Biomedical Engineering Reference
In-Depth Information
16.8 CONCLUDING REMARKS
In this chapter, activity landscape representations and activity cliffs, their most promi-
nent features, have been discussed in the context of molecular similarity and large-
scale SAR analysis. The organization and contents of this contribution partly follow
two recent perspective articles that have presented thorough reviews of the activity
landscape and activity cliff concepts [1,55].
In combination with numerical SAR analysis functions, activity landscape models
have become popular because they enable the extraction of SAR information from
large and heterogeneous compound data sets and the visualization of global and local
SAR features. A variety of activity landscape designs has already been introduced that
are often complementary in their use, such as global and compound-centric landscape
views. Moreover, the activity landscape concept is being extended in different ways:
for example, by exploring multitarget activity space.
A generally important feature of activity landscapes is their graphical nature. The
role of SAR visualization techniques for making computational results accessible
to medicinal chemists should not be underestimated. Graphical SAR exploration
is supported in different ways. For example, in network-based activity landscapes,
nodes (and sometimes edges) are often associated directly with corresponding com-
pound (or fragment) structures, which provides a basis for interactive SAR analysis
and inductive chemical interpretation of SAR networks. A characteristic feature of
activity landscape design is that it forms an interface between chemoinformatics and
medicinal chemistry. For further progress in this research area, including the devel-
opment of novel computational approaches, and for a broader acceptance of activity
landscape modeling among practicing medicinal chemists, it is also important that
a number of activity landscape methods have already been made publicly available,
including, for example, SALI graphs and various landscape tools implemented in the
SARANEA software environment [56].
The design of activity landscapes and consistent analysis of SARs critically depend
on a systematic assessment and integration of compound similarity and potency rela-
tionships. Thus, the concept of molecular similarity naturally plays a central role
in activity landscape modeling. As discussed, the dependence of activity landscape
topologies and SAR features on a chosen molecular representation and (to a lesser
extent) similarity measure is the most significant caveat for landscape design and
interpretation. Therefore, the exploration of alternative and intuitive ways to account
systematically for molecular similarity, including, for example, the application of hier-
archical structural and substructure relationships, should merit further investigation.
In addition, activity landscape analysis is also affected by the choice of experimental
measurements and their variability, which also requires careful consideration.
Despite these inherent limitations, modeling of activity landscapes provides sig-
nificant opportunities for medicinal chemistry. The data-oriented nature of activity
landscape views usually provides immediate access to SAR information available
in large data sets and makes it possible to focus the analysis on compound sub-
sets representing interesting local SARs. It is also straightforward to identify areas
that have been explored thoroughly without revealing promising SAR trends (and to
