QSAR Methods Research Articles

CAESAR models for developmental toxicity

BackgroundThe new REACH legislation requires assessment of a large number of chemicals in the European market for several endpoints. Developmental toxicity is one of the most difficult endpoints to assess, on account of the complexity, length and costs of experiments. Following the encouragement of QSAR (in silico) methods provided in the REACH itself, the CAESAR project has developed several models.ResultsTwo QSAR models for developmental toxicity have been developed, using different statistical/mathematical methods. Both models performed well. The first makes a classification based on a random forest algorithm, while the second is based on an adaptive fuzzy partition algorithm. The first model has been implemented and inserted into the CAESAR on-line application, which is java-based software that allows everyone to freely use the models.ConclusionsThe CAESAR QSAR models have been developed with the aim to minimize false negatives in order to make them more usable for REACH. The CAESAR on-line application ensures that both industry and regulators can easily access and use the developmental toxicity model (as well as the models for the other four endpoints).

Pharmacophore identification and bioactivity prediction for triaminotriazine derivatives by electron conformational-genetic algorithm QSAR method

The electron conformational–genetic algorithm (EC–GA) method has been employed as a 4D-QSAR approach to reveal the pharmacophore (Pha) and to predict anticancer activity in the N-morpholino triaminotriazine derivatives. The electron conformational matrices of congruity (ECMCs) identified by electronic and structural parameters are constructed from data of conformational analysis and electronic structure calculation of molecules. Comparing the matrices, electron conformational submatrix of activity (ECSA, Pha) are revealed that are common for these compounds within a minimum tolerance. To predict the theoretical activity of training and test set and to select important variables for describing the activities, genetic algorithm and non-linear least square regression methods were applied. Regression coefficients were found 0.9708 for training and 0.9520 for test set.

A comprehensive structure–activity analysis of protein kinase B-alpha (Akt1) inhibitors

Protein kinase B (PKB, also known as Akt) belongs to the AGC subfamily of the protein kinase superfamily. Akt1 has been reported as a central player in regulation of metabolism, cell survival, motility, transcription and cell-cycle progression, among the signalling proteins that respond to a large variety of signals. In this study an attempt was made to understand structural requirements for Akt1 inhibition using conventional QSAR, k-nearest neighbour QSAR and novel GQSAR methods. With this intention, a wide variety of structurally diverse Akt1 inhibitors were collected from various literature reports. The conventional QSAR analyses revealed the key role of Baumann's alignment independent topological descriptors along with other descriptors such as the number of hydrogen bond acceptors, hydrogen bond donors, rotatable bonds and aromatic oxygen (SaaOcount) along with molecular branching (chi3Cluster), alkene carbon atom type (SdsCHE-index) in governing activity variation. Further, the GQSAR analyses show that chemical variations like presence of hetero-aromatic ring, flexibility, polar surface area and fragment length present in the hinge binding fragment (in the present case fragment D) are highly influential for achieving highly potent Akt1 inhibitors. In addition, this study resulted in a k-nearest neighbour classification model with three descriptors suggesting the key role of oxygen (SssOE-index) and aromatic carbon (SaaCHE-index and SaasCE-index) atoms electro-topological environment that differentiate molecules binding to Akt1 kinase or PH domain. The developed models are interpretable, with good statistical and predictive significance, and can be used for guiding ligand modification for the development of potential new Akt1 inhibitors.

Forecasting CYP2D6 and CYP3A4 Risk with a Global/Local Fusion Model of CYP450 Inhibition.

