Modeling of Cyclin-Dependent Kinase Inhibition by 1H-Pyrazolo[3,4-d]Pyrimidine Derivatives Using Artificial Neural Network Ensembles

QSPR study on the polyacrylate-water partition coefficients of hydrophobic organic compounds

The partition coefficient is essential for the analysis of organic chemicals using solid-phase microextraction (SPME) techniques. In this study, a quantitative structure-property relationship (QSPR) model was developed with chemical descriptors for the prediction of the polyacrylate (PA)-water partition coefficient (K_PA-w). The major variables influencing K_PA-w in the QSPR model were CrippenlogP (crippen octanal-water partition coefficient), RNCG (relative negative charge-most negative charge/total negative charge), VE2_Dzv (average coefficient sum of the last eigenvector from the Barysz matrix/weighted by van der Waals volume), and ATSC4v (centred Broto-Moreau autocorrelation-lag 4/weighted by van der Waals volume). The relative determination coefficient (R²) and cross-validation coefficient (Q²) were 0.898 and 0.858, respectively, which implied that the model had excellent robustness. Mechanistic interpretation suggested that the factors affecting the partitioning process between PA and water are the hydrophobicity, relative negative charge, and van der Waals volume of a chemical. The results of this study provide a good tool for predicting the log K_PA-w values of diverse hydrophobic organic compounds (HOCs) within the applicability domain to reduce experimental costs and the time required for innovation.

Neural Network Modeling of AChE Inhibition by New Carbazole-Bearing Oxazolones

Acetylcholine esterase (AChE) is one of the targeted enzymes in the therapy of important neurodegenerative diseases such as Alzheimer's disease. Many studies on carbazole- and oxazolone-based compounds have been conducted in the last decade due to the importance of these compounds. New carbazole-bearing oxazolones were synthesized from several carbazole aldehydes and p-nitrobenzoyl glycine as AChE inhibitors by the Erlenmeyer reaction in the present study. The inhibitory effects of three carbazole-bearing oxazolone derivatives on AChE were studied in vitro and the experimental results were modeled using artificial neural network (ANN). The developed ANN provided sufficient correlation between several dependent systems, including enzyme inhibition. The inhibition data for AChE were modeled by a two-layered ANN architecture. High correlation coefficients were observed between the experimental and predicted ANN results. Synthesized carbazole-bearing oxazolone derivatives inhibited AChE under in vitro conditions, and further research involving in vivo studies is recommended. An ANN may be a useful alternative modeling approach for enzyme inhibition.

Investigation of the differences in activity between hydroxycycloalkyl N1 substituted pyrazole derivatives as inhibitors of B-Raf kinase by using docking, molecular dynamics, QM/MM, and fragment-based de novo design: study of binding mode of diastereomer compounds

N1 substituted pyrazole derivatives show diverse B-Raf kinase inhibitory activities when different hydroxy-substituted cycloalkyl groups are placed at this position. Docking, molecular dynamics (MD) simulations, and hybrid calculation methods (Quantum Mechanics/Molecular Mechanics (QM/MM)) were performed on the complexes, in order to explain these differences. Docking of the inhibitors showed the same orientation that X-ray crystal structure of the analogous (1E)-5-[1-(4-piperidinyl)-3-(4-pyridinyl)-1H-pyrazol-4-yl]-2,3-dihydro-1H-inden-1-one oxime. MD simulations of the most active diastereomer compounds containing cis- and trans-3-hydroxycyclohexyl substituents showed stable interactions with residue Ile463 at the entrance of the B-Raf active site. On the other hand, the less active diastereomer compounds containing cis- and trans-2-hydroxycyclopentyl substituents showed interactions with inner residues Asn580 and Ser465. We found that the differences in activity can be explained by considering the dynamic interactions between the inhibitors and their surrounding residues within the B-Raf binding site. We also explained the activity trend by using a testing scoring function derived from more reliable QM/MM calculations. In addition, we search for new inhibitors from a virtual screening carried out by fragment-based de novo design. We generated a set of approximately 200 virtual compounds, which interact with Ile463 and fulfill druglikeness properties according to Lipinski, Veber, and Ghose rules.

