link the fragments synthetically in the context
of the receptor. It is also the case that it is less common to identify fragments that could
bind simultaneously in adjacent regions of the target binding site. It may be simply because
there is insufficient space in the binding site to recognize multiple fragments simultan-
eously or adjacent sites, if available, may not provide a suitable environment for fragment
recognition. As discussed below, features in a binding site are unlikely to be distributed
evenly, meaning that there is a greater probability of identifying fragments that match a
specific region in a site. Alternatively, what has proved more amenable is the merger of two
fragments into a larger, more potent, compound or more simply the use of the structural
information to rationally improve upon a fragment hit (Figure 1.1). There is an increased
probability of identifying the separate, less complex fragments compared with the larger,
more potent, compound and also a greater chance of achieving maximum affinity.
The effectiveness of the interactions between a receptor and a small molecule is nowadays
discussed in terms of ligand efficiency. There are many different ligand efficiency indices,
but essentially they are all a means of normalising the affinity of a molecule with respect
to the size of that molecule, [ 30 ] thereby providing a measure of the quality of fit for that
molecule to the receptor. Quality of fit is an important consideration in both the selection
of compounds from a screen, whether that is a screen of drug-like molecules or a fragment-
based screen, and in the development of molecules through to preclinical candidates. The
origins of ligand efficiency can be seen in earlier work investigating ligand-receptor inter-
actions. Our understanding of small molecule-protein interactions is as yet insufficient
to be able to predict binding affinities accurately, however, from an analysis of existing
data Andrews et al . [ 31 ] formulated a means of ranking drug-receptor interactions based
on known functional group contributions. Small charged groups were found to contribute
significantly, followed by polar groups and finally nonpolar groups. The experimentally
observed binding affinity of a molecule can then be compared with an estimated value
obtained by summing the intrinsic binding energies for these constituent groups taking into
account entropic penalties. If the affinity is greater than average, then the fit to the receptor