This work presents a method to utilize the ever-expanding corporate collections of CYP450 inhibition data to forecast the future risk of compounds not yet synthesized. The global/local fusion method differs from existing QSAR methods, in that each prediction is derived from a custom-built QSAR model, constructed on-the-fly, using a customized training set assembled for each prediction. It uses a consensus of global and local descriptor-based models along with pharmacophore-based fingerprint similarity to form a prediction and to assess the uncertainty of the prediction on a case-by-case basis. We also present a new forward prediction testing and validation scheme in which the corporate dataset is split chronologically, and predictions for a molecule are based on the pool of existing data available before the molecule is registered and tested. The validation accuracy of the CYP2D6 and CYP3A4 models approaches the underlying accuracy of the data, about 0.4 log IC50 units standard error (or nearly 70% r(2) correlation) for the most confident predictions, and extends to about 0.6 log IC50 units standard error (or under 30% r(2) correlation) for the least confident predictions. As a classification model for CYP2D6 and CYP3A4 activity, the validation accuracy is about 79% for predicted actives and 85% for predicted inactives, which is consistent with existing published models.

Applications of Bond‐Based 3D‐Chiral Quadratic Indices in QSAR Studies Related to Central Chirality Codification

AbstractThe concept of bond‐based quadratic indices is generalized to codify chemical structure information for chiral drugs, making use of a trigonometric 3D‐chirality correction factor. In order to evaluate the effectiveness of this novel approach in drug design, we have modeled several well‐known data sets. In particularly, Cramer's steroid data set has become a benchmark for the assessment of novel QSAR methods. This data set has been used by several researchers using 3D‐QSAR approaches. Therefore, it is selected by us for the shake of comparability. In addition, to evaluate the effectiveness of this novel approach in drug design, we model the angiotensin‐converting enzyme inhibitory activity of perindoprilate's σ‐stereoisomers combinatorial library, as well as codify information related to a pharmacological property, highly dependent on the molecular symmetry, of a set of seven pairs of chiral N‐alkylated 3‐(3‐hydroxyphenyl)‐piperidines, which bind σ‐receptors. The validation of this method is achieved by comparison with earlier publications applied to the same data sets. The nonstochastic and stochastic bond‐based 3D‐chiral quadratic indices appear to provide a rather interesting alternative to other more common 3D‐QSAR descriptors.

Multi‐Objective Feature Selection in QSAR Using a Machine Learning Approach

AbstractThe selection of descriptor subsets for QSAR/QSPR is a hard combinatorial problem that requires the evaluation of complex relationships in order to assess the relevance of the selected subsets. In this paper, we describe the main issues in applying descriptor selection for QSAR methods and propose a novel two‐phase methodology for this task. The first phase makes use of a multi‐objective evolutionary technique which yields interesting advantages compared to mono‐objective methods. The second phase complements the first one and it enables to refine and improve the confidence in the chosen subsets of descriptors. This methodology allows the selection of subsets when a large number of descriptors are involved and it is also suitable for linear and nonlinear QSAR/QSPR models. The proposed method was tested using three data sets with experimental values for blood‐brain barrier penetration, human intestinal absorption and hydrophobicity. Results reveal the capability of the method for achieving subsets of descriptors with a high predictive capacity and a low cardinality. Therefore, our proposal constitutes a new promising technique helpful for the development of QSAR/QSPR models.

Use of QSAR methods for predicting the chemiluminescent behaviour of organic compounds upon reaction with potassium permanganate in an acid medium

In previous work, molecular connectivity computations were successfully used to predict the chemiluminescent behaviour of organic compounds upon reaction with common strong oxidants and the native fluorescence too; both of them in a liquid phase. The obtained results were used to develop new analytical procedures to the given compounds. For the first time, connectivity methods were used for a purely analytical purpose. In this work, we went deeper into the knowledge of direct chemiluminescence processes by using molecular connectivity in the form of QSAR methods to predict the chemiluminescence intensity produced by reactions between organic compounds (pharmaceuticals mainly) and potassium permanganate in a liquid phase. The choice of this oxidant was dictated by its being the most active by far in producing chemiluminescence.We used discriminant analysis to examine the results for 63 substances the emission intensity of which upon reaction with acid potassium permanganate was experimentally measured in a continuous-flow manifold. Descriptors were chosen by applying stepwise linear dicriminant analysis (LDA) to Snedecor F-values, using the smallest Mahalanobis distance, the minimum error on the test set and the lowest value of the Wilks’lambda as sorting criterion.The theoretical predictions thus obtained were checked against the experimental results for a set of 16 compounds not used in the previous theoretical computations the chemiluminescent behaviour of which was also experimentally assessed. The result was a hit rate of 87.5% in the predictions.

