Predictive Quantitative Structure-activity Relationship Models Research Articles

Improvement of Multivariate Image Analysis Applied to Quantitative Structure–Activity Relationship (QSAR) Analysis by Using Wavelet‐Principal Component Analysis Ranking Variable Selection and Least‐Squares Support Vector Machine Regression: QSAR Study of Checkpoint Kinase WEE1 Inhibitors

Inhibition of tyrosine kinase enzyme WEE1 is an important step for the treatment of cancer. The bioactivities of a series of WEE1 inhibitors have been previously modeled through comparative molecular field analyses (CoMFA and CoMSIA), but a two-dimensional image-based quantitative structure-activity relationship approach has shown to be highly predictive for other compound classes. This method, called multivariate image analysis applied to quantitative structure-activity relationship, was applied here to derive quantitative structure-activity relationship models. Whilst the well-known bilinear and multilinear partial least squares regressions (PLS and N-PLS, respectively) correlated multivariate image analysis descriptors with the corresponding dependent variables only reasonably well, the use of wavelet and principal component ranking as variable selection methods, together with least-squares support vector machine, improved significantly the prediction statistics. These recently implemented mathematical tools, particularly novel in quantitative structure-activity relationship studies, represent an important advance for the development of more predictive quantitative structure-activity relationship models and, consequently, new drugs.

Docking and Quantitative Structure–Activity Relationship Studies for the Bisphenylbenzimidazole Family of Non‐Nucleoside Inhibitors of HIV‐1 Reverse Transcriptase

Molecular docking studies on a set of bisphenylbenzimidazole derivatives were conducted to identify the compounds binding orientations within the HIV-1 reverse transcriptase non-nucleoside binding pocket. A good correlation between the calculated binding free energies and the experimental inhibitory activities suggests that the identified binding conformations of these inhibitors are reliable. Based on obtained bisphenylbenzimidazoles binding conformations, a predictive quantitative structure-activity relationship model based on radial distribution function descriptors was developed. The obtained quantitative structure-activity relationship model was predictive according to internal and external validation experiments and might provide guidelines for the design of novel non-nucleoside HIV-1 reverse transcriptase inhibitors based on the 1-benzyl-2-arylbenzimidazole scaffold.

Exploring Predictive QSAR Models Using Quantum Topological Molecular Similarity (QTMS) Descriptors for Toxicity of Nitroaromatics to Saccharomyces cerevisiae

AbstractIn view of the widespread industrial use of nitroaromatics and their consequent ecotoxicological hazard potential, we constructed predictive Quantitative Structure–Activity Relationship (QSAR) models for the toxicity of nitroaromatics to the ecologically important species Saccharomyces cerevisiae. We used Quantum Topological Molecular Similarity (QTMS) descriptors along with electrophilicity index (ELUMO) and lipid water partition coefficient (log Kow) as predictor variables. The QTMS descriptors were calculated at B3LYP/6‐311+G(2d,p) level of theory. QTMS descriptors were employed to complement the deficiency of ELUMO in setting up predictive QSAR models from the view point of external validation. The dataset was divided into a training set (18 compounds) and test set (six compounds) in a ratio of three to one. Partial Least Square (PLS) models were developed based on the training set compounds. The predictive capacity of the models was assessed by the test compounds. The models were also validated by a randomisation test and leave‐one‐seventh‐out crossvalidation test. The results suggest that Bond Critical Point (BCP) descriptors can develop predictive QSAR models for nitroaromatic toxicity to Saccharomyces cerevisiae when used along with ELUMO and log Kow. The diagnostic potential of QTMS descriptors could also reveal the importance of the nitro group for nitroaromatic toxicity.

