Development of an information-intensive structure–activity relationship model and its application to human respiratory chemical sensitizers

Structure–activity relationship (SAR) models are recognized as powerful tools to predict the toxicologic potential of new or untested chemicals and also provide insight into possible mechanisms of toxicity. Models have been based on physicochemical attributes and structural features of chemicals. We describe herein the development of a new SAR modeling algorithm called cat-SAR that is capable of analyzing and predicting chemical activity from divergent biological response data. The cat-SAR program develops chemical fragment-based SAR models from categorical biological response data (e.g. toxicologically active and inactive compounds). The database selected for model development was a published set of chemicals documented to cause respiratory hypersensitivity in humans. Two models were generated that differed only in that one model included explicate hydrogen containing fragments. The predictive abilities of the models were tested using leave-one-out cross-validation tests. One model had a sensitivity of 0.94 and specificity of 0.87 yielding an overall correct prediction of 91%. The second model had a sensitivity of 0.89, specificity of 0.95 and overall correct prediction of 92%. The demonstrated predictive capabilities of the cat-SAR approach, together with its modeling flexibility and design transparency, suggest the potential for its widespread applicability to toxicity prediction and for deriving mechanistic insight into toxicologic effects.     

A categorical structure–activity relationship analysis of GPR119 ligands

The categorical structure–activity relationship (cat-SAR) expert system has been successfully used in the analysis of chemical compounds that cause toxicity. Herein we describe the use of this fragment-based approach to model ligands for the G protein-coupled receptor 119 (GPR119). Using compounds that are known GPR119 agonists and compounds that we have confirmed experimentally that are not GPR119 agonists, four distinct cat-SAR models were developed. Using a leave-one-out validation routine, the best GPR119 model had an overall concordance of 99%, a sensitivity of 99%, and a specificity of 100%. Our findings from the in-depth fragment analysis of several known GPR119 agonists were consistent with previously reported GPR119 structure–activity relationship (SAR) analyses. Overall, while our results indicate that we have developed a highly predictive cat-SAR model that can be potentially used to rapidly screen for prospective GPR119 ligands, the applicability domain must be taken into consideration. Moreover, our study demonstrates for the first time that the cat-SAR expert system can be used to model G protein-coupled receptor ligands, many of which are important therapeutic agents.     

AI and SAR approaches for predicting chemical carcinogenicity: survey and status report

A wide variety of artificial intelligence (AI) and structure-activity relationship (SAR) approaches have been applied to tackling the general problem of predicting rodent chemical carcinogenicity. Given the diversity of chemical structures and mechanisms relative to this endpoint, the shared challenge of these approaches is to accurately delineate classes of active chemicals representing distinct biological and chemical mechanism domains, and within those classes determine the structural features and properties responsible for modulating activity. In the following discussion, we present a survey of AI and SAR approaches that have been applied to the prediction of rodent carcinogenicity, and discuss these in general terms and in the context of the results of two organized prediction exercises (PTE-1 and PTE-2) sponsored by the US National Cancer Institute/National Toxicology Program. Most models participating in these exercises were successful in identifying major structural-alerting classes of active carcinogens, but failed in modeling the more subtle modifiers to activity within those classes. In addition, methods that incorporated mechanism-based reasoning or biological data along with structural information outperformed models limited to structural information exclusively. Finally, a few recent carcinogenicity-modeling efforts are presented illustrating progress in tackling some aspects of the carcinogenicity prediction problem. The first example, a QSAR model for predicting carcinogenic potency of aromatic amines, illustrates that success is possible within well-represented classes of carcinogens. From the second example, a newly developed FDA/OTR MultiCASE model for predicting the carcinogenicity of pharmaceuticals, we conclude that the definitions of biological activity and nature of chemicals in the training set are important determinants of the predictive success and specificity/sensitivity characteristics of a derived model.     

Decision tree SAR models for developmental toxicity based on an FDA/TERIS database

Humans are exposed to thousands of environmental chemicals for which no developmental toxicity information is available. Structure-activity relationships (SARs) are models that could be used to efficiently predict the biological activity of potential developmental toxicants. However, at this time, no adequate SAR models of developmental toxicity are available for risk assessment. In the present study, a new developmental database was compiled by combining toxicity information from the Teratogen Information System (TERIS) and the Food and Drug Administration (FDA) guidelines. We implemented a decision tree modeling procedure, using Classification and Regression Tree software and a model ensemble approach termed bagging. We then assessed the empirical distributions of the prediction accuracy measures of the single and ensemble-based models, achieved by repeating our modeling experiment many times by repeated random partitioning of the working database. The decision tree developmental SAR models exhibited modest prediction accuracy. Bagging tended to enhance the accuracy of prediction. Also, the model ensemble approach reduced the variability of prediction measures compared to the single model approach. Further research with data derived from animal species- and endpoint-specific components of an extended and refined FDA/TERIS database has the potential to derive SAR models that would be useful in the developmental risk assessment of the thousands of untested chemicals.     

