Biomedical Engineering Reference
In-Depth Information
employ pharmacophore queries [8]. In addition, there are also other types of methods
for which fewer molecular representations and algorithms have been introduced, such
as shape similarity searching [49]. Pharmacophore and shape searching are primarily
three-dimensional methods (i.e., they are based three-dimensional molecular repre-
sentations. However, two-dimensional pharmacophore fingerprints are also available,
and molecular shape representations can be approximated from molecular graphs.
Fingerprints can be derived from two-dimensional or three-dimensional molecular
representations, with pharmacophore fingerprints being the most popular category of
three-dimensional fingerprints [50].
15.6.1 Pharmacophores
Currently, four-point pharmacophore models represent the standard for three-
dimensional database searching [29,50]. In this case, pharmacophore queries are
used to search conformational databases of test compounds for hits that match the
four features and six interfeature distances of the model. Importantly, precise matches
of interfeature distances are rarely possible, and hence distance ranges are used to
relax the query and identify compounds that match it closely. Pharmacophore fea-
tures (or functions) that are assigned to feature points typically include hydropho-
bic groups, aromatic centers, negatively charged groups, positively charged groups,
hydrogen-bond donors, and hydrogen-bond acceptors. Pharmacophores are derived
from spatial alignments of active compounds in known or, more frequently, puta-
tive bioactive conformations, which is often referred to as pharmacophore mapping .
This process has been automated in many modeling programs [51]. A key assump-
tion underlying pharmacophore mapping is that the bioactive compounds share the
same mechanism of action and binding mode. Pharmacophore models cannot only
be derived from ligand superpositions but also based on active-site features of targets
with known x-ray structures. In this case, pharmacophores are modeled to be com-
plementary to important interaction sites in proteins. In pharmacophore searching,
subsets of database compounds are identified that match the query and/or that can
be ranked because of three-dimensional similarity criteria. In addition to individ-
ual pharmacophore models, pharmacophore fingerprints record large ensembles of
predefined pharmacophore patterns (i.e., combinations of geometric arrangements
and specific features). Each pattern is assigned to a single bit position in the fin-
gerprint. Based on systematic conformational search, patterns are recorded that are
matched by test compounds. Pharmacophore fingerprint overlap between active ref-
erence and database compounds is then calculated as a measure of similarity. Phar-
macophore models and pharmacophore fingerprints are illustrated schematically in
Figure 15.3.
Pharmacophore searching and fingerprinting has an important distinction. In phar-
macophore fingerprinting, no match to a specific pharmacophore model is evaluated.
Rather, overlap in generally defined pharmacophore space (i.e., the union of all
predefined patterns) is quantified based on systematic conformational search. The
underlying idea is that the greater the overlap of two molecules in pharmacophore
space is, the more likely they are to have similar activity. Thus, in contrast to
Search WWH ::




Custom Search