Predictive QSAR modeling of HIV reverse transcriptase inhibitor TIBO derivatives

Comparative quantitative structure–activity relationship (QSAR) studies have been carried out on tetrahydroimidazo[4,5,1- jk][1,4]benzodiazepine (TIBO) derivatives as reverse transcriptase inhibitors ( n = 70) using topological, structural, physicochemical, electronic and spatial descriptors. The data set was divided into training and test sets using a cluster-based method. Linear models were developed using multiple regression (with stepwise regression, factor analysis and genetic function approximation (GFA) as variable selection tools) and partial least squares (PLS) and combination of factor analysis and partial least squares (FA–PLS). Genetic function approximation (spline) and artificial neural networks (ANN) were used for the development of non-linear models. Using topological and structural descriptors, the best equation was obtained from GFA (spline) based on internal validation ( Q 2 = 0.737), but the model with the best external validation characteristics was obtained with FA–PLS ( R pred 2 = 0.707). When structural, physicochemical, electronic and spatial descriptors were used, the best Q 2 (0.740) value was obtained from GFA (spline) whereas PLS provided the best R pred 2 (0.784) value. When all descriptors were used in combination, the best R pred 2 (0.760) value and the best Q 2 (0.800) value were obtained from ANN and GFA (spline), respectively. The majority of the models satisfied the criteria of external validation recommended by Golbraikh and Tropsha (2002) and the criteria of modified r 2 ( r m 2) values of the test set for external validation as suggested by Roy and Roy (2008). In order to further validate selected models, an external set of 10 TIBO derivatives, which fall within the applicability domain of the models and are not shared with the compounds of the present data set, was taken from a different source, and reverse transcriptase inhibitory activity of these compounds was predicted. Acceptable values of squared correlation coefficients between the observed and predicted values of the external set compounds were obtained from the selected models suggesting true predictive potential of the models.

Considerations and recent advances in QSAR models for cytochrome P450-mediated drug metabolism prediction

Quantitative structure-activity relationships (QSAR) methods are urgently needed for predicting ADME/T (absorption, distribution, metabolism, excretion and toxicity) properties to select lead compounds for optimization at the early stage of drug discovery, and to screen drug candidates for clinical trials. Use of suitable QSAR models ultimately results in lesser time-cost and lower attrition rate during drug discovery and development. In the case of ADME/T parameters, drug metabolism is a key determinant of metabolic stability, drug-drug interactions, and drug toxicity. QSAR models for predicting drug metabolism have undergone significant advances recently. However, most of the models used lack sufficient interpretability and offer poor predictability for novel drugs. In this review, we describe some considerations to be taken into account by QSAR for modeling drug metabolism, such as the accuracy/consistency of the entire data set, representation and diversity of the training and test sets, and variable selection. We also describe some novel statistical techniques (ensemble methods, multivariate adaptive regression splines and graph machines), which are not yet used frequently to develop QSAR models for drug metabolism. Subsequently, rational recommendations for developing predictable and interpretable QSAR models are made. Finally, the recent advances in QSAR models for cytochrome P450-mediated drug metabolism prediction, including in vivo hepatic clearance, in vitro metabolic stability, inhibitors and substrates of cytochrome P450 families, are briefly summarized.

Predictive QSAR Models for Polyspecific Drug Targets: The Importance of Feature Selection

Since the advent of QSAR (quantitative structure activity relationship) modeling quantitative representations of molecular structures are encoded in terms of information-preserving descriptor values. Nowadays, a nearly infinite variety of potential descriptors is available and descriptor selection is no longer a task which can be done manually. There is an increasing need for automation in order to reduce the dimensionality of the descriptor space. Classical feature selection (FS) and dimensionality reduction (DR) methods like principal component analysis, which relies on the selection of those descriptors that contribute most to the variance of a data set, often fail in providing the best classification result. More sophisticated methods like genetic algorithms, self-organizing-maps and stepwise linear discriminant analysis have proven to be useful techniques in the process of selecting descriptors with a significant discriminative power. The topic FS and DR becomes even more important when predictive models are approached which should describe the QSAR of highly promiscuous target proteins. The ABC-transporter family, the cardiac hERG-potassium channel, and the hepatic cytochrom-P450-family are classical representatives of such poly-specific proteins. In this case the interaction pattern is a rather complex one and thus the selection of the most predictive descriptors needs advanced methods. This review surveys FS and DR methods that have recently been successfully applied to classify ligands of poly-specific target proteins. Keywords: Feature selection, dimensionality reduction, machine learning, chemical descriptors, ABC-transporter, nuclear hormone receptors, cytochrome P450, hERG

Combinatorial QSAR Modeling of Specificity and Subtype Selectivity of Ligands Binding to Serotonin Receptors 5HT1E and 5HT1F

