Compound Activity Classes Research Articles

Large-Scale Predictions of Compound Potency with Original and Modified Activity Classes Reveal General Prediction Characteristics and Intrinsic Limitations of Conventional Benchmarking Calculations.

Predicting compound potency is a major task in computational medicinal chemistry, for which machine learning is often applied. This study systematically predicted compound potency values for 367 target-based compound activity classes from medicinal chemistry using a preferred machine learning approach and simple control methods. The predictions produced unexpectedly similar results for different classes and comparably high accuracy for machine learning and simple control models. Based on these findings, the influence of different data set modifications on relative prediction accuracies was explored, including potency range balancing, removal of nearest neighbors, and analog series-based compound partitioning. The predictions were surprisingly resistant to these modifications, leading to only small error margin increases. These findings also show that conventional benchmark settings are unsuitable for directly comparing potency prediction methods.

Large-scale prediction of activity cliffs using machine and deep learning methods of increasing complexity

Activity cliffs (AC) are formed by pairs of structural analogues that are active against the same target but have a large difference in potency. While much of our knowledge about ACs has originated from the analysis and comparison of compounds and activity data, several studies have reported AC predictions over the past decade. Different from typical compound classification tasks, AC predictions must be carried out at the level of compound pairs representing ACs or nonACs. Most AC predictions reported so far have focused on individual methods or comparisons of two or three approaches and only investigated a few compound activity classes (from 2 to 10). Although promising prediction accuracy has been reported in most cases, different system set-ups, AC definitions, methods, and calculation conditions were used, precluding direct comparisons of these studies. Therefore, we have carried out a large-scale AC prediction campaign across 100 activity classes comparing machine learning methods of greatly varying complexity, ranging from pair-based nearest neighbor classifiers and decision tree or kernel methods to deep neural networks. The results of our systematic predictions revealed the level of accuracy that can be expected for AC predictions across many different compound classes. In addition, prediction accuracy did not scale with methodological complexity but was significantly influenced by memorization of compounds shared by different ACs or nonACs. In many instances, limited training data were sufficient for building accurate models using different methods and there was no detectable advantage of deep learning over simpler approaches for AC prediction. On a global scale, support vector machine models performed best, by only small margins compared to others including simple nearest neighbor classifiers.Graphical

Prediction of activity cliffs on the basis of images using convolutional neural networks

An activity cliff (AC) is formed by a pair of structurally similar compounds with a large difference in potency. Accordingly, ACs reveal structure–activity relationship (SAR) discontinuity and provide SAR information for compound optimization. Herein, we have investigated the question if ACs could be predicted from image data. Therefore, pairs of structural analogs were extracted from different compound activity classes that formed or did not form ACs. From these compound pairs, consistently formatted images were generated. Image sets were used to train and test convolutional neural network (CNN) models to systematically distinguish between ACs and non-ACs. The CNN models were found to predict ACs with overall high accuracy, as assessed using alternative performance measures, hence establishing proof-of-principle. Moreover, gradient weights from convolutional layers were mapped to test compounds and identified characteristic structural features that contributed to successful predictions. Weight-based feature visualization revealed the ability of CNN models to learn chemistry from images at a high level of resolution and aided in the interpretation of model decisions with intrinsic black box character.

