Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 8

receptor (e.g. in micromolar range). In this case, again, the X-ray snapshot of
the receptor cocrystallized with the ligand may not represent the optimal
binding mode, nor the ensemble of binding conformations, for the ligand. In
these cases, the binding modes have to be refined by computational sampling
of various spatial positions of the ligand within the receiving site of the
receptor to sample orientations in which the ligand binds the receptor more
tightly. The roles of entropy and enthalpy of complex formation are not
independent and simple comparisons of affinity (DG) without dissection
into its components, DHandDS, can be misleading [12].
These considerations illustrate the point that in peptide and protein
design, whether structure-based or target-based, there are certain pro-
blems for which we require computational modelling. At the same time,
one should not overestimate the precision of current computational
results. For typical applications, computational approaches generate
plausible suggestions regarding structural aspects of the functional
recognition of peptides and proteins. In all cases, these suggestions
have to be independently validated either by direct experimental
structural measurements or by confirmation of predictions through
biological experiments.

2.2.2 Tools and Procedures: Force Fields and Sampling

General protocols of any computational approach to peptide and pro-
tein design regularly include two key aspects: the generation of possible
molecular conformations and relative orientations of interacting mole-
cules (sampling) and the evaluation of the plausibility of the generated
conformations or orientations in terms of their relative energies (scor-
ing). The more thorough the sampling protocol and the more accurate
the scoring function, the more reliable the predictions. Ultimately, the
best results may be obtained when all possible states of a system in
question (conformations and relative orientations) are sampled and the
energy for each of the states is calculated employing high-level quantum
calculations. However, this best-case scenario is seldom applicable in
peptide and protein design, simply because of the system size, which over-
whelms available computer resources. Even with current rapid expansion
of computer capacity, it is unrealistic to expect adequate quantum chemi-
cal calculations for, say, a linear octapeptide in water - the system featur-
ing thousands of possible conformations of the peptide and millions of
configurations of the solvent - within the next decade.