The Quantitative Structure-Activity Relationship (QSAR) approach has been applied to model binding affinity and receptor subtype selectivity of human 5HT1E and 5HT1F receptor-ligands. The experimental data were obtained from the PDSP Ki Database. Several descriptor types and data-mining approaches have been used in the context of combinatorial QSAR modeling. Data mining approaches included k Nearest Neighbor, Automated Lazy Learning (ALL), and PLS; descriptor types included MolConnZ, MOE, DRAGON, Frequent Subgraphs (FSG), and Molecular Hologram Fingerprints (MHFs). Highly predictive QSAR models were generated for all three data sets (i.e., for ligands of both receptor subtypes and for subtype selectivity), and different individual techniques were proved best in each case. For real value activity data available for 5HT1E and 5HT1F ligand binding, models were characterized by leave-one-out cross-validated R(2) (q(2)) for the training sets and predictive R(2) values for the test sets. The best models for 5HT1E ligands were obtained with the kNN approach combined with MolConnZ descriptors (q(2)=0.69, R(2)=0.92); for 5HT1F ligands ALL QSAR method using MolConnZ descriptors gave the best results (R(2)=0.92). Rigorously validated classification models were also developed for the 5HT1E/5HT1F subtype selectivity data set with high correct classification accuracy for both training (CCRtrain=0.88) and test (CCRtest=1.00) sets using kNN with MolConnZ descriptors. The external predictive power of QSAR models was further validated by virtual screening of The Scripps Research Institute (TSRI) screening library to recover 5HT1E agonists and antagonists (not present in the original PDSP data set) with high enrichment factors. The successful development of externally predictive and interpretative QSAR models affords further design and discovery of novel subtype specific GPCR agents.

On some aspects of validation of predictive QSAR models

Quantitative structure-activity relationships (QSARs) represent predictive models derived from application of statistical tools correlating biological activity (including therapeutic and toxic) of chemicals (drugs/toxicants/environmental pollutants) with descriptors representative of molecular structure and/or property. The success of any QSAR model depends on accuracy of the input data, selection of appropriate descriptors and statistical tools, and most importantly validation of the developed model. Validation is the process by which the reliability and relevance of a procedure are established for a specific purpose. Leave one-out cross-validation generally leads to an overestimation of predictive capacity, and even with external validation, no one can be sure whether the selection of training and test sets was manipulated to maximize the predictive capacity of the model being published. In this paper, we present some representative examples of validation of QSAR models in order to explore possible importance of the method of selection of training set compounds, setting training set size and impact of variable selection for training set models for determining the quality of prediction. The major conclusions from the study are: (1) K-means cluster based division of training and prediction sets can be used as a reliable method of division of data set into training and test sets for developing predictive QSAR models; (2) the training set size should be set at an optimal level so that the model is developed with proper training (learning) process and the developed model is able to satisfactorily predict the activity values of the test set compounds; (3) choice of variables for regression based only on Q2 value may not be optimum. Furthermore, predictive R2 value may not be considered as the only criterion to indicate external predictability of a model.

Development of linear and nonlinear predictive QSAR models and their external validation using molecular similarity principle for anti-HIV indolyl aryl sulfones

Quantitative structure–activity relationship (QSAR) studies have been carried out on indolyl aryl sulfones, a class of novel HIV-1 non-nucleoside reverse transcriptase inhibitors, using physicochemical, topological and structural parameters along with appropriate indicator variables. The statistical tools used were linear methods (e.g., stepwise regression analysis, partial least squares (PLS), factor analysis followed by multiple regression (FA-MLR), genetic function approximation combined with multiple linear regression (GFA-MLR) and GFA followed by PLS or G/PLS and nonlinear method (artificial neural network or ANN). In case of physicochemical parameters, GFA-MLR generated the best Equation (n = 97, R2 = 0.862, Q2 = 0.821). Using topological parameters, the best Equation (based on leave-one-out Q2) was obtained with stepwise regression technique (n = 97, R2 = 0.867, Q2 = 0.811). When topological and physicochemical parameters were used in combination, statistical quality increased to a great extent (n = 97, R2 = 0.891, Q2 = 0.849 from stepwise regression). Furthermore, the whole dataset had been divided into test (25% of whole dataset) and training (remaining 75%) sets. Models were developed based on the training set and predictive potential of such models was checked from the test set. The selection of the training set was based on K-means clustering of the standardized descriptors (topological and physicochemical). In this case also the best results were obtained with stepwise regression (n = 72, R2 = 0.906, Q2 = 0.853) but external predictive capacity of this model () was inferior to the model developed from GFA-MLR technique (R2 = 0.883, Q2 = 0.823, ). However, the squared regression coefficient between observed activity and predicted activity values of the test set compounds for the best linear model, i.e., GFA-MLR (r2 = 0.736) was lower in comparison to the best nonlinear model developed using artificial neural network (r2 = 0.781). Thus, based on external validation, the ANN models were superior to the linear models. The predictive potential of the best linear Equation (stepwise regression model) was superior to that of the previously published CoMFA (Q2 = 0.81, SDEPTest = 0.89) on the same data set (Ragno R. et al., J Med Chem 2006, 49, 3172–3184). Furthermore, the physicochemical parameter based models also supported the previous observations based on docking (Ragno R. et al., J Med Chem 2005, 48, 213–223).