Activity landscape image analysis using convolutional neural networks

Activity landscapes (ALs) are graphical representations that combine compound similarity and activity data. ALs are constructed for visualizing local and global structure–activity relationships (SARs) contained in compound data sets. Three-dimensional (3D) ALs are reminiscent of geographical maps where differences in landscape topology mirror different SAR characteristics. 3D AL models can be stored as differently formatted images and are thus amenable to image analysis approaches, which have thus far not been considered in the context of graphical SAR analysis. In this proof-of-concept study, 3D ALs were constructed for a variety of compound activity classes and 3D AL image variants of varying topology and information content were generated and classified. To these ends, convolutional neural networks (CNNs) were initially applied to images of original 3D AL models with color-coding reflecting compound potency information that were taken from different viewpoints. Images of 3D AL models were transformed into variants from which one-dimensional features were extracted. Other machine learning approaches including support vector machine (SVM) and random forest (RF) algorithms were applied to derive models on the basis of such features. In addition, SVM and RF models were trained using other features obtained from images through edge filtering. Machine learning was able to accurately distinguish between 3D AL image variants with different topology and information content. Overall, CNNs which directly learned feature representations from 3D AL images achieved highest classification accuracy. Predictive performance for CNN, SVM, and RF models was highest for image variants emphasizing topological elevation. In addition, SVM models trained on rudimentary images from edge filtering classified such images with high accuracy, which further supported the critical role of altitude-dependent topological features for image analysis and predictions. Taken together, the findings of our proof-of-concept investigation indicate that image analysis has considerable potential for graphical SAR exploration to systematically infer different SAR characteristics from topological features of 3D ALs.

Advances in exploring activity cliffs

The activity cliff (AC) concept is of comparable relevance for medicinal chemistry and chemoinformatics. An AC is defined as a pair of structurally similar compounds with a large potency difference against a given target. In medicinal chemistry, ACs are of interest because they reveal small chemical changes with large potency effects, a concept referred to as structure–activity relationship (SAR) discontinuity. Computationally, ACs can be systematically identified, going far beyond individual compound series considered during lead optimization. Large-scale analysis of ACs has revealed characteristic features across many different compound activity classes. The way in which the molecular similarity and potency difference criteria have been addressed for defining ACs distinguishes between different generations of ACs and mirrors the evolution of the AC concept. We discuss different stages of this evolutionary path and highlight recent advances in AC research.

Large-Scale Comparison of Alternative Similarity Search Strategies with Varying Chemical Information Contents.

Similarity searching (SS) is a core approach in computational compound screening and has a long tradition in pharmaceutical research. Over the years, different approaches have been introduced to increase the information content of search calculations and optimize the ability to detect compounds having similar activity. We present a large-scale comparison of distinct search strategies on more than 600 qualifying compound activity classes. Challenging test cases for SS were identified and used to evaluate different ways to further improve search performance, which provided a differentiated view of alternative search strategies and their relative performance. It was found that search results could not only be improved by increasing compound input information but also by focusing similarity calculations on database compounds. In the presence of multiple active reference compounds, asymmetric SS with high weights on chemical features of target compounds emerged as an overall preferred approach across many different activity classes. These findings have implications for practical virtual screening applications.

Rationalizing the Formation of Activity Cliffs in Different Compound Data Sets.

Activity cliffs are formed by structurally analogous compounds with large potency variations and are highly relevant for the exploration of discontinuous structure–activity relationships and compound optimization. So far, activity cliffs have mostly been studied on a case-by-case basis or assessed by global statistical analysis. Different from previous investigations, we report a large-scale analysis of activity cliff formation with a strong focus on individual compound activity classes (target sets). Compound potency distributions were systematically analyzed and categorized, and structural relationships were dissected and visualized on a per-set basis. Our study uncovered target set-dependent interplay of potency distributions and structural relationships and revealed the presence of activity cliffs and origins of cliff formation in different structure–activity relationship environments.

Exploring differential evolution for inverse QSAR analysis

Inverse quantitative structure-activity relationship (QSAR) modeling encompasses the generation of compound structures from values of descriptors corresponding to high activity predicted with a given QSAR model. Structure generation proceeds from descriptor coordinates optimized for activity prediction. Herein, we concentrate on the first phase of the inverse QSAR process and introduce a new methodology for coordinate optimization, termed differential evolution (DE), that originated from computer science and engineering. Using simulation and compound activity data, we demonstrate that DE in combination with support vector regression (SVR) yields effective and robust predictions of optimized coordinates satisfying model constraints and requirements. For different compound activity classes, optimized coordinates are obtained that exclusively map to regions of high activity in feature space, represent novel positions for structure generation, and are chemically meaningful.

