Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 37

All energy calculations were performed with the ECEPP force field
using an in-house-developed program. Employing the ECEPP force field
provided at least two significant advantages compared to employment of
the force fields with flexible valence geometry. First, energy minimization
occurred in the space of the dihedral angles (about 700 dihedral angles in
the TM region of CXCR4) and not in the space of the Cartesian atomic
coordinates (about 6000 atomic coordinates), which ensured much faster
computations. Second, practical experience showed that the process of
energy minimization over the large number of Cartesian atomic coordi-
nates using a force field with flexible valence geometry almost always
converged to a local energy minimum closest to the starting point, due to
accumulation of small deviations of the starting Cartesian atomic coor-
dinates. As a result, the final 3D structure of a TM bundle was very close
to the initial TM bundle of rhodopsin. In other words, modelling of the
TM bundles for different GPCRs by homology to rhodopsin using the
force fields with flexible valence geometry yielded 3D structures virtually
the same as the X-ray structures of rhodopsin. These results could hardly
be consistent with the real 3D structures for the amino acid sequences of
different GPCRs.

2.3.2.1 Caveats and comments

Since there were no energy estimations for different conformations of the
helix bundle, many uncertainties discussed above, such as selection of the
values for the energy cutoffs, were not applicable in this case. However,
the packing protocol we have developed [108] assumes that the backbone
conformations of the TM helices do not change during packing and only
the side chains undergo mutual adjustments (the 'hard core' and 'soft
shell' assumption [109]). This assumption allows computation runs to be
accelerated significantly, since energy minimization can be performed
only in the space of the dihedral angles for the side chains, and not for
those of the backbone, reducing the dimension of the search space from
about 700 dihedral angles to about 300. Also, there is no need to calculate
the energy of interatomic interactions within the peptide backbones of
the TM helices, since the energy cannot change during minimization. The
assumption is based on several experimental findings. For instance, the
average diameter of the TM helices in TM proteins is about 10 ˚ [110],
which suggests that the backbone elements of the helices are not involved
in direct interaction. Also, the (f, c) values in the X-ray structures of the
TM helices rarely exceed the limits of 60-40 used in energy