On some aspects of validation of predictive quantitative structure–activity relationship models

The success of any quantitative structure–activity relationship model depends on the accuracy of the input data, selection of appropriate descriptors and statistical tools and, most importantly, the validation of the developed model. Validation is the process by which the reliability and relevance of a procedure are established for a specific purpose. This review focuses on the importance of validation of quantitative structure–activity relationship models and different methods of validation. Some important issues, such as internal versus external validation, method of selection of training set compounds and training set size, applicability domain, variable selection and suitable parameters to indicate external predictivity, are also discussed.

Exploring the impact of size of training sets for the development of predictive QSAR models

While building a predictive quantitative structure-activity relationship (QSAR), validation of the developed model is a very important task. However, a truly new set of data being often unavailable for checking predictability and robustness of the developed model, a typical external validation in QSAR studies is commonly performed by splitting the available data into training and test sets. In the present work we have attempted to explore the impact of training set size on the quality of prediction using different topological descriptors and three different statistical techniques. Three different data sets of moderate size have been used for the present study: cytoprotection data of anti-HIV thiocarbamates (n=62), HIV reverse transcriptase inhibition data of 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT) derivatives (n=107) and bioconcentration factor data of diverse functional compounds (n=122). In each case, the data set was divided into different combinations of training and test sets maintaining different size ratios in several iterations. In cases of the first two data sets, significant impact of reduction of training set size was found on the predictive ability of the models while the first data set showing higher dependence on the size than the second one. However, in case of modeling of bioconcentration factor, no significant impact of training set size on the quality of prediction could be found. Hence, no general rule can be formulated regarding the impact of training set size on the quality of prediction. Optimum size of the training set should be set based on a particular data set and types of descriptors and statistical analysis being used.

Probing the Anticancer Activity of Nucleoside Analogues: A QSAR Model Approach Using an Internally Consistent Training Set

The cancer research community has begun to address the in silico modeling approaches, such as quantitative structure-activity relationships (QSAR), as an important alternative tool for screening potential anticancer drugs. With the compilation of a large dataset of nucleosides synthesized in our laboratories, or elsewhere, and tested in a single cytotoxic assay under the same experimental conditions, we recognized a unique opportunity to attempt to build predictive QSAR models. Here, we report a systematic evaluation of classification models to probe anticancer activity, based on linear discriminant analysis along with 2D-molecular descriptors. This strategy afforded a final QSAR model with very good overall accuracy and predictability on external data. Finally, we search for similarities between the natural nucleosides, present in RNA/DNA, and the active nucleosides well-predicted by the model. The structural information then gathered and the QSAR model per se shall aid in the future design of novel potent anticancer nucleosides.

Quantitative structure-activity relationship of some 1-benzylbenzimidazole derivatives as antifungal agents

In the present study, the antifungal activity of some 1-benzylbenzimidazole derivatives against yeast Saccharomyces cerevisiae was investigated. The tested benzimidazoles displayed in vitro antifungal activity and minimum inhibitory concentration (MIC) was determined for all the compounds. Quantitative structure-activity relationship (QSAR) has been used to study the relationships between the antifungal activity and lipophilicity parameter, logP, calculated by using CS Chem-Office Software version 7.0. The results are discussed on the basis of statistical data. The best QSAR model for prediction of antifungal activity of the investigated series of benzimidazoles was developed. High agreement between experimental and predicted inhibitory values was obtained. The results of this study indicate that the lipophilicity parameter has a significant effect on antifungal activity of this class of compounds, which simplify design of new biologically active molecules.

