Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 6

with its specific receptor. The apparent reason is the inherent conforma-
tional flexibility of peptides. Most peptides exist under physiological con-
ditions as a mixture of more or less well-defined, interconverting
conformers. The interconversion rate is such that, for instance, NMR
spectroscopy with characteristic resolution times of 10 5 -10 3 seconds
does not distinguish separate conformers of linear peptides in the kilo-
dalton range in solution (with the exception of cis / trans peptide bond
isomers). The same is true for CD, IR and ESR spectroscopy. In the absence
of one highly predominant conformer, the 3D peptide structure deduced
from physicochemical measurements (e.g. from NMR parameters such as
NOEs, vicinal coupling constants, etc.) reflects the average over the ensem-
ble of conformers present in solution and, in this sense, could not be related
to any of the 'real' peptide conformers at all. On the other hand, the
conformation of peptide ligands corresponding to the 3D pharmacophore,
i.e. to the ligand conformation in the complex with receptor, may not
necessarily be the one with the highest statistical weight in solution, since
some other conformers may acquire the highest statistical weight in the
peptide-receptor complex, being compensated by much more favourable
interaction in the complex with the receptor. At the same time, X-ray
crystallography produces only individual 'snapshots' of peptides, each
representing a single 3D structure stabilized by the crystalline lattice
from among the set of possible conformers existing in solution. For the
highly flexible enkephalin molecule, for instance, X-ray crystallography
obtained snapshots of the four drastically different 3D structures ranging
from fully extended to various types of b-reversals (see review [8]).
On the other hand, computational methods, being applied to various
analogues of the same peptide that differ by values of affinity (or activity)
toward a specific receptor, may model all 3D structures feasible for the
parent peptide and its analogues from the energetic and/or sterical point
of view. Then one may compare sets of those structures to one another
and select those among the biologically active analogues in which the
important functional groups are arranged in space similarly. These struc-
tures may be regarded as reasonable candidates for 3D pharmacophores,
which in turn may be stabilized by introducing constraints through
chemical synthesis. The structure-based design employing this approach
has been successful in developing novel cyclic analogues of linear pep-
tides, many of which are biologically active (such as analogues of opioid
peptides, angiotensin, a-melanotropin, etc.; see earlier review [3]).
Historically, target-based design came after structure-based design, since
detailed information on the receptor molecules only became readily avail-
able in recent decades. This progress was made largely due to rapid