CAUTION: Popular “Benchmark” Data Sets Do Not Distinguish the Merits of 3D QSAR Methods

The quality of 3D QSAR models obtained using extremely simple descriptors was examined for nine popular data sets, including the well-known set of 31 steroids, which for 20 years has been the standard for benchmarking 3D QSAR methods. The atomic numbers of atoms coinciding with vertices of the molecular alignment as well as binary descriptors indicating the occupancy of those vertices were compared to models obtained using SAMFA descriptors, which have been shown to yield models statistically indistinguishable from CoMFA. For most data sets, only a minor loss in model performance was observed, even for the occupancy descriptors, where all chemical information is neglected. As a further simplification, models were fitted using descriptors from just a few atomic positions occupied in the majority of active or inactive compounds. No further loss in performance was observed, even though in one case descriptors from a single atomic position were used, and in all cases the number of atomic positions required was fewer than twelve. The resulting models suggest that simply filling space at a few key atomic positions is responsible for enhanced activity. At least for the steroids, this finding is at odds with known SAR and binding interactions with the relevant receptor. Using a simulated data set, we illustrate that this paradoxical outcome is a symptom of having too few observations to describe the response in a data set. It is concluded that none of the nine data sets examined can reliably distinguish the merits of different 3D QSAR descriptors and that they should not be used for this purpose. We advocate the use of simulated data, instead.

Group‐Based QSAR (G‐QSAR): Mitigating Interpretation Challenges in QSAR

AbstractSeveral approaches are widely being used as important tools for drug discovery. These approaches include Hansch method, Free‐Wilson method and conventional 2‐D/3‐D QSAR methods. The Hansch analysis assumes that substituents are independent of each other and does not include explicit interactions of groups. In the conventional QSAR method, the interpretation of model generated is rather difficult, as one does not a get clear direction about the site for improvement. A new Group‐Based QSAR (G‐QSAR) method is proposed which uses descriptors evaluated for the fragments of the molecules generated using specific fragmentation rules defined for a given dataset. Herein, we describe the application of G‐QSAR method on two different datasets belonging to a simple congeneric series and a complex noncongeneric series. This method provides models with predictive ability similar or better to conventional methods and in addition provides hints for sites of improvement in the molecules.

Design and synthesis of aryl ether inhibitors of the Bacillus anthracis enoyl-ACP reductase.

The problem of increasing bacterial resistance to the current generation of antibiotics is well documented. Known resistant pathogens such as methicillin-resistant Staphylococcus aureus are becoming more prevalent, while the potential exists for developing drug-resistant pathogens for use as bioweapons, such as Bacillus anthracis. The biphenyl ether antibacterial agent, triclosan, exhibits broad-spectrum activity by targeting the fatty acid biosynthetic pathway through inhibition of enoyl-acyl carrier protein reductase (ENR) and provides a potential scaffold for the development of new, broad-spectrum antibiotics. We used a structure-based approach to develop novel aryl ether analogues of triclosan that target ENR, the product of the fabI gene, from B. anthracis (BaENR). Structure-based design methods were used for the expansion of the compound series including X-ray crystal structure determination, molecular docking, and QSAR methods. Structural modifications were made to both phenyl rings of the 2-phenoxyphenyl core. A number of compounds exhibited improved potency against BaENR and increased efficacy against both the Sterne strain of B. anthracis and the methicillin-resistant strain of S. aureus. X-ray crystal structures of BaENR in complex with triclosan and two other compounds help explain the improved efficacy of the new compounds and suggest future rounds of optimization that might be used to improve their potency.