Development of Quantitative Structure-Activity Relationships and Its Application in Rational Drug Design

Over forty years have elapsed since Hansch and Fujita published their pioneering work of quantitative structure-activity relationships (QSAR). Following the introduction of Comparative Molecular Field Analysis (CoMFA) by Cramer in 1998, other three-dimensional QSAR methods have been developed. Currently, combination of classical QSAR and other computational techniques at three-dimensional level is of greatest interest and generally used in the process of modern drug discovery and design. During the last several decades, a number of different mythologies incorporating a range of molecular descriptors and different statistical regression ways have been proposed and successfully applied in developing of new drugs, thus QSAR method has been proven to be indispensable in not only the reliable prediction of specific properties of new compounds, but also the help to elucidate the possible molecular mechanism of the receptor-ligand interactions. Here, we review the recent developments in QSAR and their applications in rational drug design, focusing on the reasonable selection of novel molecular descriptors and the construction of predictive QSAR models by the help of advanced computational techniques.

QSAR and primary docking studies of trans-stilbene (TSB) series of imaging agents for β-amyloid plaques

In vivo imaging β-amyloid plaques can help to diagnose Alzheimer's disease (AD). Some trans-stilbene derivatives are hopeful as probes for in vivo evaluation of β-amyloid plaques. In order to explore the interaction mechanism, quantitative structure-activity relationship (QSAR) and molecular docking studies were performed on trans-stilbene (TSB) compounds. First, 22 TSB-based analogues were optimized using DFT method. Through QSAR analysis, highly predictive QSAR model was developed with r 2 value of 0.857. Based on the QSAR model, one possible binding site was proposed and validated by molecular docking studies. The Lamarckian Genetic Algorithm (LGA) was applied to deal with the ligand-protein interactions. A good correlation between the calculated binding energies and the experimental binding affinities suggests that the identified binding site is reliable. The QSAR model and the information of the ligand–protein interaction would be useful in developing new imaging agents for β-amyloid plaques.

Application of Validated QSAR Models of D1Dopaminergic Antagonists for Database Mining

Rigorously validated quantitative structure-activity relationship (QSAR) models have been developed for 48 antagonists of the dopamine D1 receptor and applied to mining chemical datasets to discover novel potential antagonists. Several QSAR methods have been employed, including comparative molecular field analysis (CoMFA), simulated annealing-partial least squares (SA-PLS), k-nearest neighbor (kNN), and support vector machines (SVM). With the exception of CoMFA, these approaches employed 2D topological descriptors generated with the MolConnZ software package (EduSoft, LLC. MolconnZ, version 4.05; http://www.eslc.vabiotech.com/ [4.05], 2003). The original dataset was split into training and test sets to allow for external validation of each training set model. The resulting models were characterized by cross-validated R2 (q2) for the training set and predictive R2 values for the test set of (q2/R2) 0.51/0.47 for CoMFA, 0.7/0.76 for kNN, R2 for the training and test sets of 0.74/0.71 for SVM, and training set fitness and test set R2 values of 0.68/0.63 for SA-PLS. Validated QSAR models with R2 > 0.7, (i.e., kNN and SVM) were used to mine three publicly available chemical databases: the National Cancer Institute (NCI) database of ca. 250,000 compounds, the Maybridge Database of ca. 56,000 compounds, and the ChemDiv Database of ca. 450,000 compounds. These searches resulted in only 54 consensus hits (i.e., predicted active by all models); five of them were previously characterized as dopamine D1 ligands, but were not present in the original dataset. A small fraction of the purported D1 ligands did not contain a catechol ring found in all known dopamine full agonist ligands, suggesting that they may be novel structural antagonist leads. This study illustrates that the combined application of predictive QSAR modeling and database mining may provide an important avenue for rational computer-aided drug discovery.