Study of differences in the VEGFR2 inhibitory activities between semaxanib and SU5205 using 3D-QSAR, docking, and molecular dynamics simulations

Semaxanib (SU5416) and 3-[4'-fluorobenzylidene]indolin-2-one (SU5205) are structurally similar drugs that are able to inhibit vascular endothelial growth factor receptor-2 (VEGFR2), but the former is 87 times more effective than the latter. Previously, SU5205 was used as a radiolabelled inhibitor (as surrogate for SU5416) and a radiotracer for positron emission tomography (PET) imaging, but the compound exhibited poor stability and only a moderate IC(50) toward VEGFR2. In the current work, the relationship between the structure and activity of these drugs as VEGFR2 inhibitors was studied using 3D-QSAR, docking and molecular dynamics (MD) simulations. First, comparative molecular field analysis (CoMFA) was performed using 48 2-indolinone derivatives and their VEGFR2 inhibitory activities. The best CoMFA model was carried out over a training set including 40 compounds, and it included steric and electrostatic fields. In addition, this model gave satisfactory cross-validation results and adequately predicted 8 compounds contained in the test set. The plots of the CoMFA fields could explain the structural differences between semaxanib and SU5205. Docking and molecular dynamics simulations showed that both molecules have the same orientation and dynamics inside the VEGFR2 active site. However, the hydrophobic pocket of VEGFR2 was more exposed to the solvent media when it was complexed with SU5205. An energetic analysis, including Embrace and MM-GBSA calculations, revealed that the potency of ligand binding is governed by van der Waals contacts.

Docking and quantitative structure-activity relationship studies for 3-fluoro-4-(pyrrolo[2,1-f][1,2,4]triazin-4-yloxy)aniline, 3-fluoro-4-(1H-pyrrolo[2,3-b]pyridin-4-yloxy)aniline, and 4-(4-amino-2-fluorophenoxy)-2-pyridinylamine derivatives as c-Met kinase inhibitors

We have performed docking of 3-fluoro-4-(pyrrolo[2,1-f][1,2,4]triazin-4-yloxy)aniline (FPTA), 3-fluoro-4-(1H-pyrrolo[2,3-b]pyridin-4-yloxy)aniline (FPPA), and 4-(4-amino-2-fluorophenoxy)-2-pyridinylamine (AFPP) derivatives complexed with c-Met kinase to study the orientations and preferred active conformations of these inhibitors. The study was conducted on a selected set of 103 compounds with variations both in structure and activity. Docking helped to analyze the molecular features which contribute to a high inhibitory activity for the studied compounds. In addition, the predicted biological activities of the c-Met kinase inhibitors, measured as IC(50) values were obtained by using quantitative structure-activity relationship (QSAR) methods: Comparative molecular similarity analysis (CoMSIA) and multiple linear regression (MLR) with topological vectors. The best CoMSIA model included steric, electrostatic, hydrophobic, and hydrogen bond-donor fields; furthermore, we found a predictive model containing 2D-autocorrelation descriptors, GETAWAY descriptors (GETAWAY: Geometry, Topology and Atom-Weight AssemblY), fragment-based polar surface area (PSA), and MlogP. The statistical parameters: cross-validate correlation coefficient and the fitted correlation coefficient, validated the quality of the obtained predictive models for 76 compounds. Additionally, these models predicted adequately 25 compounds that were not included in the training set.

A quantitative structure-activity relationship study of tetrabutylphosphonium bromide analogs as muscarinic acetylcholine receptors

Quantitative structure-activity relationship (QSAR) of tetrabutylphosphonium bromide (TBPB) analogs as muscarinic acetylcholine receptors (mAChRs) was studied. A suitable set of molecular descriptors was calculated and stepwise multiple linear regression (SW-MLR) was employed to select those descriptors that resulted in the best fitted models. A MLR model with three selected descriptors was obtained. Furthermore, the MLR model was validated using the leave-one-out (LOO) and leave-group-out (LGO) crossvalidation, and the Y-randomization test. This model, with high statistical significance (R2 train = 0.982, F = 388.715, Q2 LOO = 0.973, Q2 LGO = 0.977 and R2 test = 0.986) could predict the activity of the molecules with a percentage prediction error lower than 5 %.

