Preeti Iyer

Activity Landscapes, Information Theory, and Structure – Activity Relationships

Activity landscapes provide a comprehensive description of structure‐activity relationships (SARs). An information theoretic assessment of their features, namely, activity cliffs, similarity cliffs, smooth‐SAR, and featureless regions, is presented based on the probability of occurrence of these features. It is shown that activity cliffs provide highly informative SARs compared to smooth‐SAR regions, although the latter are the basis for most QSAR studies. This follows since small structural changes in the former are coupled with relatively large changes in activity, thus pinpointing specific structural features associated with the changes in activity. In contrast, Smooth‐SAR regions are typically associated with relatively small changes in both structure and activity. Surprisingly, similarity cliffs, which occur when both compounds in a compound‐pair have approximately equal activities but significantly different structures, are the most prevalent feature of activity landscapes. Hence, from an information theoretic point of view, they are the least informative landscape feature. Nevertheless, similarity cliffs do provide SAR information on potentially new active compound classes, and in that sense they are quite useful in drug discovery programs since they provide alternative possibilities should ADMET or other issues arise during the discovery and earlier preclinical development phases of drug research.

Assessing the Target Differentiation Potential of Imidazole-Based Protein Kinase Inhibitors

A library of 484 imidazole-based candidate inhibitors was tested against 24 protein kinases. The resulting activity data have been systematically analyzed to search for compounds that effectively differentiate between kinases. Six imidazole derivatives with high kinase differentiation potential were identified. Nearest neighbor analysis revealed the presence of close analogues with varying differentiation potential. Small structural modifications of active compounds were found to shift their inhibitory profiles toward kinases with different functions.

Advances in Computational Medicinal Chemistry: Matched Molecular Pair Analysis

Preclinical Research

Comparison of two- and three-dimensional activity landscape representations for different compound data sets

Modeling of activity landscapes provides a basis for the analysis of structure–activity relationships (SARs) in large compound data sets. Activity landscape models enable visual access to SAR features. Regardless of their specific details, these models generally have in common that they integrate molecular similarity and potency relationships between active compounds. Different two-dimensional (2D) landscape representations have been introduced and recently also the first detailed three-dimensional (3D) model. Herein we compare advanced 2D and 3D activity landscape models for compound data sets having different SAR character. Although the compared 2D and 3D representations are conceptually distinct, it is found that global SAR features of compound data sets can be equally well deduced from them. However, local SAR information is often captured in different ways by these representations. Since these 2D and 3D landscape modeling tools have been made freely available, the analysis also provides guidelines for how to best utilize these alternative landscape representations for practical SAR analysis.

Representation of Multi‐Target Activity Landscapes Through Target Pair‐Based Compound Encoding in Self‐Organizing Maps

Activity landscape representations provide access to structure‐activity relationships information in compound data sets. In general, activity landscape models integrate molecular similarity relationships with biological activity data. Typically, activity against a single target is monitored. However, for steadily increasing numbers of compounds, activity against multiple targets is reported, resulting in an opportunity, and often a need, to explore multi‐target structure‐activity relationships. It would be attractive to utilize activity landscape representations to aid in this process, but the design of activity landscapes for multiple targets is a complicated task. Only recently has a first multi‐target landscape model been introduced, consisting of an annotated compound network focused on the systematic detection of activity cliffs. Herein, we report a conceptually different multi‐target activity landscape design that is based on a 2D projection of chemical reference space using self‐organizing maps and encodes compounds as arrays of pair‐wise target activity relationships. In this context, we introduce the concept of discontinuity in multi‐target activity space. The well‐ordered activity landscape model highlights centers of discontinuity in activity space and is straightforward to interpret. It has been applied to analyze compound data sets with three, four, and five target annotations and identify multi‐target structure‐activity relationships determinants in analog series.

Conditional probabilities of activity landscape features for individual compounds.