Toxicophores and Quantitative Structure -Toxicity Relationships for Some Environmental Pollutants

The electron-conformational (EC) method is employed to reveal the toxicophore and to predict aquatic toxicity quantitatively using as a training set a series of 51 compounds that have aquatic toxicity to fish. By performing conformational analysis (optimization of geometries of the low-energy conformers by the PM3 method) and electronic structure calculations (by ab initio method corrected within the SM54/PM3 solvatation model), the Electron-Conformational Matrix of Congruity (ECMC) was constructed for each conformation of these compounds. The toxicophore defined as the EC sub-matrix of activity (ECSA), a sub-matrix with matrix elements common to all the active compounds under consideration within minimal tolerances, is determined by an iterative procedure of comparison of their ECMC’s, gradually minimizing the tolerances. Starting with only the four most toxic compounds, their ECSA (toxicophore) was found to consists of a 4x4 matrix (four sites with certain electronic and topologic characteristics) which was shown to be present in 17 most active compounds. A structure-toxicity correlation between three toxicophore parameters and the activities of these 17 compounds with R2=0.94 was found. It is shown that the same toxicophore with larger tolerances satisfies the compounds with les activity, thus explicitly demonstrating how the activity is controlled by the tolerances quantitatively and which atoms (sites) are most flexible in this respect. This allows for getting slightly different toxicophores for different levels of activity. For some active compounds that have no toxicophore a bimolecular mechanism of activity is suggested. Distinguished from other QSAR methods, no arbitrary descriptors and no statistics are involved in this EC structure-activity investigation.

Molecular characterization of CYP2B6 substrates.

CYP2B6 has not been as fully characterized at the molecular level as other members of the human cytochrome P450 family. As more widely used in vitro probes for characterizing the involvement of this enzyme in the metabolism of xenobiotics have become available, the number of molecules identified as CYP2B6 substrates has increased. In this study we have analyzed the available kinetic data generated by multiple laboratories with human recombinant expressed CYP2B6 and along with calculated molecular properties derived from the ChemSpider database, we have determined the molecular features that appear to be important for CYP2B6 substrates. In addition we have applied 2D and 3D QSAR methods to generate predictive pharmacophore and 2D models. For 28 molecules with K(m) data, the molecular weight (mean +/- SD) is 253.78+/-74.03, ACD/logP is 2.68+/-1.51, LogD(pH 5.5) is 1.51+/-1.43, LogD(pH 7.4) is 2.02+/-1.25, hydrogen bond donor (HBD) count is 0.57 +/-0.57, hydrogen bond acceptor (HBA) count is 2.57+/-1.37, rotatable bonds is 3.50+/-2.71 and total polar surface area (TPSA) is 27.63+/-19.42. A second set of 15 molecules without K(m) data possessed similar mean molecular property values. These properties are comparable to those of a set of 21 molecules used in a previous pharmacophore modeling study (Ekins et al., J Pharmacol Exp Ther 288 (1), 21-29, 1999). Only the LogD and HBD values were statistically significantly different between these different datasets. We have shown that CYP2B6 substrates are generally small hydrophobic molecules that are frequently central nervous system active, which may be important for drug discovery research.

Structure alerts for carcinogenicity, and the Salmonella assay system: A novel insight through the chemical relational databases technology