Computational study on the interaction of N1 substituted pyrazole derivatives with B-raf kinase: an unusual water wire hydrogen-bond network and novel interactions at the entrance of the active site

Docking and molecular dynamics (MD) simulations of N1 substituted pyrazole derivatives complexed with B-Raf kinase were performed to gain insight into the structural and energetic preferences of these inhibitors. First, a comparative study of fully automated docking programs AutoDock, ICM, GLIDE, and Surflex-Dock in closely approximating the X-ray crystal structure of the inhibitor (1E)-5-[1-(4-piperidinyl)-3-(4-pyridinyl)-1H-pyrazol-4-yl]-2,3-dihydro-1H-inden-1-one oxime was performed. Afterward, the dynamics of the above-mentioned compound and the less active analogous compounds with 1-methyl-4-piperidinyl and tetrahydro-2H-pyran-4-yl groups at position N1 of pyrazole ring inside the B-Raf active site were analyzed by MD simulations. We found that the most active compound has stable interactions with residues Ile463 and His539 at the entrance of the B-Raf active site. Those interactions were in very good agreement with more reliable quantum mechanics/molecular mechanics calculations performed on the torsional angle phi between the pyrazole ring and the substituents at position N1. In addition, we identified a water wire connecting N2 of the pyrazole ring, Cys532, and Ser536, which is composed of three water molecules for the most active compound. We found some differences in the water wire hydrogen-bond network formed by less active compounds. We suggest that the differences between these structural features are responsible for the differences in activity among the studied compounds.

Computational study of the interactions between guanine derivatives and cyclin-dependent kinase 2 (CDK2) by CoMFA and QM/MM

Comparative molecular field analysis (CoMFA) and QM/MM hybrid calculations were performed on 9H-purine derivatives as CDK2 inhibitors. CoMFA was carried out to describe the activities of 78 analogues. The models were applied to a training set including 64 compounds. The best CoMFA model included steric and electrostatic fields, had a good Q(2) value of 0.845, and adequately predicted the compounds contained in the test set. Furthermore, plots of the steric CoMFA field allowed conclusions to be drawn for the choice of suitable inhibitors. In addition, the dynamical behavior of compounds with 4-(aminosulfonyl)phenyl, 4-[(methylamino)sulfonyl]phenyl, 4-[(dimethylamino)sulfonyl]phenyl, and [3-methoxy-4-(aminosulfonyl)]phenyl groups at position 2 of the 9H-purine scaffold inside the CDK2 active site were analyzed by QM/MM calculations. The interactions of these compounds with residues Lys89, Asp86, and Ile10 were characterized.

Genetic algorithm optimization in drug design QSAR: Bayesian-regularized genetic neural networks (BRGNN) and genetic algorithm-optimized support vectors machines (GA-SVM)

Many articles in "in silico" drug design implemented genetic algorithm (GA) for feature selection, model optimization, conformational search, or docking studies. Some of these articles described GA applications to quantitative structure-activity relationships (QSAR) modeling in combination with regression and/or classification techniques. We reviewed the implementation of GA in drug design QSAR and specifically its performance in the optimization of robust mathematical models such as Bayesian-regularized artificial neural networks (BRANNs) and support vector machines (SVMs) on different drug design problems. Modeled data sets encompassed ADMET and solubility properties, cancer target inhibitors, acetylcholinesterase inhibitors, HIV-1 protease inhibitors, ion-channel and calcium entry blockers, and antiprotozoan compounds as well as protein classes, functional, and conformational stability data. The GA-optimized predictors were often more accurate and robust than previous published models on the same data sets and explained more than 65% of data variances in validation experiments. In addition, feature selection over large pools of molecular descriptors provided insights into the structural and atomic properties ruling ligand-target interactions.

