Biomedical Engineering Reference
In-Depth Information
However, other links between activity cliffs and SAR continuity can also be observed.
This has been revealed by searching for SAR pathways, which represent a prede-
fined SARmodel designed to mine NSG representations systematically [51,52]. Such
pathways are formed by series of pairwise structurally similar compounds that follow
a potency gradient. The underlying SAR model prioritizes pathways that consist of
many overlapping pairs of similar compounds, span a large potency interval between
the start and end compounds, and ideally follow linear potency gradient with small
potency increases between neighboring compounds. Because these pathways consist
of pairwise similar compounds, “scaffold hops” might occur along a path. Accord-
ingly, SAR pathways are designed to navigate regions of local SAR continuity in
activity landscapes. In systematic NSG mining, SAR pathways have often been iden-
tified that directly connect regions of SAR continuity to prominent activity cliffs
[51,52], thereby establishing a close connection between SAR continuity and dis-
continuity. These insights have been further corroborated and extended by results
of SAR-oriented particle swarm optimization [53,54]. An optimization strategy was
devised to identify activity cliffs formed by compounds proximal to initialized particle
positions and to select other compounds from the immediate structural neighborhood
of prominent activity cliffs that form continuous local SARs [54]. In seven of 32
diverse compound data sets, subsets of at least five compounds were identified that
represented SAR continuity and had structural neighbors forming large-magnitude
activity cliffs. These findings revealed the presence of highly heterogeneous SAR
microenvironments and provided further evidence for close links between SAR con-
tinuity and discontinuity.
16.7.2
Information Extraction from Activity Cliffs
The inspection of isolated activity cliffs in compound data sets reveals individual
chemical changes leading to large potency differences between pairs of compounds.
This information alone might not be sufficient to aid in the design of better com-
pounds. The interpretation of activity cliffs becomes more informative if they are
viewed in the context of other SAR trends and if the structural neighborhood of cliffs
is analyzed. Moreover, because activity cliffs often do not occur in isolation but in
a coordinated manner involving multiple compounds (see above), a search for cliffs
might initially be better directed toward ridge-like structures from which more SAR
information can be deduced than from isolated cliffs.
Because most compound data sets analyzed thus far were found to contain mul-
tiple activity cliffs, often in different local arrangements, the investigation of new
compound series for established targets should involve mining of already known
compound activity data. Together with prior knowledge of activity cliff-forming
molecular building blocks, this might help to prioritize substitution patterns for
chemical exploration in different structural environments. Finally, for the chemical
interpretation of activity cliff information, alternative activity landscape represen-
tations should be taken into account, including those that complement or replace
calculated similarity values with defined substructure relationships.
