Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 34

Figure 2.4 Superimposition of the two conformations of FC131, together represent-
ing a plausible 3D pharmacophoric model for binding to CXCR4

2.3.1.1 Caveats and comments

It is noteworthy that the final results of the above computational protocol
significantly depended on several empirical (user-specified) parameters
that were applied at different steps of the procedure, such as selection of
the functionally important atomic centres, the values of energy cutoffs for
selection of low-energy conformations, or RMSD cutoffs for establishing
similarity criteria between conformations. The choice of such parameters
is always somewhat arbitrary, and in general varies from one specific
application to another. Some other uncertainties, mainly those from the
inevitable errors in energy estimation (inadequate force field, simplifica-
tions in description of electrostatic interactions and/or the molecular
environment) are present in any computational procedure, so it is essen-
tial to validate the resulting hypotheses either by comparison with experi-
mental data (when available) or by additional independent calculations.
In the case of the 3D pharmacophoric model for FC131, one aspect of
validation was convergence of the entire computational procedure to the
hypothetical structures in Figure 2.4 - it might happen that no choice of
the energy and RMSD cutoffs will yield a clear distinction between
analogues with high and low affinity toward CXCR4. Another aspect
was additional calculations performed for new analogues of FC131 with
different conformational possibilities that were not used in deducing the
3D pharmacophore. Such calculations showed that the hypothetical
models were still present in the sets of low-energy conformations for
the new analogues with high affinity to CXCR4 and at the same time
were absent in the sets of low-energy conformations for all new analogues
with low affinity. In other words, self-consistency is a requirement for
model building that must be maintained when confronted with new data.