In the past decades, chemical carcinogenicity has been the object of mechanistic studies that have been translated into valuable experimental (e.g., the Salmonella assays system) and theoretical (e.g., compilations of structure alerts for chemical carcinogenicity) models. These findings remain the basis of the science and regulation of mutagens and carcinogens. Recent advances in the organization and treatment of large databases consisting of both biological and chemical information nowadays allows for a much easier and more refined view of data. This paper reviews recent analyses on the predictive performance of various lists of structure alerts, including a new compilation of alerts that combines previous work in an optimized form for computer implementation. The revised compilation is part of the Toxtree 1.50 software (freely available from the European Chemicals Bureau website). The use of structural alerts for the chemical biological profiling of a large database of Salmonella mutagenicity results is also reported. Together with being a repository of the science on the chemical biological interactions at the basis of chemical carcinogenicity, the SAs have a crucial role in practical applications for risk assessment, for: (a) description of sets of chemicals; (b) preliminary hazard characterization; (c) formation of categories for e.g., regulatory purposes; (d) generation of subsets of congeneric chemicals to be analyzed subsequently with QSAR methods; (e) priority setting. An important aspect of SAs as predictive toxicity tools is that they derive directly from mechanistic knowledge. The crucial role of mechanistic knowledge in the process of applying (Q)SAR considerations to risk assessment should be strongly emphasized. Mechanistic knowledge provides a ground for interaction and dialogue between model developers, toxicologists and regulators, and permits the integration of the (Q)SAR results into a wider regulatory framework, where different types of evidence and data concur or complement each other as a basis for making decisions and taking actions.

Antibody Recognition and Conformational Flexibility of a Plaque-Specific β-Amyloid Epitope Modulated by Non-native Peptide Flanking Regions

Here we report on the synthesis, antibody binding, and QSAR studies of a series of linear and cyclic peptides containing a beta-amyloid plaque-specific epitope (Abeta(4-10); FRHDSGY). In these constructs, two or three alpha- l-Ala, alpha- d-Ala, or beta-Ala residues were introduced at both N- and C-termini of the epitope as non-native flanking sequences. Cyclization of the linear Abeta(4-10) epitope peptide resulted in reduced antibody binding. However, the antibody binding could be fully compensated by insertion of alanine flanks into the corresponding cyclic peptides. These results indicate that the modification of a beta-amyloid plaque-specific epitope by combination of cyclization and flanking sequences could generate highly antigenic peptides compared to the native sequence. A novel 3D QSAR method, which explicitly handles conformational flexibility, was developed for the case of such molecular libraries. This method led to the prediction of the binding conformation for the common FRHDSGY sequence.

QSAR without arbitrary descriptors: the electron-conformational method

The electron-conformational (EC) method in QSAR problems employs a unique (based on first principles) descriptor of molecular properties that incorporates the electronic structure and topology of the molecule and is presented in a digital-matrix form suitable for computer processing, the EC matrix of congruity (ECMC). Its matrix elements have clear-cut physical meanings of interatomic distances, bond orders, and atomic reactivity (interaction indices). By comparing these matrices for several active compounds of the training set a group of matrix elements is revealed that are common for these compounds within a minimum tolerance, the EC submatrix of activity (ECSA). The latter is the numerical pharmacophore for the level of activity and diversity of the tried compounds. The EC method was described in detail and used for pharmacophore identification and quantitative bioactivity prediction elsewhere. In this paper we give further general considerations of its uniqueness and emphasize its advantages as compared with traditional QSAR methods, outlining the following three novel points: (1) The unique, non-arbitrary descriptor employed in the EC method avoids the shortcomings of the arbitrary chosen descriptors and statistical estimation of their weight in the evaluation of the pharmacophore used in traditional QSAR methods. Arbitrary descriptors may be interdependent ("non-orthogonal") and their sets are necessarily incomplete, hence they may lead to chance correlations and artifacts. The EC pharmacophore is void of these failures, thus deemed to be absolutely reliable within the accuracy of the experimental data and the diversity of the molecules used in its evaluation; (2) The tolerances in the matrix elements of the ECSA play a special role reflecting the flexibilities of the pharmacophore parameters and the dependence of the activity on the latter quantitatively; they are obtained in a minimization procedure; by increasing the tolerances one can get pharmacophores for larger intervals of activity. An advanced formula is derived for the activity as a function of the drug-receptor bonding energy which handles also the multi-conformational problem, and a regressional procedure is suggested to represent the interaction energy and the activity by the ECSA matrix elements or tolerances; (3) The possibility of bimolecular activity is discussed when a single molecule of the active compound has no pharmacophore, but the latter is present in the bimolecular structure. Examples are given from the problem of aquatic toxicity to fish.

