Biology Reference
In-Depth Information
strategy for generating peptide libraries. In a cyclic or iterative
process, virtual peptide libraries are generated by variation of a seed
sequence (parent), the activity of each peptide (offspring) is predicted
by the ANN, and the peptide (child) with the highest predicted activ-
ity is selected as the parent for the next cycle. This is repeated until no
further optimization can be observed. To evaluate the ANN, the
selected peptides were tested for their activity and compared to the
predicted activities. Despite some conflicts in the degree of binding
activity, all peptides predicted as active by the ANN showed activity
when tested. The most active peptides shared only one common
residue, aspartic acid, at position 7 with the parent or seed sequence.
Furthermore, a peptide predicted by the ANN as being the most inac-
tive peptide, which contained favorable binding motifs in two positions,
in fact tested to be non-binding. Most striking was the observation of
the difference in activities of the
de novo
peptides when compared to
natural-derived peptides at different concentrations. Some of the
ANN-designed peptides had similar or significantly higher activities at
concentrations 10-fold lower than those of natural peptides.
This example for peptide/protein design shows that a mathemat-
ical approach using ANN methods to derive knowledge and search in
sequence space can be successful for the identification of mimetopes
(antigens mimicking a natural epitope sequence) with significantly
different residue sequences. It also shows that the most active peptide
represents a good starting point for the expanded search and further
optimization. One may describe the virtual strategy presented here as
the mathematical “molding” of molecular structural features to a
desired function. Thus, computer-based strategies for
de novo
mole-
cule candidate design have emerged as a complimentary approach to
high-throughput techniques and should prove beneficial for develop-
ment of immunotherapeutic and diagnostic candidates.
166,230-232
Other computer-based techniques have been used for the
sequence-oriented design of focused libraries. Wrede
et al
. (1998)
developed two computer-based techniques for the generation of arti-
ficial
E. coli
signal peptidase I (SP I) cleavage target sites. The first
