Mathematical Modeling and Optimization Methods for De Novo Protein Design - Systems Biology: Networks, Models, and Applications - page 58

performing massive chemical synthesis of peptide analogs, a few selected
analogs were tested against the theoretical prediction. Table 2.2 shows the
experimentally measured percent complement inhibition and peptide D1
is currently the most active compstatin analog available. The C2A/C12A
analog is inactive [73] and has been used as a negative control for the
inhibition measurements. Table 2.2 summarizes the results from the
inhibitory activity experiments in comparison to the theoretical fold
stability results.
Qualitatively, the predicted increases in fold stability and specificity
are in excellent agreement with the results from the experimental studies.
This is especially significant, given that the predictions correspond
more directly to fold stability enhancements, while the experiments
directly test inhibitory function.
The comparison between experimental and computational results
indicates that the most active compstatin analogs are sequences D1 and
B1, as suggested by the optimization study. The common characteristic
of these two sequences is the substitutions at positions 4 and 9, the two
positions flanking the b- turn residues, Gln 5 -Asp 6 -Trp 7 -Gly 8 . In partic-
ular, the combination of tyrosine at position 4 and alanine at position 9
are key residues for increased activity and lead to a 16-fold improve-
ment over the parent peptide compstatin (see table 2.2).
CONCLUSIONS AND FUTURE WORK
A novel computational structure/activity-based methodology for the de
novo design of peptides and proteins was presented. The method is com-
pletely general in nature, with the main steps of the approach being the
availability of NMR-derived structural templates, combinatorial selection
of sequences based on optimization of parameterized pairwise residue
interaction potentials, and validation of fold stability and specificity using
deterministic global optimization. The optimization study led to the iden-
tification of many active analogs including a 16-fold more active analog,
as validated through immunological activity measurements. Allowing
tryptophan in position 4, the in silico sequence prediction framework
demonstrates that tryptophan is preferred over tyrosine and tyrosine is
preferred over valine. This is in agreement with recent experimental
results [81] which showed a 45-fold improvement in the inhibitory activ-
ity of the peptide Ac-I[CVWQDWGAHRC]T-NH 2 .
These results are extremely impressive and represent significant
enhancements in inhibitory activity over analogs identified by either
purely rational or experimental combinatorial design techniques. The
work provides direct evidence that an integrated experimental and the-
oretical approach can make possible the engineering of compounds
with enhanced immunological properties. Future work will be focused