Docking and quantitative structure-activity relationship studies for sulfonyl hydrazides as inhibitors of cytosolic human branched-chain amino acid aminotransferase

We have performed the docking of sulfonyl hydrazides complexed with cytosolic branched-chain amino acid aminotransferase (BCATc) to study the orientations and preferred active conformations of these inhibitors. The study was conducted on a selected set of 20 compounds with variation in structure and activity. In addition, the predicted inhibitor concentration (IC(50)) of the sulfonyl hydrazides as BCAT inhibitors were obtained by a quantitative structure-activity relationship (QSAR) method using three-dimensional (3D) vectors. We found that three-dimensional molecule representation of structures based on electron diffraction (3D-MoRSE) scheme contains the most relevant information related to the studied activity. The statistical parameters [cross-validate correlation coefficient (Q(2) = 0.796) and fitted correlation coefficient (R(2) = 0.899)] validated the quality of the 3D-MoRSE predictive model for 16 compounds. Additionally, this model adequately predicted four compounds that were not included in the training set.

QSAR study of 2-(1-Propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide as PARP inhibitors for treatment of cancer

Quantitative structure-activity relationship of the 2-(1-propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide as a potent inhibitor of poly(ADP-ribose) polymerase for cancer treatment was studied. A suitable set of molecular descriptors was calculated and the genetic algorithm was employed to select those descriptors that resulted in the best fitted models. Excellent results were obtained employing multiple linear regressions and critically discussed using a variety of statistical parameters. Furthermore, the model was validated using leave-one-out and leave-group-out cross-validation, external test set and chance correlation. A genetic algorithm-multiple linear regression model with seven selected descriptors was obtained. This model, with high statistical significance (R(2) = 0.935, Q(2)_(LOO)= 0.894, Q(2)_(LGO)= 0.875, F = 53.481), could be used to predict poly(ADP-ribose) polymerase inhibitor activity of the molecules.

Modeling of acetylcholinesterase inhibition by tacrine analogues using Bayesian-regularized Genetic Neural Networks and ensemble averaging

Acetylcholinesterase inhibition was modeled for a set of 136 tacrine analogues using Bayesian-regularized Genetic Neural Networks (BRGNNs). In the BRGNN approach the Bayesian-regularization avoids overtraining/overfitting and the genetic algorithm (GA) allows exploring an ample pool of 3D-descriptors. The predictive capacity of our selected model was evaluated by averaging multiple validation sets generated as members of diverse-training set neural network ensembles (NNEs). The ensemble averaging provides reliable statistics. When 40 members are assembled, the NNE provides a reliable measure of training and test set R values of 0.921 and 0.851 respectively. In other respects, the ability of the nonlinear selected GA space for differentiating the data was evidenced when the total data set was well distributed in a Kohonen Self-Organizing Map (SOM). The location of the inhibitors in the map facilitates the analysis of the connection between compounds and serves as a useful tool for qualitative predictions.

Protein radial distribution function (P-RDF) and Bayesian-Regularized Genetic Neural Networks for modeling protein conformational stability: Chymotrypsin inhibitor 2 mutants

Development of novel computational approaches for modeling protein properties is a main goal in applied Proteomics. In this work, we reported the extension of the radial distribution function (RDF) scores formalism to proteins for encoding 3D structural information with modeling purposes. Protein-RDF (P-RDF) scores measure spherical distributions on protein 3D structure of 48 amino acids/residues properties selected from the AAindex data base. P-RDF scores were tested for building predictive models of the change of thermal unfolding Gibbs free energy change (DeltaDeltaG) of chymotrypsin inhibitor 2 upon mutations. In this sense, an ensemble of Bayesian-Regularized Genetic Neural Networks (BRGNNs) yielded an optimum nonlinear model for the conformational stability. The ensemble predictor described about 84% and 70% variance of the data in training and test sets, respectively.

Computational chemistry development of a unified free energy Markov model for the distribution of 1300 chemicals to 38 different environmental or biological systems

Predicting tissue and environmental distribution of chemicals is of major importance for environmental and life sciences. Most of the molecular descriptors used in computational prediction of chemicals partition behavior consider molecular structure but ignore the nature of the partition system. Consequently, computational models derived up-to-date are restricted to the specific system under study. Here, a free energy-based descriptor (DeltaG(k)) is introduced, which circumvent this problem. Based on DeltaG(k), we developed for the first time a single linear classification model to predict the partition behavior of a broad number of structurally diverse drugs and other chemicals (1300) for 38 different partition systems of biological and environmental significance. The model presented training/predicting set accuracies of 91.79/88.92%. Parametrical assumptions were checked. Desirability analysis was used to explore the levels of the predictors that produce the most desirable partition properties. Finally, inversion of the partition direction for each one of the 38 partition systems evidences that our models correctly classified 89.08% of compounds with an uncertainty of only +/-0.17% independently of the direction of the partition process used to seek the model. Other 10 different classification models (linear, neural networks, and genetic algorithms) were also tested for the same purposes. None of these computational models favorably compare with respect to the linear model indicating that our approach capture the main aspects that govern chemicals partition in different systems.