Exploring differential evolution for inverse QSAR analysis

Inverse quantitative structure-activity relationship (QSAR) modeling encompasses the generation of compound structures from values of descriptors corresponding to high activity predicted with a given QSAR model. Structure generation proceeds from descriptor coordinates optimized for activity prediction. Herein, we concentrate on the first phase of the inverse QSAR process and introduce a new methodology for coordinate optimization, termed differential evolution (DE), that originated from computer science and engineering. Using simulation and compound activity data, we demonstrate that DE in combination with support vector regression (SVR) yields effective and robust predictions of optimized coordinates satisfying model constraints and requirements. For different compound activity classes, optimized coordinates are obtained that exclusively map to regions of high activity in feature space, represent novel positions for structure generation, and are chemically meaningful.

Exploring differential evolution for inverse QSAR analysis.

Inverse quantitative structure-activity relationship (QSAR) modeling encompasses the generation of compound structures from values of descriptors corresponding to high activity predicted with a given QSAR model. Structure generation proceeds from descriptor coordinates optimized for activity prediction. Herein, we concentrate on the first phase of the inverse QSAR process and introduce a new methodology for coordinate optimization, termed differential evolution (DE), that originated from computer science and engineering. Using simulation and compound activity data, we demonstrate that DE in combination with support vector regression (SVR) yields effective and robust predictions of optimized coordinates satisfying model constraints and requirements. For different compound activity classes, optimized coordinates are obtained that exclusively map to regions of high activity in feature space, represent novel positions for structure generation, and are chemically meaningful.

Is scaffold hopping a reliable indicator for the ability of computational methods to identify structurally diverse active compounds?

Computational scaffold hopping aims to identify core structure replacements in active compounds. To evaluate scaffold hopping potential from a principal point of view, regardless of the computational methods that are applied, a global analysis of conventional scaffolds in analog series from compound activity classes was carried out. The majority of analog series was found to contain multiple scaffolds, thus enabling the detection of intra-series scaffold hops among closely related compounds. More than 1000 activity classes were found to contain increasing proportions of multi-scaffold analog series. Thus, using such activity classes for scaffold hopping analysis is likely to overestimate the scaffold hopping (core structure replacement) potential of computational methods, due to an abundance of artificial scaffold hops that are possible within analog series.

Influence of Varying Training Set Composition and Size on Support Vector Machine-Based Prediction of Active Compounds

Support vector machine (SVM) modeling is one of the most popular machine learning approaches in chemoinformatics and drug design. The influence of training set composition and size on predictions currently is an underinvestigated issue in SVM modeling. In this study, we have derived SVM classification and ranking models for a variety of compound activity classes under systematic variation of the number of positive and negative training examples. With increasing numbers of negative training compounds, SVM classification calculations became increasingly accurate and stable. However, this was only the case if a required threshold of positive training examples was also reached. In addition, consideration of class weights and optimization of cost factors substantially aided in balancing the calculations for increasing numbers of negative training examples. Taken together, the results of our analysis have practical implications for SVM learning and the prediction of active compounds. For all compound classes under study, top recall performance and independence of compound recall of training set composition was achieved when 250–500 active and 500–1000 randomly selected inactive training instances were used. However, as long as ∼50 known active compounds were available for training, increasing numbers of 500–1000 randomly selected negative training examples significantly improved model performance and gave very similar results for different training sets.

Target enhanced 2D similarity search by using explicit biological activity annotations and profiles

