Consensus Models of Activity Landscapes with Multiple Chemical, Conformer, and Property Representations

Abstract

We report consensus Structure-Activity Similarity (SAS) maps that address the dependence of activity landscapes on molecular representation. As a case study, we characterized the activity landscape of 54 compounds with activities against human cathepsin B (hCatB), human cathepsin L (hCatL), and Trypanosoma brucei cathepsin B (TbCatB). Starting from an initial set of 28 descriptors we selected ten representations that capture different aspects of the chemical structures. These included four 2D (MACCS keys, GpiDAPH3, pairwise, and radial fingerprints) and six 3D (4p and piDAPH4 fingerprints with each including three conformers) representations. Multiple conformers are used for the first time in consensus activity landscape modeling. The results emphasize the feasibility of identifying consensus data points that are consistently formed in different reference spaces generated with several fingerprint models, including multiple 3D conformers. Consensus data points are not meant to eliminate data, disregarding, for example, "true" activity cliffs that are not identified by some molecular representations. Instead, 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 molecular representations. Systematic description of the SARs of two targets give rise to the identification of pairs of compounds located in the same region of the activity landscape of hCatL and TbCatB suggesting similar mechanisms of action for the pairs involved. We also explored the relationship between property similarity and activity similarity and found that property similarities are suitable to characterize SARs. We also introduce the concept of structure-property-activity (SPA) similarity in SAR studies.

Activity cliffs in drug discovery: Dr Jekyll or Mr Hyde?

“…As discussed above, one approach to address such dependence is using MMPs or discrete R-group substitutions around a core scaffold. Another approach proposed by Medina-Franco and colleagues [25,[41][42][43] is using multiple representations [44] to derive a consensus activity landscape model [25]. Using this strategy, the SALI approach [22] was extended to compute consensus SALI values obtained over multiple, diverse and orthogonal 2D and 3D representations.…”

“…[ 43] www.drugdiscoverytoday.com 5 A new approach for the identification of structure-based activity cliffs (ISAC) by analyzing interaction energies of proteinligand complexes.…”

Scanning Structure–Activity Relationships with Structure–Activity Similarity and Related Maps: From Consensus Activity Cliffs to Selectivity Switches

“…Examples of thresholds used are an activity difference of one log unit; 27 two thresholds of one and two log units, 23 or the mean activity similarity of the data set. 29 Defining thresholds for structure similarity has the issue of defining "high similarity". Statistical approaches have been proposed to address this point.…”

“…For example, the large effect of the type of 2D and 3D fingerprints used to compute structure similarity on the distribution of the pairs of compounds in an SAS map has been shown for a number of data sets. 23,27,29 In light of this dependence and to reduce the effect of the activity landscape with structure representation, it was proposed the analysis of the SAR of data sets using a number of SAS maps, each generated using a different 2D and 3D structure fingerprints, and then derive general conclusions common to several representations. 27,33 This approach is based on the fact that there are not perfect representations or descriptors that capture all possible relevant aspects of the molecule structures.…”

“…35 To address this issue, one can compute 3D similarities calculated for multiple conformers. 29 As part of the further development of the SAS maps to analyze SARs, it was proposed using not only fingerprint-based representations but also molecular properties, e.g., drug-like properties, to compute property similarity using scaled Euclidean distance. 23 (4) where sd ij is the scaled distance, while min d ij and max d ij specify the range of the computed distances.…”