Ensemble feature selection: consistent descriptor subsets for multiple QSAR models

Selecting a small subset of descriptors from a large pool to build a predictive quantitative structure-activity relationship (QSAR) model is an important step in the QSAR modeling process. In general, subset selection is very hard to solve, even approximately, with guaranteed performance bounds. Traditional approaches employ deterministic or stochastic methods to obtain a descriptor subset that leads to an optimal model of a single type (such as linear regression or a neural network). With the development of ensemble modeling approaches, multiple models of differing types are individually developed resulting in different descriptor subsets for each model type. However, it is advantageous, from the point of view of developing interpretable QSAR models, to have a single set of descriptors that can be used for different model types. In this paper, we describe an approach to the selection of a single, optimal, subset of descriptors for multiple model types. We apply this approach to three data sets, covering both regression and classification, and show that the constraint of forcing different model types to use the same set of descriptors does not lead to a significant loss in predictive ability for the individual models considered. In addition, interpretations of the individual models developed using this approach indicate that they encode similar structure-activity trends.

Amino acid sequence autocorrelation vectors and bayesian-regularized genetic neural networks for modeling protein conformational stability: Gene V protein mutants

Development of novel computational approaches for modeling protein properties from their primary structure is the main goal in applied proteomics. In this work, we reported the extension of the autocorrelation vector formalism to amino acid sequences for encoding protein structural information with modeling purposes. Amino acid sequence autocorrelation (AASA) vectors were calculated by measuring the autocorrelations at sequence lags ranging from 1 to 15 on the protein primary structure of 48 amino acid/residue properties selected from the AAindex data base. A total of 720 AASA descriptors were tested for building predictive models of the change of thermal unfolding Gibbs free energy change (delta deltaG) of gene V protein upon mutation. In this sense, ensembles of Bayesian-regularized genetic neural networks (BRGNNs) were used for obtaining an optimum nonlinear model for the conformational stability. The ensemble predictor described about 88% and 66% variance of the data in training and test sets respectively. Furthermore, the optimum AASA vector subset not only helped to successfully model unfolding stability but also well distributed wild-type and gene V protein mutants on a stability self-organized map (SOM), when used for unsupervised training of competitive neurons.

Ensembles of Bayesian-regularized genetic neural networks for modeling of acetylcholinesterase inhibition by huprines

Acetylcholinesterase inhibition was modeled for a set of huprines using ensembles of Bayesian-regularized Genetic Neural Networks. In the Bayesian-regularized Genetic Neural Network approach the Bayesian regularization avoids overfitted regressions and the genetic algorithm allows exploring a wide pool of three-dimensional descriptors. The predictive capacity of our selected model was evaluated by averaging multiple validation sets generated as members of neural network ensembles. When 60 members are assembled, the neural network ensemble provides a reliable measure of training and test set R(2)-values of 0.945 and 0.850 respectively. In other respects, the ability of the nonlinear selected genetic algorithm space for differentiate the data were evidenced when total data set was well distributed in a Kohonen self-organizing map. The analysis of the self-organizing map zones allows establishing the main structural features differentiated by our vectorial space.

QSAR models for predicting the activity of non-peptide luteinizing hormone-releasing hormone (LHRH) antagonists derived from erythromycin A using quantum chemical properties

Multiple linear regression (MLR) combined with genetic algorithm (GA) and Bayesian-regularized Genetic Neural Networks (BRGNNs) were used to model the binding affinity (pK(I)) of 38 11,12-cyclic carbamate derivatives of 6-O-methylerythromycin A for the Human Luteinizing Hormone-Releasing Hormone (LHRH) receptor using quantum chemical descriptors. A multiparametric MLR equation with good statistical quality was obtained that describes the features relevant for antagonistic activity when the substituent at the position 3 of the erythronolide core was varied. In addition, four-descriptor linear and nonlinear models were established for the whole dataset. Such models showed high statistical quality. However, the BRGNN model was better than the linear model according to the external validation process. In general, our linear and nonlinear models reveal that the binding affinity of the compounds studied for the LHRH receptor is modulated by electron-related terms.

