Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 32

similar in all analogues with high affinity, but dissimilar to all conformations
available to stereoisomers with low affinity. This selected set of common
active conformers would be the most plausible candidate for the 3D phar-
macophoric model of FC131.
Conformational sampling for several cyclopentapeptide compounds
with high and low affinities to CXCR4 required several steps. First, all
possible templates for backbone conformations were generated system-
atically by combining all local minima on the Ramachandran map char-
acteristic for each chiral residue in a given cyclopentapeptide. This
approach was justified by the analysis of experimentally determined cyclo-
pentapeptide structures [105], showing that those compounds almost
exclusively adopted combinations of (f, c) dihedral angles that were
close to allowed regions for linear peptides in the Ramachandran map.
Backbone conformations that allowed closing of the pentapeptide ring
were selected as starting points for further energy minimization. Energy
minimization was performed employing the ECEPP force field, and all
redundant conformations (those conformers with similar values of all f
and c angles) were removed. The choice of the ECEPP force field at this
step avoided low-energy conformers with unrealistic (f, c)values(suchas
those in the lower-right quadrant of the Ramachandran map for L -amino
acid residues), which might have occurred in sampling of cyclopentapep-
tides employing other force fields with flexible valence geometry [106].
The important side chains in FC131 and its analogues (Tyr 2 ,Arg 3 ,Arg 4 ,
Nal 5 ) are of a size comparable with the cyclic backbone, so sampling of the
side-chain rotamers should be rather exhaustive. Starting conformations for
the side-chain rotamers were generated systematically for each backbone by
including all the relevant side-chain rotamers (i.e. 60, 60 and 180)for
most of the w i dihedral angles. Then each starting conformation was sub-
jected to energy minimization that was, in turn, performed in two sequen-
tial steps. Since the number of generated starting ring conformations varied
between 10 5 and 10 6 from analogue to analogue, energy minimization was
first performed with the ECEPP/2 force field, which employs rigid valence
geometry that drastically reduces the computational time required for each
minimization run. This in-house-developed program was used for all calcu-
lations employing the ECEPP force field. The calculations were performed
in vacuo with a dielectric constant (e) of 80.0. There are considerable
uncertainties involved in mimicking the heterogeneous transmembrane
(TM) protein environment of the CXCR4 receptor, and this treatment
was chosen to dampen the strong electrostatic interactions between charged
groups, thereby allowing exploration of a wider set of low-energy confor-
mations. After selection of low-energy conformations by a rather high