Decision tree SAR models for developmental toxicity based on an FDA/TERIS database

Humans are exposed to thousands of environmental chemicals for which no developmental toxicity information is available. Structure-activity relationships (SARs) are models that could be used to efficiently predict the biological activity of potential developmental toxicants. However, at this time, no adequate SAR models of developmental toxicity are available for risk assessment. In the present study, a new developmental database was compiled by combining toxicity information from the Teratogen Information System (TERIS) and the Food and Drug Administration (FDA) guidelines. We implemented a decision tree modeling procedure, using Classification and Regression Tree software and a model ensemble approach termed bagging. We then assessed the empirical distributions of the prediction accuracy measures of the single and ensemble-based models, achieved by repeating our modeling experiment many times by repeated random partitioning of the working database. The decision tree developmental SAR models exhibited modest prediction accuracy. Bagging tended to enhance the accuracy of prediction. Also, the model ensemble approach reduced the variability of prediction measures compared to the single model approach. Further research with data derived from animal species- and endpoint-specific components of an extended and refined FDA/TERIS database has the potential to derive SAR models that would be useful in the developmental risk assessment of the thousands of untested chemicals.     

Structure–activity relationship models for rat carcinogenesis and assessing the role mutagens play in model predictivity

We previously demonstrated that fragment based cat-SAR carcinogenesis models consisting solely of mutagenic or non-mutagenic carcinogens varied greatly in terms of their predictive accuracy. This led us to investigate how well the rat cancer cat-SAR model predicted mutagens and non-mutagens in their learning set. Four rat cancer cat-SAR models were developed: Complete Rat, Transgender Rat, Male Rat and Female Rat, with leave-one-out (LOO) validation concordance values of 69%, 74%, 67% and 73%, respectively. The mutagenic carcinogens produced concordance values in the range 69–76% compared with only 47–53% for non-mutagenic carcinogens. As a surrogate for mutagenicity, comparisons between single site and multiple site carcinogen SAR models were analysed. The LOO concordance values for models consisting of 1-site, 2-site and 4+-site carcinogens were 66%, 71% and 79%, respectively. As expected, the proportion of mutagens to non-mutagens also increased, rising from 54% for 1-site to 80% for 4+-site carcinogens. This study demonstrates that mutagenic chemicals, in both SAR learning sets and test sets, are influential in assessing model accuracy. This suggests that SAR models for carcinogens may require a two-step process in which mutagenicity is first determined before carcinogenicity can be accurately predicted.     

Integration of structure-activity relationship and artificial intelligence systems to improve in silico prediction of ames test mutagenicity

The Ames mutagenicity test in Salmonella typhimurium is a bacterial short-term in vitro assay aimed at detecting the mutagenicity caused by chemicals. Mutagenicity is considered as an early alert for carcinogenicity. After a number of decades, several (Q)SAR studies on this endpoint yielded enough evidence to make feasible the construction of reliable computational models for prediction of mutagenicity from the molecular structure of chemicals. In this study, we propose a combination of a fragment-based SAR model and an inductive database. The hybrid system was developed using a collection of 4337 chemicals (2401 mutagens and 1936 nonmutagens) and tested using 753 independent compounds (437 mutagens and 316 nonmutagens). The overall error of this system on the external test set compounds is 15% (sensitivity = 15%, specificity = 15%), which is quantitatively similar to the experimental error of Ames test data (average interlaboratory reproducibility determined by the National Toxicology Program). Moreover, each single prediction is provided with a specific confidence level. The results obtained give confidence that this system can be applied to support early and rapid evaluation of the level of mutagenicity concern.     

Local and global quantitative structure-activity relationship modeling and prediction for the baseline toxicity

The predictive accuracy of the model is of the most concern for computational chemists in quantitative structure-activity relationship (QSAR) investigations. It is hypothesized that the model based on analogical chemicals will exhibit better predictive performance than that derived from diverse compounds. This paper develops a novel scheme called "clustering first, and then modeling" to build local QSAR models for the subsets resulted from clustering of the training set according to structural similarity. For validation and prediction, the validation set and test set were first classified into the corresponding subsets just as those of the training set, and then the prediction was performed by the relevant local model for each subset. This approach was validated on two independent data sets by local modeling and prediction of the baseline toxicity for the fathead minnow. In this process, hierarchical clustering was employed for cluster analysis, k-nearest neighbor for classification, and partial least squares for the model generation. The statistical results indicated that the predictive performances of the local models based on the subsets were much superior to those of the global model based on the whole training set, which was consistent with the hypothesis. This approach proposed here is promising for extension to QSAR modeling for various physicochemical properties, biological activities, and toxicities.     

