Without any knowledge of potential hidden conformations, a
search with molecular simulation can rely on computational brute
force, which, despite many recent advances in high performance
computing, is still far from sufficient in most cases. DESRES has
reported examples where long unbiased simulations attempting to
make difficult transitions between known active and inactive con-
formations have found intermediate states which might be amena-
ble to structure-based design [132, 133]. Seeded adaptive,
unbiased sampling has achieved similar feats to DESRES and
Anton by exploring known transitions with unbiased dynamics
and finding useful conformations and hidden pockets [12, 46, 75].
These authors have been unable to find examples of any
enhanced simulation applied to a system with a single known
conformation and finding significantly kinetically separated new
conformations. While it is likely that these hidden conformations
are waiting to be found, current simulation methods may not be
sufficiently powerful to find them. Combinations of adaptive cel-
ling and potential biasing approaches may hold the key to this type
of powerful search routine (seeNote 19).4.4 Understanding
Allostery Without
Conformational
Change
Allostery without conformational change is a concept which has
been discussed for many years [134]. This allostery is not restricted
to those systems which are more rigid and ordered but just excludes
a significant conformational change in geometry as the allosteric
mechanism. The most relevant situation in drug discovery is the
discovery of an allosteric compound through experimental screen-
ing where comparisons of available crystal structures of actives and
non-actives or apo-structures seem the same. Often an allosteric
small molecule will need to be optimized, which can be difficult
when the molecular mechanism of action is not understood.
Understanding the potential for allostery when evaluating target
tractability is probably a less common problem, but this is a require-
ment for evaluation of a target or an allosteric pocket. Understand-
ing transitions or allosteric signals without an obvious or significant
geometrical change has many of the same challenges discussed
above for transitions with obvious changes. The added difficulty is
characterizing the transition even if this has been discovered or
sampled as their structures can be very similar.4.4.1 Challenge of Long
Timescales and Hidden
Transitions
This difficulty in predescribing cryptic allosteric transitions has
caused a dearth of studies investigating these problems with driven
simulation methods which use specific CVs. An interesting but rare
example of this is the recent study of the KIX domain of CREB-
binding protein with metadynamics [135](seeNote 20). aMD is
also a sensible sampling method to sample cryptic allosteric transi-
tions within single geometric conformations [136, 137].
Given the difficulty of sampling long timescale cryptic transi-
tions, less computationally expensive approaches may offer a way352 Benjamin P. Cossins et al.