hardware and advanced interprocessor networks [53–55](seeNote
14 ). Published work shows very long timescale simulations and
detailed study of how important drug targets (tyrosine kinases)
function [56, 57]. Emphasis for DESRES seems to be on the details
of interaction networks and change points in protein dynamics
rather than thermodynamic and kinetic details found in similar
work from academic groups [58–61]. This emphasis has led to
the development of advanced statistical analysis to replace painstak-
ing visual analysis [62, 63].4.1 Target
Tractability
Many of the most valuable activities related to conformational
transitions in structure-enabled drug discovery can be classed as
target tractability work. This is mainly due to the fact that it is
sensible to understand the details of a target’s dynamics and func-
tion before embarking on a therapeutic project. Of course such
understanding is also valuable during lead optimization; however as
definition of conformational transitions can be time-consuming, it
can be more difficult to impact rapid chemistry cycles at this later
stage of a project.4.2 Searching
for Hidden Pockets
for New Chemical
Entities (NCEs)
An important aspect to NCE target tractability is the presence of a
pocket/cavity capable of sustaining strong binding to a small mol-
ecule. For many years there have been computational analyses able
to find pockets in protein crystal structures; some are physics-based
[64] while others are trained on large sets of known (substrate)
pockets [65, 66]. Even when a potential pocket is discovered with
one of these approaches, there is still much work required to
discover a compound which will experimentally validate it as a
binding site; additionally generating structural information on the
bound complex is even rarer.
There are many important target proteins which do not seem,
in available crystal structures, to have a cavity suitable for traditional
small molecules. However, there are pockets which only reveal
themselves when a compound, generally found only through exper-
imental screening, binds.
It is unclear how numerous these hidden or cryptic pockets
might be, however, molecular dynamics-based approaches have
been emerging which are able to suggest some of these pockets.
The simplest examples of this are studies where a relatively short,
normal MD simulation reveals a cavity which is predicted to be
ligandable by a prediction algorithm or a subcavity in a known
pocket [67–72]. These studies move on from those which analyze
static crystal structures to allow for a more realistic representation
of the protein target. However, these MD-enabled studies are
limited in that they can only hope to discover cavities which emerge
on relatively fast timescales, and often an element of chance is
involved. Other studies attempt to remove chance and use some
form of enhanced sampling to search for useful pockets more
thoroughly [73–76].Computational Study of Protein Conformational Transitions 347