Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 17

reproduce the results of high-level quantum calculations as closely as
possible.
Direct comparison with experimental data was recently performed by
simulating the Ramachandran map of Ac-Ala-OMe in water with the
AMBER polarizable force field [32] and by calculating the same free-
energy map in water with the polarizable AMOEBA (Atomic Multipole
Optimized Energetics for Biomolecular Applications) force field (Ponder,
personal communication). While the former study produced the
Ramachandran map with heavy populated b-region and alpha R region,
but not the pass or the alpha L regions (see [32], Figure 4), the AMOEBA
map appears very consistent with the experimental distribution of the
(f, c) points (see Figure 2.2(b)). Keeping in mind that calculation of the
AMOEBA map required computation only eight times larger than that
for the same calculation with the nonpolarizable AMBER force field
(Ponder, personal communication), prospects for the application of the
AMOEBA force field to peptides and proteins look encouraging.
However, polarizable force fields still have to be validated for their ability
to reproduce thermodynamics in a variety of examples to ensure their
wide applicability to peptide and protein design.
In summary, as comparison with the experimental data shows, devel-
opments of recent novel approaches to molecular force fields should lead
to a much wider use of the QM/MM and polarizable force fields in
peptide and protein design. Surprisingly, however, the comparison also
shows that the relatively simple ECEPP force field, which was developed
more than two decades ago, reproduced the distribution of experimental
(f, c) points on a Ramachandran map with accuracy comparable to that
of the QM/MM and polarizable force field approaches, and surpassing
that of the more sophisticated GROMOS and OPLS force fields.

2.2.2.2 Scoring functions

The other current trend in describing interactions in peptides and espe-
cially in proteins suggests not making the interactions more complex but,
on the contrary, simplifying them. In such a system, each amino acid
residue, and some functional atomic groups, is reduced to one of several
interaction centres. (In fact, combining the aliphatic hydrogens together
with the bonded carbon atom into a united centre such as CH n , which is a
common feature of many atom-atom force fields, is already an example
of such simplification.) Potential functions for interactions between the
centres may be deduced by averaging atom-atom interactions between