Biomedical Engineering Reference
In-Depth Information
that emphasize SAR hotspots. Thus, this compound-centric activity landscape design
represents a versatile data structure for local SAR analysis and SAR data mining.
Global and local SAR analysis using activity landscape models can also be com-
bined in an effective manner. This has been accomplished, for example, through
NSG-SPT analysis [30]. Here, the basic idea is to initially monitor a data set globally
using an NSG representation, select from it the most interesting local SAR environ-
ments (compound subsets), and then inspect them at high resolution in series of SPTs
where each subset compound is used once as a reference. NSG-SPT analysis has been
applied successfully to extract SAR information from a large and noisy phenotypic
screening data set with more than 13,000 potential antimalarial hits [30].
16.4.4.2 Chemical Neighborhood Graph
Another compound-centric activity
landscape representation is provided by a chemical neighborhood graph (CNG) [31],
shown in Figure 16.1. In this case, the neighborhood of a reference compound is
also defined as the subset of compounds that exceed a predefined similarity threshold
relative to the reference. However, the graphical analysis scheme is distinct from the
NSG or SPT designs. Different from molecular network-based activity landscapes,
the CNG does not contain edges and basically represents a set of potency-colored
nodes (rather than a formal graph). In CNGs, similarity relationships are captured
by arranging all compound nodes falling into a neighborhood on concentric circles
around the reference. Each circle represents a range of similarity values, and increas-
ing radii of circles reflect decreasing similarity to the central node. CNGs provide a
detailed view of the potency distribution within organized (and also often overlap-
ping) chemical neighborhoods and are ranked by quantifying similarity and potency
value distributions that are indicative of interpretable SAR information [31].
16.4.4.3 Representation of Analog Series
Analog series can also be analyzed in
detail in “mini-landscape” formats, which might be regarded as a special case of
compound-centric activity landscapes. In medicinal chemistry, the standard way to
represent analog series is the use of R-group tables. These tables contain the invariant
structural core of a series with labeled substitution sites and report R-groups available
at each site together with the potency of the corresponding analogs. As an extension
of R-group tables, different versions of SAR maps have been introduced [32,33] that
report analog series after standard R-group decomposition in a matrix format where
each cell represents a unique combination of R-groups applying a potency-based
color code.
Going beyond the conventional R-group table format and its extensions, hierarchi-
cal tree-like local activity landscape representations have been introduced specifically
for analyzing analog series [34,35]. These combinatorial analog graphs (CAGs) sys-
tematically organize analog series on the basis of substitution site combinations
and calculate discontinuity scores for (overlapping) compound subsets sharing the
same substitution site or combinations of sites. Because the data structure is applied
exclusively to analyze series of analogs, whole-molecular similarity values for dis-
continuity score calculations are best replaced by local similarity measures such as
R-group-based pharmacophore edit distances [35]. CAGs reveal substitution sites