QSAR study of natural, synthetic and environmental endocrine disrupting compounds for binding to the androgen receptor

A large data set of 146 natural, synthetic and environmental chemicals belonging to a broad range of structural classes have been tested for their relative binding affinity (expressed as log (RBA)) to the androgen receptor (AR). These chemicals commonly termed endocrine disrupting compounds (EDCs) present a variety of adverse effects in humans and animals. As assays for binding affinity remains a time-consuming task, it is important to develop predictive methods. In this work, quantitative structure–activity relationships (QSARs) were determined using three methods, multiple linear regression (MLR), radical basis function neural network (RBFNN) and support vector machine (SVM). Five descriptors, accounting for hydrogen-bonding interaction, distribution of atomic charges and molecular branching degree, were selected from a heuristic method to build predictive QSAR models. Comparison of the results obtained from three models showed that the SVM method exhibited the best overall performances, with a RMS error of 0.54 log (RBA) units for the training set, 0.59 for the test set, and 0.55 for the whole set. Moreover, six linear QSAR models were constructed for some specific families based on their chemical structures. These predictive toxicology models, should be useful to rapidly identify potential androgenic endocrine disrupting compounds.

Improved prediction of fish bioconcentration factor of Hydrophobic Chemicals

Using a large heterogeneous data-set of 640 organic chemicals, we have developed predictive Quantitative Structure-Activity Relationship models for fish bioconcentration factor (BCF). For 539 chemicals with a log K ow (octanol-water partition coefficient) range of −2.3 to 6.0, we developed a model with [Formula: See Text] and a standard error of 0.661; the primary descriptor was log K ow, and others were polarisability, number of amino groups, hydrogen bond acceptor ability and a molecular shape factor. For 101 chemicals with a log K ow range of 6.0-12.7, we developed a model with [Formula: See Text] and a standard error of 0.777; the descriptors were aqueous solubility (reflecting the importance of this property in governing uptake from aqueous solution), polarity, polarisability, hydrogen bond donor ability and molecular size. Bearing in mind the very great range of BCF values of highly hydrophobic chemicals, our model offers good predictivity of this important environmental property.

Quantitative structure-activity relationship study of histone deacetylase inhibitors.

Histone deacetylases (HDACs) play a critical role in gene transcription and have become a novel target for the discovery of drugs against cancer and other diseases. During the past several years there have been extensive efforts in the identification and optimization of histone deacetylase inhibitors (HDACIs) as novel anticancer drugs. Here we report a comprehensive quantitative structure-activity relationship (QSAR) study of HDACIs in the hope of identifying the structural determinants for anticancer activity. We have identified, collected, and verified the structural and biological activity data for 124 compounds from various literature sources and performed an extensive QSAR study on this comprehensive data set by using various QSAR and classification methods. A highly predictive QSAR model with R(2) of 0.76 and leave-one-out cross-validated R(2) of 0.73 was obtained. The overall rate of cross-validated correct prediction of the classification model is around 92%. The QSAR and classification models provided direct guidance to our internal programs of identifying and optimizing HDAC inhibitors. Limitations of the models were also discussed.

Application of Predictive QSAR Models to Database Mining: Identification and Experimental Validation of Novel Anticonvulsant Compounds

We have developed a drug discovery strategy that employs variable selection quantitative structure-activity relationship (QSAR) models for chemical database mining. The approach starts with the development of rigorously validated QSAR models obtained with the variable selection k nearest neighbor (kNN) method (or, in principle, with any other robust model-building technique). Model validation is based on several statistical criteria, including the randomization of the target property (Y-randomization), independent assessment of the training set model's predictive power using external test sets, and the establishment of the model's applicability domain. All successful models are employed in database mining concurrently; in each case, only variables selected as a result of model building (termed descriptor pharmacophore) are used in chemical similarity searches comparing active compounds of the training set (queries) with those in chemical databases. Specific biological activity (characteristic of the training set compounds) of external database entries found to be within a predefined similarity threshold of the training set molecules is predicted on the basis of the validated QSAR models using the applicability domain criteria. Compounds judged to have high predicted activities by all or the majority of all models are considered as consensus hits. We report on the application of this computational strategy for the first time for the discovery of anticonvulsant agents in the Maybridge and National Cancer Institute (NCI) databases containing ca. 250,000 compounds combined. Forty-eight anticonvulsant agents of the functionalized amino acid (FAA) series were used to build kNN variable selection QSAR models. The 10 best models were applied to mining chemical databases, and 22 compounds were selected as consensus hits. Nine compounds were synthesized and tested at the NIH Epilepsy Branch, Rockville, MD using the same biological test that was employed to assess the anticonvulsant activity of the training set compounds; of these nine, four were exact database hits and five were derived from the hits by minor chemical modifications. Seven of these nine compounds were confirmed to be active, indicating an exceptionally high hit rate. The approach described in this report can be used as a general rational drug discovery tool.