simulations into one analysis and hence can be easily parallelized
across any size of computer resource.
MSM analyses of very rare events are more difficult as the
simulations which provide kinetic data will not sample these events
often. One way around this difficulty is to “seed” simulations with
data from potential bias methods which explore these rare events
more readily. These seeded kinetic analyses have the potential to
provide the highest quality biophysics data on rare events of interest
with a lower cost [46]. Even in the absence of seed data, adaptive
sampling arrangements can be utilized to enhance sampling effi-
ciency [47, 48].
One of the likely difficulties for large scale celling approaches is
analyzing the huge amount of MD data produced. D. E. Shaw’s
research has developed an open-source code for analyzing many
terabyte trajectories using highly parallel computation [49]. On-
the-fly analysis of data as it is produced such that the best starting
points for new trajectories can be used will be critical for explora-
tion with high performance celling approaches (seeNote 13). If a
specialized hardware/software solution for an enhanced celling
approach could be designed it would be more efficient than the
single very long trajectory as the simulation would not spend time
exploring space it has already characterized.4 What Is Useful for Working with Protein Targets in Drug Discovery?
The field of molecular simulation is now being taken more seriously
in drug discovery research as projects driven by computational
biosimulation have very publically produced valuable candidate
compounds [50]. These nice examples have so far used MD to
predict small molecule binding affinities allowing the rapid explo-
ration of larger swathes of chemical space.
As have been described by various experts, computational
approaches are not yet close to being the first choice technology
in the drug discovery process [51]. However, it is the belief of these
authors that molecular simulation can offer more to drug discovery
research than these examples of binding affinity prediction. A con-
cept for this could be the use of dynamical models to predict target
tractability prior to the engagement of experimental approaches
like protein production allowing a more developed strategy from
the early stages of a therapeutic project [52].
Another concept for the use of molecular dynamics in drug
discovery research is exemplified by the approach of D.E. Shaw
research (DESRES). Since 2002 David Shaw has focused on build-
ing MD technology and applying it to biochemistry and drug
discovery. DESRES has built specialized computers for MD which
encode the calculation into ASIC (applications specific circuit)346 Benjamin P. Cossins et al.