BackgroundThe enriched biological activity information of compounds in large and freely-accessible chemical databases like the PubChem Bioassay Database has become a powerful research resource for the scientific research community. Currently, 2D fingerprint based conventional similarity search (CSS) is the most common widely used approach for database screening, but it does not typically incorporate the relative importance of fingerprint bits to biological activity.ResultsIn this study, a large-scale similarity search investigation has been carried out on 208 well-defined compound activity classes extracted from PubChem Bioassay Database. An analysis was performed to compare the search performance of three types of 2D similarity search approaches: 2D fingerprint based conventional similarity search approach (CSS), iterative similarity search approach with multiple active compounds as references (ISS), and fingerprint based iterative similarity search with classification (ISC), which can be regarded as the combination of iterative similarity search with active references and a reversed iterative similarity search with inactive references. Compared to the search results returned by CSS, ISS improves recall but not precision. Although ISC causes the false rejection of active hits, it improves the precision with statistical significance, and outperforms both ISS and CSS. In a second part of this study, we introduce the profile concept into the three types of searches. We find that the profile based non-iterative search can significantly improve the search performance by increasing the recall rate. We also find that profile based ISS (PBISS) and profile based ISC (PBISC) significantly decreases ISS search time without sacrificing search performance.ConclusionsOn the basis of our large-scale investigation directed against a wide spectrum of pharmaceutical targets, we conclude that ISC and ISS searches perform better than 2D fingerprint similarity searching and that profile based versions of these algorithms do nearly as well in less time. We also suggest that the profile version of the iterative similarity searches are both better performing and potentially quicker than the standard algorithm.Electronic supplementary materialThe online version of this article (doi:10.1186/s13321-015-0103-5) contains supplementary material, which is available to authorized users.

Activity Cliff Networks for Medicinal Chemistry

Network representations are widely used in bioinformatics but have only been little explored in chemistry. Thus far, only a few attempts have been made to generate and analyze compound networks. Among these are the first activity cliff networks. In medicinal chemistry, activity cliffs are focal points of structure-activity relationships (SAR) analysis. Activity cliffs have generally been defined as pairs of structurally similar or analogous active compounds that have a large difference in potency against their target. However, most activity cliffs are not formed in isolation but in a coordinated manner involving multiple highly and weakly potent compounds. Recently, a comprehensive activity cliff network has been generated for current public domain bioactive compounds, hence providing a first global view of activity cliff formation. The design of activity cliff networks is discussed herein. From the global activity cliff network, local networks can be extracted for individual compound activity classes that provide graphical access to high-level SAR information for compound optimization efforts.

Systematic Assessment of Compound Series with SAR Transfer Potential

Compound series with different core structures that contain pairs of analogs with corresponding substitution patterns and similar activity represent structure-activity relationship (SAR) transfer events. On the basis of the matched molecular pair (MMP) formalism and linear regression analysis of compound potencies, a general approach is introduced for the identification of SAR transfer series (SAR-TS) and SAR-TS with regular potency progression (SAR-TS-RP). We have systematically extracted such series from public domain compound data and analyzed their size distribution and structural characteristics. More than 900 SAR-TS and 500 SAR-TS-RP with high-confidence potency annotations were identified in various compound activity classes. These series provide a substantial knowledge base for the analysis and prediction of SAR transfer and are made publicly available.

Target Family-Directed Exploration of Scaffolds with Different SAR Profiles

The scaffold concept is widely applied in chemoinformatics and medicinal chemistry to organize bioactive compounds according to common core structures or associate compound classes with specific biological activities. A variety of scaffold analyses have been carried out to derive statistics for scaffold distributions, generate structural organization schemes, or identify scaffolds that preferentially occur in given compound activity classes. Herein we further extend scaffold analysis by identifying scaffolds that display defined SAR profiles consisting of multiple properties. A structural relationship-based scaffold network has been designed as the basic data structure underlying our analysis. From network representations of scaffolds extracted from compounds active against 32 different target families, scaffolds with different SAR profiles have been extracted on the basis of decision trees that capture structural and functional characteristics of scaffolds in different ways. More than 600 scaffolds and 100 scaffold clusters were assigned to 10 SAR profiles. These scaffold sets represent different activity and target selectivity profiles and are provided for further SAR investigations including, for example, the exploration of alternative analog series for a given target of target family or the design of novel compounds on the basis of scaffold(s) with desired SAR profiles.