Use of 3D QSAR Models for Database Screening: A Feasibility Study

The applicability and scope of 3D QSAR methods (CoMFA, CoMSIA) to screen databases are examined. A protocol requiring minimal user intervention has been established to align training and test set molecules using FlexS. As model system isozymes of human carbonic anhydrase (hCA) are used, all results are exemplified studying affinity toward hCA II and selectivity between hCA I and II. The predictive power of the obtained models is assessed through prediction of 663 compounds not included in the training set and compared to 2D QSAR models derived from fragment (MACCS) or property (VSA) based descriptors. The predictive power is evaluated with respect to the following criteria: the numerical, concerning the absolute accuracy of prediction, and the categorical, characterizing the ability to assign a compound to the correct activity class.

Classical and Fragment-Based Hologram Structure-Activity Relationships for a Series of Analgesic Cyclic Imides

Cyclic imides have been widely employed in drug design research due to their multiple pharmacological and biological properties. In the present study, two-dimensional quantitative structure-activity relationship (2D QSAR) studies were conducted on a series of potent analgesic cyclic imides using both classical and hologram QSAR (HQSAR) methods, yielding significant statistical models (classical QSAR, q2 = 0.80; HQSAR, q2 = 0.84). The models were then used to evaluate an external data test, and the predicted values were in good agreement with the experimental results, indicating their consistency for untested compounds. Keywords: Imides, Analgesics, Drug design, 2D QSAR, HQSAR

Combined 4D‐Fingerprint and Clustering Based Membrane‐Interaction QSAR Analyses for Constructing Consensus Caco‐2 Cell Permeation Virtual Screens

A set of 30 structurally diverse molecules, for which Caco-2 cell permeation coefficients were determined, formed the training set for construction of Caco-2 cell permeation models based upon membrane-interaction (MI) QSAR analysis and a new QSAR method called 4D-fingerprint QSAR analysis. The descriptor terms of the 4D-fingerprints equation are molecular similarity eigenvalues, and this set of descriptors is being evaluated as a potential "universal" QSAR descriptor set. The 4D-fingerprint model suggests that Caco-2 cell permeation is governed by the spatial distribution of hydrogen bonding and nonpolar groups over the molecular shape of a molecule. Moreover, a complementary resampling of the original Caco-2 cell permeation training set, followed by the construction of several "clustered" MI-QSAR models, led to a consensus model consistent in interpretation with the 4D-fingerprint model.

Ordinal Classification Using Comparative Molecular Field Analysis

Comparative Molecular Field Analysis (CoMFA) is most widely used as one of the 3-dimensional QSAR (3D-QSAR) methods to identify the relationship between chemical structure and biological activity. Conventional CoMFA requires at least 3 orders of experimental data, such as IC50 and Ki, to obtain a good model, although practically there are many screening assays where biological activity is measured only by a rating scale. Hence, rating classification-oriented CoMFA coupled with ordinal logistic regression has been developed, and its predictive ability and 3D graphical analysis ability have been investigated. As a result, this novel CoMFA (Logistic CoMFA) has been found to be more robust than conventional CoMFAs in both predictive and 3D graphical analysis abilities. Furthermore, Logistic CoMFA is useful since it can provide the probability of each rank.

Multivariate QSAR: From Classical Descriptors to New Perspectives

This review describes an overview of multivariate QSAR methods, from classical analysis to 3D approaches and new perspectives. Data exploration, multivariate regression and molecular descriptors are some topics also appraised here. Special emphasis is given to a recently developed 2D image-based approach, known as MIA-QSAR, which is an improved method in many aspects, namely computing cost, simplicity and prediction performance. Remarks on the MIA-QSAR technique, numerical examples and comparison with traditional methodologies, in addition to a description of limitations and potentialities of this method, are also discussed. Keywords: Descriptors, ligand approach, MIA-QSAR, multivariate QSAR