Peptide-Based Drug Design

Cationic Antimicrobial Peptides 135
peptides. In addition, depolarization experiments of the cytosolic membrane
of tested bacteria showed only minor depolarization effects of the inactive
peptides compared to active peptide sets. Both results—structural features and
depolarization—suggest that the inactive peptides are hampered in their ability
to interact with lipids/membranes (36). In addition, the screening of 1400 biasedrandom
peptides once again confirmed the importance of the balance between
charge and hydrophobicity, as well as the fact that the “right” composition alone
is not sufficient to create an active peptide (K. Hilpert et al., manuscript in
preparation).
3. Predicting Antimicrobial Peptide Activity Using Quantitative
Structure-Activity Relationships
3.1. Introduction
A QSAR seeks to relate quantitative properties (descriptors) of a compound
with other properties such as drug-like activity or toxicity. The essential
assumption of QSAR is that quantities that can be conveniently measured or
calculated for a compound can be used to accurately predict another property of
interest (e.g., antibacterial activity) in a nontrivial way. QSAR has become an
integral part of screening programs in pharmaceutical drug-discovery pipelines
of small compounds and more recently in toxicological studies (69). However,
the use of QSAR modeling applied to the search for antimicrobial peptides is
relatively recent. Advances in this area are reviewed in brief here.
There are two separate but interrelated aspects to QSAR modeling of antibacterial
peptides: the choice of QSAR descriptors and the choice of numerical
analysis techniques used to relate these values to antibacterial activity. A simple
example of a QSAR descriptor is the total charge of a peptide. A large number
of QSAR descriptors is available for small compounds in the literature and
from commercial software products that may be considered. A smaller subset
is used in QSAR studies of antibacterial peptides and may be separated into
two categories: descriptors based on empirical values and calculated descriptors.
An example of an empirical value is HPLC retention time, which is a surrogate
measure of solubility or hydrophilicity/hydrophobicity. An example of a calculated
descriptor is total peptide charge at pH 7.
In addition to the choice of QSAR descriptors, many statistical learning
methods are available to relate the descriptors to the predicted value. There are
two main categories of prediction to answer two different questions: regression
models (for predicting the activity of a peptide as a continuous variable such as
MIC or a surrogate such as in the luminescence assay) or classification where
the model is trained to classify as simply active or inactive. Historically, linear