How Do 2D Fingerprints Detect Structurally Diverse Active Compounds? Revealing Compound Subset-Specific Fingerprint Features through Systematic Selection

In independent studies it has previously been demonstrated that two-dimensional (2D) fingerprints have scaffold hopping ability in virtual screening, although these descriptors primarily emphasize structural and/or topological resemblance of reference and database compounds. However, the mechanism by which such fingerprints enrich structurally diverse molecules in database selection sets is currently little understood. In order to address this question, similarity search calculations on 120 compound activity classes of varying structural diversity were carried out using atom environment fingerprints. Two feature selection methods, Kullback-Leibler divergence and gain ratio analysis, were applied to systematically reduce these fingerprints and generate alternative versions for searching. Gain ratio is a feature selection method from information theory that has thus far not been considered in fingerprint analysis. However, it is shown here to be an effective fingerprint feature selection approach. Following comparative feature selection and similarity searching, the compound recall characteristics of original and reduced fingerprint versions were analyzed in detail. Small sets of fingerprint features were found to distinguish subsets of active compounds from other database molecules. The compound recall of fingerprint similarity searching often resulted from a cumulative detection of distinct compound subsets by different fingerprint features, which provided a rationale for the scaffold hopping potential of these 2D fingerprints.

Large-Scale Similarity Search Profiling of ChEMBL Compound Data Sets

A large-scale similarity search investigation has been carried out on 266 well-defined compound activity classes extracted from the ChEMBL database. The analysis was performed using two widely applied two-dimensional (2D) fingerprints that mark opposite ends of the current performance spectrum of these types of fingerprints, i.e., MACCS structural keys and the extended connectivity fingerprint with bond diameter four (ECFP4). For each fingerprint, three nearest neighbor search strategies were applied. On the basis of these search calculations, a similarity search profile of the ChEMBL database was generated. Overall, the fingerprint search campaign was surprisingly successful. In 203 of 266 test cases (∼76%), a compound recovery rate of at least 50% was observed with at least the better performing fingerprint and one search strategy. The similarity search profile also revealed several general trends. For example, fingerprint searching was often characterized by an early enrichment of active compounds in database selection sets. In addition, compound activity classes have been categorized according to different similarity search performance levels, which helps to put the results of benchmark calculations into perspective. Therefore, a compendium of activity classes falling into different search performance categories is provided. On the basis of our large-scale investigation, the performance range of state-of-the-art 2D fingerprinting has been delineated for compound data sets directed against a wide spectrum of pharmaceutical targets.

Identification of Descriptors Capturing Compound Class-Specific Features by Mutual Information Analysis

The identification of molecular descriptors that contain compound class-specific information is of high relevance in chemoinformatics. A generally applicable way to identify such descriptors is to determine and compare their information content in a given compound activity class and in large databases where the vast majority of compounds do not have the desired activity. For this purpose, the Shannon entropy concept from information theory can in principle be employed. However, previous adaptations of this concept for descriptor profiling are insufficient to select discriminatory descriptors for data sets that dramatically differ in size. Therefore, we introduce a methodology to reliably select such descriptors by transforming the previously introduced differential Shannon entropy formalism into mutual information analysis, another concept from information theory. The newly introduced approach is evaluated by descriptor ranking and correlation analysis on 168 compound activity classes.

Reduction and Recombination of Fingerprints of Different Design Increase Compound Recall and the Structural Diversity of Hits

We report an advanced 'hybrid fingerprint' design concept specifically for the purpose of scaffold hopping. The generation of hybrid fingerprints includes two major steps. In the 'fingerprint reduction' step, bit positions of different types of fingerprints (e.g. substructural and pharmacophore fingerprints) are ranked according to their statistical significance and ability to discriminate between specifically active compounds and database decoys. On the basis of bit ranking, subsets containing the most discriminatory bit positions are determined. In the subsequent 'fingerprint recombination' step, bit subsets from different fingerprints are combined to yield a new compound class-directed fingerprint representation for similarity searching. Here, we generate hybrids from multiple fingerprints and analyze their search performance in comparison with parental fingerprints on compound activity classes that exclusively consist of molecules with unique core structures and that exhibit different levels of intra-class structural diversity. Fingerprint reduction is found to be a critical component of hybrid design. The resulting compound class-directed hybrid fingerprints further increase the similarity search performance and scaffold hopping potential of their parental fingerprints. Thus, fingerprint reduction and recombination improve compound recall and increase the structural diversity of hits.