Biomedical Engineering Reference
In-Depth Information
large numbers of compounds are tested experimentally. By contrast, VS attempts to
rationalize candidate selection by reducing the number of test compounds as much
as possible, and hence yields in successful instances only small numbers of novel
active compounds. In fact, a major attraction of the VS approach is its ability to sig-
nificantly reduce the number of candidate compounds for experimental evaluation.
For example, current screening libraries often contain several million compounds
whose chemical information content can be processed computationally in a target-
directed manner to identify small subsets of compounds, on the order of hundreds
of thousands, which have a high likelihood to be active. If required assay systems
are complicated, this reductionist approach is highly attractive, even if HTS capacity
is available. Although active compounds are usually enriched in database selection
sets, the specificity of VS calculations is generally limited regardless of the methods
that are employed. Consequently, the majority of candidate compounds in database
selections are usually false positives. Thus, if VS is successful, it can realistically
be expected to find a few novel active compounds in a database selection set: for
example, one to five among 100 to 200 database candidates [1,7]. On the other hand,
HTS is often plagued with large numbers of apparent (weakly potent) hits, many of
which turn out to be false positives in secondary assays. Hence, HTS hit triaging and
assay follow-up usually require much time and effort. Clearly, any experimental or
computational screening technology has its limitations. While hit discovery in phar-
maceutical research and development is clearly dominated by HTS, VS approaches
are established in drug discovery settings and their use is expected to further increase
as even larger numbers of compounds and screening targets become available in the
future. Over the past approximately 15 years, VS has become increasingly popular
in both pharmaceutical research and academia [10-13], and there is no end in sight.
In addition to practical relevance, the compound reductionist concept underlying
VS approaches, the search for “active needles” in large “compound haystacks,” and
the prediction of structure-activity relationships (SARs) present intellectually stim-
ulating tasks and challenging aspects for computational method development. These
challenges catalyze many exploratory activities in academia and in the pharmaceu-
tical industry to further improve current VS methodologies or develop conceptually
new methods. Hence, it is fair to say that the VS field is in constant flux, in terms of
theoretical developments and practical applications.
15.2 BASIC VIRTUAL SCREENING CONCEPTS
VS efforts require specific input information. In principle, this can be the three-
dimensional structure of the target protein or knowledge of already known active
compounds. Accordingly, one generally distinguishes between structure-based VS
(SBVS) [14,15] and ligand-based VS (LBVS) [16,17].
15.2.1 Structure- and Ligand-Based Virtual Screening
SBVS makes explicit use of target structures as screening templates, whereas LBVS
extrapolates from known active compounds to identify structurally diverse molecules
