Biomedical Engineering Reference
In-Depth Information
be positioned onto this map using interpolation in terms of PCA score prediction. Fur-
ther details of this method are provided elsewhere [74]. Figure 10.3 clearly illustrates
that libraries obtained using the LOL method (see above) occupy different regions of
chemical space. ChemGPS-NPWeb has recently been used to compare the chemical
space of approved drugs with TCM and compounds derived from in-house combina-
torial libraries available in PubChem [77]. Several other examples of the comparison
of compound collections relevant in drug discovery are reviewed elsewhere [31,61].
10.4.2 Molecular Complexity
Molecular complexity has emerged as an attractive criterion to guide the selection of
chemical libraries to explore the currently neglected chemical space [78,79]. Several
measures of complexity have been proposed [80-83]. Using the fraction of saturated
carbons as a simple and intuitive measure of complexity, Lovering et al. showed that
more complex molecules have higher success rates in the drug discovery process [84].
The authors of that work also suggested that compounds with increased complexity,
as captured by a measure of carbon bond saturation, might increase selectivity.
The hypothesis of Lovering et al. was supported experimentally by Clemons et al.
[85], who screened three major types of compound collections—commercial com-
pounds, natural products, and synthetic molecules obtained from academic groups—
across 100 diverse proteins using techniques such as DOS. Figure 10.4 illustrates a
comparison of the complexity of the three collections using two measures employed
in that work: stereochemical complexity, calculated as the portion of carbon atoms
that are stereogenic (Figure 10.4a), and shape complexity, calculated as the ratio of
sp
3
-hydridized carbon atoms to total sp
3
- and sp
2
-hybridized carbons (Figure 10.4b).
The distributions were generated with the data reported by Clemons et al. [85]. The
box plots, for which a rich statistical interpretation is presented by Ritchie and Mac-
donald [86], clearly show the increased complexity of the natural products collection
over the commercial compounds and the set of diverse synthetic molecules. By con-
trast, the commercial compounds showed the lowest complexity. After screening
these three compound databases, which differed in molecular complexity, the authors
of that work concluded that increasing the content of sp
3
-hydridized and stereogenic
atoms relative to compounds from commercial sources enhanced the binding selectiv-
ity and frequency. The results of Clemons et al. are in agreement with the complexity
model of Hann, Leach, and co-workers [87,88].
The pivotal work of Lovering et al. [84] using the fraction of saturated carbons
to measure molecular complexity (or “aliphatic indicator of a molecule” as defined
initially by Yan and Gasteiger [89]) encouraged the use of this measure to survey the
complexity of different compound databases. Thus, analysis of the chemical com-
plexity of the compounds in the Molecular Libraries Small Molecule Repository
using the fraction of saturated carbons leads to the conclusion that natural products
and collections from academic and other research institutes are more complex than
libraries from commercial vendors [79]. Chen et al. demonstrated the increased com-
plexity of natural products over drugs, clinical candidates, and bioactive molecules
[90]. Walters et al. analyzed the distribution of the fraction of saturated carbons for
