Biomedical Engineering Reference
In-Depth Information
FIGURE 10.8
Distribution of compounds in the top 10 most populated scaffolds found in
1490 approved drugs, a collection of 2472 natural products, 6623 synthetic diverse compounds
obtained from academic groups using techniques such as DOS, and 6152 compounds from
a commercial vendor library. The last three data sets were obtained from the work reported
by Clemons et al. [85] (see also Figure 10.5). Scaled Shannon entropy (SSE) values close to
1.0 indicate that the compounds are more evenly distributed in the scaffolds, suggesting large
scaffold diversity. Smaller SSE values denote that most of the compounds are distributed in
fewer scaffolds, indicating lower diversity.
chemical space with structure representation appears visually as in Figure 10.9, which
shows a comparison of the chemical space of nine compound libraries with 250 com-
pounds each. The compound databases are five in-house combinatorial libraries (red,
yellow, gray, black, and cyan), natural products from a commercial provider (green),
approved drugs (blue), synthetic commercial compounds (magenta), and compounds
targeted to adenosine receptors (orange). Each visual representation of the chemical
space was obtained by PCA of the similarity matrix computed using the Tanimoto
similarity function and two different fingerprint representations: MACCS keys [128]
and typed graph distance (TGD) [129] fingerprints. MACCS keys are a predefined set
of structural keys; and TGD fingerprints are graph distances employing the following
atom typing: donor, acceptor polar, anion, cation, and hydrophobe. The first three PCs
account for 85.9% (MACCS) and 90.3% (TGD) of the variance, respectively. Fig-
ure 10.9 clearly illustrates the significant changes in the neighborhood relationships
of the databases when different fingerprint representations are employed. Of course,
this dependence may have profound effects not only on the visual representation of
chemical spaces, as shown in Figure 10.9, but also on similarity searching and mining
of the chemical space. Other examples of the visualization of chemical space with
fingerprint representations are presented elsewhere [126].
