Molecular Simulation and Systems Biology - Systems Biology: Networks, Models, and Applications - page 93

Greater computational capacity allows the use of more sophisticated
models for the molecular interactions (table 3.1). Since McCammon
and Harvey's 1987 work, great strides have been taken even in the use
of fixed charge force fields: united atom simulations, the norm in the
1980s, are now much less common; explicit representation of solvent,
rare in biomolecular simulations of the 1980s, has now become the
standard; Ewald electrostatics, rare in the 1980s, are now used for the
more accurate modeling of electrostatic interactions. These kinds of
trends in the quality of modeling will certainly continue into the future.
Finally, with greater computing capability and capacity comes also
improved sampling. The future should bring much greater and much
needed statistical precision associated with the properties measured
from simulations and this will greatly increase their utility and drive
further improvements.

Table 3.1 This table gives some indication of how the scale of simulation has
increased over time and particularly how the increase in the scale
of simulations tracks the increase in supercomputer capability
shown in figure 3.2.

Computational

Year Investigators System platform Time scale

1955 Fermi et al. [76] 64-particle chain MANIAC I 50,000 cycles

with nonlinear
interactions

1959 Alder and 108 hard spheres IBM 704 3000 collisions

Wainwright [77]

1964 Rahman [78] Lennard-Jones CDC 3600 10 ps

liquid (argon)
864 particles

1974 Stillinger and Water 216 IBM 360/195 22 ps

Rahman [79] molecules

1977 McCammon BPTI (in vacuum) 8.8 ps

et al. [25]

1983 Van Gunsteren BPTI (in hydrated Cyber 170/760 20 ps

et al. [27] crystal)

1994 York et al. [80] BPTI (in hydrated Cray Y-MP 1 ns

crystal)

1998 Sheinerman and Segment B1 of CrayT3D,T3E, 7.6 ns

Brooks [81] protein G C-90

1998 Duan and Villin headpiece Cray T3D,T3E 1 µ s

Kollman [21] (36 residues) in

3000 water
molecules