Contemporary QSAR Classifiers Compared

We present a comparative assessment of several state-of-the-art machine learning tools for mining drug data, including support vector machines (SVMs) and the ensemble decision tree methods boosting, bagging, and random forest, using eight data sets and two sets of descriptors. We demonstrate, by rigorous multiple comparison statistical tests, that these techniques can provide consistent improvements in predictive performance over single decision trees. However, within these methods, there is no clearly best-performing algorithm. This motivates a more in-depth investigation into the properties of random forests. We identify a set of parameters for the random forest that provide optimal performance across all the studied data sets. Additionally, the tree ensemble structure of the forest may provide an interpretable model, a considerable advantage over SVMs. We test this possibility and compare it with standard decision tree models.

Quantitative Structure?Activity Relationship Modeling of Growth Hormone Secretagogues Agonist Activity of some Tetrahydroisoquinoline 1-Carboxamides

Growth hormone secretagogue agonist activities for a data set of 45 tetrahydroisoquinoline 1-carboxamides were modeled using several kinds of molecular descriptors from dragon software. A linear model with six variables selected from a large pool of two-dimensional descriptors described 80% of cross-validation data variance. Similar results were found for a model obtained from a pool of three-dimensional descriptors. Size and hydrophilicity-related atomic properties such as mass, polarizability, and van der Waals volume were determined to be the most relevant for the differential growth hormone secretagogue agonist activities of the compounds studied. In addition, Artificial Neural Networks were trained using optimum variables from the linear models; however, they were found to overfit the data and resulted in similar or lower predictive power.

Linear and nonlinear QSAR study of N-hydroxy-2-[(phenylsulfonyl)amino]acetamide derivatives as matrix metalloproteinase inhibitors

The inhibitory activity (IC50) toward matrix metalloproteinases (MMP-1, MMP-2, MMP-3, MMP-9, and MMP-13) of N-hydroxy-2-[(phenylsulfonyl)amino]acetamide derivatives (HPSAAs) has been successfully modeled using 2D autocorrelation descriptors. The relevant molecular descriptors were selected by linear and nonlinear genetic algorithm (GA) feature selection using multiple linear regression (MLR) and Bayesian-regularized neural network (BRANN) approaches, respectively. The quality of the models was evaluated by means of cross-validation experiments and the best results correspond to nonlinear ones (Q2>0.7 for all models). Despite the high correlation between the studied compound IC50 values, the 2D autocorrelation space brings different descriptors for each MMP inhibition. On the basis of these results, these models contain useful molecular information about the ligand specificity for MMP S'1, S1, and S'2 pockets.

Amino Acid Sequence Autocorrelation Vectors and Ensembles of Bayesian-Regularized Genetic Neural Networks for Prediction of Conformational Stability of Human Lysozyme Mutants

Development of novel computational approaches for modeling protein properties from their primary structure is a main goal in applied proteomics. In this work, we reported the extension of the autocorrelation vector formalism to amino acid sequences for encoding protein structural information with modeling purposes. Amino Acid Sequence Autocorrelation (AASA) vectors were calculated by measuring the autocorrelations at sequence lags ranging from 1 to 15 on the protein primary structure of 48 amino acid/residue properties selected from the AAindex database. A total of 720 AASA descriptors were tested for building predictive models of the thermal unfolding Gibbs free energy change of human lysozyme mutants. In this sense, ensembles of Bayesian-Regularized Genetic Neural Networks (BRGNNs) were used for obtaining an optimum nonlinear model for the conformational stability. The ensemble predictor described about 88% and 68% variance of the data in training and test sets, respectively. Furthermore, the optimum AASA vector subset was shown not only to successfully model unfolding thermal stability but also to distribute wild-type and mutant lysozymes on a stability Self-organized Map (SOM) when used for unsupervised training of competitive neurons.