Activity landscape representations aid in the analysis of structure-activity relationships (SARs) of large compound data sets. Landscapes are characterized by features with different SAR information content such as, for example, regions formed by structurally diverse compounds having similar activity or, alternatively, structurally similar compounds with large activity differences, so-called activity cliffs. Modeling of activity landscapes typically requires pairwise comparisons of molecular similarity and potency relationships of compounds in a data set. Consequently, landscape features are generally resolved at the level of compound pairs. Herein, we introduce a methodology to assign feature probabilities to individual compounds. This makes it possible to organize compounds comprising activity landscapes into well-defined SAR categories. Specifically, the calculation of conditional feature probabilities of active compounds provides a balanced and further refined view of activity landscapes with a focus on individual molecules.

Navigating High-Dimensional Activity Landscapes: Design and Application of the Ligand-Target Differentiation Map

The transformation of high-dimensional bioactivity spaces into activity landscape representations is as of yet an unsolved problem in computational medicinal chemistry. High-dimensional activity spaces result from the experimental evaluation of compound sets on large numbers of targets. We introduce a first concept to represent and navigate high-dimensional activity landscapes that is based on a data structure termed ligand-target differentiation (LTD) map. This approach is designed to reduce the complexity of high-dimensional bioactivity spaces and enable the identification and further analysis of compound subsets with interesting activity and structural relationships. Its utility has been demonstrated using a set of more than 1400 inhibitors with exact activity measurements for varying numbers of 172 kinases.

Exploring SAR Continuity in the Vicinity of Activity Cliffs

Activity cliffs are formed by structurally similar compounds with significant differences in potency and represent an extreme form of structure–activity relationships discontinuity. By contrast, regions of structure–activity relationships continuity in compound data sets result from the presence of structurally increasingly diverse compounds retaining similar activity. Previous studies have revealed that structure–activity relationships information extracted from large compound data sets is often heterogeneous in nature containing both continuous and discontinuous structure–activity relationships components. Structure–activity relationships discontinuity and continuity are often represented by different compound series, independent of each other. Here, we have searched different compound data sets for the presence of structure–activity relationships continuity within the vicinity of prominent activity cliffs. For this purpose, we have designed and implemented a computational approach utilizing particle swarm optimization to examine the structural neighborhood of activity cliffs for continuous structure–activity relationships components. Structure–activity relationships continuity in the structural neighborhood of activity cliffs was relatively rarely observed. However, in a number of cases, notable structure–activity relationships continuity was detected in the vicinity of prominent activity cliffs. Exemplary local structure–activity relationships environments displaying these characteristics were analyzed in detail. Thus, the structure–activity relationships environment of activity cliffs must not necessarily be discontinuous in nature, and local structure–activity relationships continuity and discontinuity can occur in a concerted manner in series of structurally related compounds.

Prediction of Individual Compounds Forming Activity Cliffs Using Emerging Chemical Patterns

Activity cliffs are formed by structurally similar or analogous compounds having large potency differences. In medicinal chemistry, pairs or groups of compounds forming activity cliffs are of interest for structure-activity relationship (SAR) analysis and compound optimization. Thus far, activity cliff assessment has mostly been descriptive, i.e., compound data sets and activity landscape representations have been searched for activity cliffs in the context of SAR analysis. Only recently, first attempts have also been made to depart from descriptive analysis and predict activity cliffs. This has been done by building computational models that distinguish compound pairs forming activity cliffs from non-cliff pairs. However, it is principally more challenging to predict single compounds that participate in activity cliffs. Here, we show that individual compounds having high or low potency can be accurately predicted to form activity cliffs on the basis of emerging chemical patterns.

Predicting potent compounds via model-based global optimization.

Finding potent compounds for a given target in silico can be viewed as a constraint global optimization problem. This requires the use of an optimization function for which evaluations might be costly. The major task is maximizing the function while minimizing the number of evaluation steps. To solve this problem, we propose a machine learning algorithm, which first builds a statistical QSAR-model of the SAR landscape and then uses the model to identify regions in compound space having a high probability to contain a highly potent compound. For this purpose, we devise the so-called expected potency improvement (EI) criterion to rank candidate compounds with respect to their likelihood to exhibit higher potency than the most active compound in the training data. Therefore, this approach significantly differs from a purely prediction-oriented classical QSAR model. The method is superior to a nearest neighbor approach as significantly fewer evaluation steps are needed to identify the most potent compound for the given target.