Bayesian analysis and inference from QSAR predictive model results

QSAR models have been under development for decades but acceptance and utilization of model results have been slow, in part, because there is no widely accepted metric for assessing their reliability. We reapply a method commonly used in quantitative epidemiology and medical decision-making for evaluating the results of screening tests to assess reliability of a QSAR model. It quantifies the accuracy (expressed as sensitivity and specificity) of QSAR models as conditional probabilities of correct and incorrect classification of chemical characteristic, given a true characteristic. Using Bayes formula, these conditional probabilities are combined with prior information to generate a posterior distribution to determine the probability a specific chemical has a particular characteristic, given a model prediction. As an example, we apply this approach to evaluate the predictive reliability of a CATABOL model and base on it a "ready" and "not ready" biodegradability classification. Finally, we show how predictive capability of the model can be improved by sequential use of two models, the first one with high sensitivity and the second with high specificity.     

QSAR modeling of large heterogeneous sets of molecules

In aquatic toxicology, QSAR models are generally designed for chemicals presenting the same mode of toxic action. Their proper use provides good simulation results. Problems arise when the mechanism of toxicity of a chemical is not clearly identified. Indeed, in that case, the inappropriate application of a specific QSAR model can lead to a dramatic error in the toxicity estimation. With the advent of powerful computers and easy access to them, and the introduction of soft modeling and artificial intelligence in SAR and QSAR, radically different models, designed from large noncongeneric sets of chemicals have been proposed. Some of these new QSAR models are reviewed and their originality, advantages, and limitations are stressed.     

Experimental determination and prediction of bilitranslocase transport activity

The transport activity of a membrane protein, bilitranslocase (T.C. # 2.A.65.1.1), which acts as a transporter of bilirubin from blood to liver cells, was experimentally determined for a large set of various endogenous compounds, drugs, purine and pyrimidine derivatives. On these grounds, the structure-activity models were developed following the OECD principles of QSAR models and their predictive ability for new chemicals was evaluated. The applicability domain of the models was estimated by Euclidean distances criteria according to the applied modeling method. The selection of the most influential structural variables was an important stage in the adopted modeling methodology. The interpretation of selected variables was performed in order to get an insight into the mechanism of transport through the cell membrane via bilitranslocase. Validation of the optimized models was performed by a previously determined validation set. The classification model was build to separate active from inactive compounds. The resulting accuracy, sensitivity, and specificity were 0.73, 0.89, and 0.64, respectively. Only active compounds were used to develop a predictive model for bilitranslocase inhibition constants. The model showed good predictive ability; Root Mean Squared error of the validation set, RMS(V)=0.29 log units.     

Qualitative and quantitative structure–activity relationship modelling for predicting blood-brain barrier permeability of structurally diverse chemicals

In this study, structure–activity relationship (SAR) models have been established for qualitative and quantitative prediction of the blood–brain barrier (BBB) permeability of chemicals. The structural diversity of the chemicals and nonlinear structure in the data were tested. The predictive and generalization ability of the developed SAR models were tested through internal and external validation procedures. In complete data, the QSAR models rendered ternary classification accuracy of >98.15%, while the quantitative SAR models yielded correlation (r²) of >0.926 between the measured and the predicted BBB permeability values with the mean squared error (MSE) <0.045. The proposed models were also applied to an external new in vitro data and yielded classification accuracy of >82.7% and r² > 0.905 (MSE < 0.019). The sensitivity analysis revealed that topological polar surface area (TPSA) has the highest effect in qualitative and quantitative models for predicting the BBB permeability of chemicals. Moreover, these models showed predictive performance superior to those reported earlier in the literature. This demonstrates the appropriateness of the developed SAR models to reliably predict the BBB permeability of new chemicals, which can be used for initial screening of the molecules in the drug development process.     

QSAR Modeling of Large Heterogeneous Sets of Molecules

Abstract

In aquatic toxicology, QSAR models are generally designed for chemicals presenting the same mode of toxic action. Their proper use provides good simulation results. Problems arise when the mechanism of toxicity of a chemical is not clearly identified. Indeed, in that case, the inappropriate application of a specific QSAR model can lead to a dramatic error in the toxicity estimation. With the advent of powerful computers and easy access to them, and the introduction of soft modeling and artificial intelligence in SAR and QSAR, radically different models, designed from large non-congeneric sets of chemicals have been proposed. Some of these new QSAR models are reviewed and their originality, advantages, and limitations are stressed.     

Bayesian analysis and inference from QSAR predictive model results

QSAR models have been under development for decades but acceptance and utilization of model results have been slow, in part, because there is no widely accepted metric for assessing their reliability. We reapply a method commonly used in quantitative epidemiology and medical decision-making for evaluating the results of screening tests to assess reliability of a QSAR model. It quantifies the accuracy (expressed as sensitivity and specificity) of QSAR models as conditional probabilities of correct and incorrect classification of chemical characteristic, given a true characteristic. Using Bayes formula, these conditional probabilities are combined with prior information to generate a posterior distribution to determine the probability a specific chemical has a particular characteristic, given a model prediction. As an example, we apply this approach to evaluate the predictive reliability of a CATABOL model and base on it a "ready" and "not ready" biodegradability classification. Finally, we show how predictive capability of the model can be improved by sequential use of two models, the first one with high sensitivity and the second with high specificity.