Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 38

calculations for individual TM helices (see e.g. [10,111]). Nevertheless,
independent experimental validation of the computational results
obtained using this assumption would be highly desirable. However, in
the absence of direct structural data on CXCR4, the only available
experimental data are those from site-directed mutagenesis. Compari-
son with those data would be more meaningful for results obtained after
computational docking of FC131 to CXCR4.

2.3.3 Docking of FC131 to CXCR4

Docking of FC131 to CXCR4 [7] was achieved in several sequential
steps. In fact, docking was performed not only for FC131, but also for
10 analogues of FC131 with high or medium affinity to CXCR4 in order
to elucidate orientations (binding modes) common for all potent cyclo-
pentapeptides. The previously deduced 3D pharmacophore model (see
Figure 2.4) was docked to the obtained 3D models of the TM region of
CXCR4 using the AutoDock program, which assumed a rigid receptor
(CXCR4) with flexibility (rotation of bonds) of the ligand (cyclopenta-
peptides). Sampling of possible orientations of the ligand employed a grid
map of 1273 points with a spacing of 0.375 ˚ , which covered the whole
TM helical bundle of CXCR4, with the exception of a few residues on the
intracellular side. The Lamarckian genetic algorithm (LGA), which uses a
combination of a genetic algorithm and a local search, was used as the
sampling method, and the standard scoring function implemented in
AutoDock was used to evaluate ligand orientations.
LGA docking selected 100 orientations with the most preferable values
of the scoring function for each of the 11 compounds. The orientations
obtained for each compound were then divided into clusters using an
RMS cutoff of 1.0 ˚ (all atoms, no translation). Only clusters found near
or within the extracellular pocket formed by the seven TM helices of
CXCR4, as judged by visual inspection, were considered further, i.e.
clusters representing binding to the lipid-oriented transmembrane faces
of the receptor were discarded. For each of the remaining clusters, the
orientation with the lowest value of the scoring function was extracted as
a representative. Based on the 10 atoms defining the 3D pharmacophore
(see Figure 2.3), an in-house program was used to compare orientations
within the TM helices of CXCR4 for different compounds with the less
stringent RMS cutoff of 3.0 ˚ (no translation). This resulted in the
identification of two binding modes (A and B) that were common for