Activity cliffs/landscapes

Activity cliff, Structure-activity similarity maps, Structure-activity landscape index, Structure-activity... 

Structure-activity similarity (SAS) maps, first described by Shanmugasundaram and Maggiora (35), are pairwise plots of the structure similarity against the activity similarity. The resultant plot can be divided into four quadrants, allowing one to identify molecules characteristic of one of four possible behaviors smooth regions of the SAR space (rough), activity cliffs, nondescript (i.e., low structural similarity and low activity similarity), and scaffold hops (low structural similarity but high activity similarity). Recently, SAS maps have been extended to take into account multiple descriptor representations (two and three dimensions) (36, 37). In addition to SAS maps, other pairwise metrics to characterize and visualize SAR landscapes have been developed such as the structure-activity landscape index (SALI) (38) and the structure-activity index (SARI) (39). 

NSG that connect regions of low and high SAR discontinuity). Such SAR pathways represent a set of compounds that when ordered appropriately exhibit a continuous series of SAR changes. While network-based analyses of landscapes have seen much activity, an alternative visualization approach described by Seebeck et al. (42) abstracted the idea of the SALI metric and extended it to include the receptor. Using this technique they were able to highlight specific regions within protein-binding sites that are most likely to lead to activity cliffs. 

The concept of activity cliffs and the landscape paradigm have also been applied to R-groups, where an R-cliff occurs when a pair of compounds differs in a single R-group. This is clearly a specialization of the activity cliff concept, placing this type of analysis in the context of analogue series derived via R-group decompositions (43, 44). 

Medina-Franco JL, Martinez-Mayorga K, Bender A et al (2009) Characterization of activity landscapes using 2D and 3D similarity methods consensus activity cliffs. J Chem Inf Model 49(2) 477-491... 

Guha R, Van Drie JH (2008) The structure-activity landscape index identifying and quantifying activity-cliffs. J Chem Inf Model 48(3) 646-658... 

In systematic SAR analysis, molecular structure and similarity need to be represented and related to each other in a measurable form. Just like any molecular similarity approach, SAR analysis critically depends on molecular representations and the way similarity is measured. The nature of the chemical space representation determines the positions of the molecules in space and thus ultimately the shape of the activity landscape. Hence, SARs may differ considerably when changing chemical space and molecular representations. In this context, it becomes clear that one must discriminate between SAR features that reflect the fundamental nature of the underlying molecular structures as opposed to SAR features that are merely an artifact of the chosen chemical space representation. Consequently, activity cliffs can be viewed as either fundamental or descriptor- and metrics-dependent. The latter occur as a consequence of an inappropriate molecular representation or similarity metrics and can be smoothed out by choosing a more suitable representation, e.g., by considering activity-relevant physicochemical properties. By contrast, activity cliffs fundamental to the underlying SARs cannot be circumvented by changing the reference space. In this situation, molecules that should be recognized as... 

A quantification of the concept of presence of activity cliffs was proposed in terms of the SAL Index (or Structure-Activity Landscape Index), which for a pair of compounds is defined as [Maggiora, 2006 Guha and Ven Drie, 2008]... 

The introduction of a structure-activity landscape index (SALI), as proposed by Guha and Van Drie [51], helps to identify and quantify activity cliffs. This SALI is defined as follows ... 

The systematic exploration of activity cliffs from various representations of activity landscapes is summarized in an interesting article by Stumpfe and Bajorath [79]. These representations could include either distinct chemical transformations (MMPs) or similarity-based relationships between molecules. In a classical medicinal chemistry approach, activity cliffs are extracted from R-group tables, as shown in Figure 10.3 for 3-oxybenzamides as factor Xa inhibitors [58, 59]. This has to be repeated for each scaffold of interest with a predefined view on informative attachment points at the molecular core. Hence this approach is feasible for smaller datasets with only a few informative attachment points for SAR investigation, typically a narrow series in lead optimization. 

The SAR index (SARI) provides a composite score of individual SAR continuity and discontinuity functions, which have been introduced by Peltason et al. [83,47]. However, in order to identify activity cliffs, local scoring functions focusing directly on SAR discontinuity in a local environment around a chemical structure of interest are more informative than global activity landscape overviews. Regions of SAR discontinuity provide a lot of information on the SAR of a particular chemical series for deriving a pharmacophore hypothesis. For this case, the compound discontinuity score was introduced [80]. This variation of the SARI formalism is aiming to quantify individual molecular contributions to local SAR discontinuity. Typically, the... 

Peltason L, Iyer P, Bajorath J. Rationalizing three-dimensional activity landscapes and the influence of molecular representations on landscape topology and the formation of activity cliffs. J Chem Inf Model 2010 50 1021-1033. 

Dimova D, Wawer M, Wassermann AM, et al. Design of multi-target activity landscapes that capture hierarchical activity cliff distributions. J Chem Inf Model 2011 51 256-288. 

The concept of data fusion has also been extended to activity landscape modeling." It is well known that activity landscapes will be largely influenced by the choice of the molecular representation that is used to define the chemical space. In an effort to address this issue, multiple structural representations are combined using data fusion to derive a consensus model of activity landscapes and identify consensus activity cliffs [143-145]. Consensus models are designed to prioritize the SAR analysis of activity cliffs and other consistent regions in the activity landscape that are captured by several structure representations. They are not meant to be a means for eliminating data by disregarding, for example, true activity cliffs that are not identified by some structure representations. 

The concept of activity landscapes can also be extended to property landscapes where any set of measurable molecular properties can be added as another dimension to the chemical space of a compound dataset [146], In line with the concept of activity cliffs, odor cliffs [181] and flavor cliffs [182] have been recently described. 

As will be discussed in more detail in Section 15.6.3, structure-activity similarity (S AS) maps also afford a means not only for characterizing activity cliffs but also the entire activity landscapes within which they are embedded. 

See Section 15.6 for a more detailed discussion of activity landscapes and activity cliffs. The chemical space plots are constructed by carrying out principle component analysis (PCA) on the 2250 x 2250-dimensional similarity matrix associated with the complete compound dataset. Although, similarity matrices are not equivalent to the data matrices typically used in PCA, they can be used for this purpose [41]. Each data point (i.e., molecule) is then plotted with respect to the three PCs with the greatest variance. 

The paper by Tanaka et al. [168] investigates small world phenomena in several libraries obtained directly from the ZINC DB [184] and from virtual libraries constructed from structurally diverse fragments. By contrast, the paper of Kiein and Sukumar [169] undertakes a much more comprehensive analysis based on a nnmber of different sets of CS descriptors applied not only to CSs but also to their subspaces associated with activity cliffs. A recent paper from Bajorath s group [172] also addresses subnetworks associated with activity cliffs. Obviously, these analyses can be extended to other landscape features such as similarity cliffs (see Sect 1.2.4). 

---