Quantitative structure carcinogenicity relationship for detecting structural alerts in nitroso-compounds

Machine learning-assisted data filtering and QSAR models for prediction of chemical acute toxicity on rat and mouse

Machine learning (ML) methods provide a new opportunity to build quantitative structure-activity relationship (QSAR) models for predicting chemicals' toxicity based on large toxicity data sets, but they are limited in insufficient model robustness due to poor data set quality for chemicals with certain structures. To address this issue and improve model robustness, we built a large data set on rat oral acute toxicity for thousands of chemicals, then used ML to filter chemicals favorable for regression models (CFRM). In comparison to chemicals not favorable for regression models (CNRM), CFRM accounted for 67% of chemicals in the original data set, and had a higher structural similarity and a smaller toxicity distribution in 2-4 log₁₀ (mg/kg). The performance of established regression models for CFRM was greatly improved, with root-mean-square deviations (RMSE) in the range of 0.45-0.48 log₁₀ (mg/kg). Classification models were built for CNRM using all chemicals in the original data set, and the area under receiver operating characteristic (AUROC) reached 0.75-0.76. The proposed strategy was successfully applied to a mouse oral acute data set, yielding RMSE and AUROC in the range of 0.36-0.38 log₁₀ (mg/kg) and 0.79, respectively.

Mutagenic impurities in pharmaceuticals: A critical assessment of the cohort of concern with a focus on N-nitrosamines

The TTC (Threshold of Toxicological Concern; set at 1.5 μg/day for pharmaceuticals) defines an acceptable patient intake for any unstudied chemical posing a negligible risk of carcinogenicity or other toxic effects. A group of high potency mutagenic carcinogens, defined solely by the presence of particular structural alerts, are referred to as the "cohort of concern" (CoC); aflatoxin-like-, N-nitroso-, and alkyl-azoxy compounds are considered to pose a significant carcinogenic risk at intakes below the TTC. Kroes et al.2004, derived values for the TTC and CoC in the context of food components, employing a non-transparent dataset never placed in the public domain. Using a reconstructed all-carcinogen dataset from relevant publications, it is now clear that there are exceptions for all three CoC structural classes. N-Nitrosamines represent 62% of the N-nitroso class in the reconstructed dataset. Employing a contemporary dataset, 20% are negative in rodent carcinogenicity bioassays with less than 50% of N-nitrosamines estimated to fall into the highest risk category. It is recommended that CoC nitrosamines are identified by compound-specific data rather than structural alerts. Thus, it should be possible to distinguish CoC from non-CoC N-nitrosamines in the context of mutagenic impurities described in ICH M7 (R1).

Application of a developed triple-classification machine learning model for carcinogenic prediction of hazardous organic chemicals to the US, EU, and WHO based on Chinese database

Cancer, the second largest human disease, has become a major public health problem. The prediction of chemicals' carcinogenicity before their synthesis is crucial. In this paper, seven machine learning algorithms (i.e., Random Forest (RF), Logistic Regression (LR), Support Vector Machines (SVM), Complement Naive Bayes (CNB), K-Nearest Neighbor (KNN), XGBoost, and Multilayer Perceptron (MLP)) were used to construct the carcinogenicity triple classification prediction (TCP) model (i.e., 1A, 1B, Category 2). A total of 1444 descriptors of 118 hazardous organic chemicals were calculated by Discovery Studio 2020, Sybyl X-2.0 and PaDEL-Descriptor software. The constructed carcinogenicity TCP model was evaluated through five model evaluation indicators (i.e., Accuracy, Precision, Recall, F1 Score and AUC). The model evaluation results show that Accuracy, Precision, Recall, F1 Score and AUC evaluation indicators meet requirements (greater than 0.6). The accuracy of RF, LR, XGBoost, and MLP models for predicting carcinogenicity of Category 2 is 91.67%, 79.17%, 100%, and 100%, respectively. In addition, the constructed machine learning model in this study has potential for error correction. Taking XGBoost model as an example, the predicted carcinogenicity level of 1,2,3-Trichloropropane (96-18-4) is Category 2, but the actual carcinogenicity level is 1B. But the difference between Category 2 and 1B is only 0.004, indicating that the XGBoost is one optimum model of the seven constructed machine learning models. Besides, results showed that functional groups like chlorine and benzene ring might influence the prediction of carcinogenic classification. Therefore, considering functional group characteristics of chemicals before constructing the carcinogenicity prediction model of organic chemicals is recommended. The predicted carcinogenicity of the organic chemicals using the optimum machine leaning model (i.e., XGBoost) was also evaluated and verified by the toxicokinetics. The RF and XGBoost TCP models constructed in this paper can be used for carcinogenicity detection before synthesizing new organic substances. It also provides technical support for the subsequent management of organic chemicals.

Quantum-Mechanical Approach to Predicting the Carcinogenic Potency of N-Nitroso Impurities in Pharmaceuticals

N-Nitroso contaminants in medicinal products are of concern due to their high carcinogenic potency; however, not all these compounds are created equal, and some are relatively benign chemicals. Understanding the structure-activity relationships (SARs) that drive hazards in one molecule versus another is key to both protecting human health and alleviating costly and sometimes inaccurate animal testing. Here, we report on an extension of the CADRE (computer-aided discovery and REdesign) platform, which is used broadly by the pharmaceutical and personal care industries to assess environmental and human health endpoints, to predict the carcinogenic potency of N-nitroso compounds. The model distinguishes compounds in three potency categories with 77% accuracy in external testing, which surpasses the reproducibility of rodent cancer bioassays and constraints imposed by limited (high-quality) data. The robustness of predictions for more complex pharmaceuticals is maximized by capturing key SARs using quantum mechanics, that is, by hinging the model on the underlying chemistry versus chemicals in the training set. To this end, the present approach can be leveraged in a quantitative hazard assessment and to offer qualitative guidance using electronic structure comparisons between well-studied analogues and unknown contaminants.

Systematic QSAR and iQCCR modelling of fused/non-fused aromatic hydrocarbons (FNFAHs) carcinogenicity to rodents: reducing unnecessary chemical synthesis and animal testing

The prediction of new or untested FNFAHs will reduce unnecessary chemical synthesis and animal testing, and contribute to the design of safer chemicals for production activities.

Use of less-than-lifetime (LTL) durational limits for nitrosamines: Case study of N-Nitrosodiethylamine (NDEA)

The ICH M7(R1) guideline describes a framework to assess the carcinogenic risk of mutagenic and carcinogenic pharmaceutical impurities following less-than-lifetime (LTL) exposures. This LTL framework is important as many pharmaceuticals are not administered for a patient's lifetime and as clinical trials typically involve LTL exposures. While there has been regulatory caution about applying LTL concepts to cohort of concern (COC) impurities such as N-nitrosamines, ICH M7 does not preclude this and indeed literature data suggests that the LTL framework will be protective of patient safety for N-nitrosamines. The goal was to investigate if applying the LTL framework in ICH M7 would control exposure to an acceptable excess cancer risk in humans. Using N-nitrosodiethylamine as a case study, empirical data correlating exposure duration (as a percentage of lifespan) and cancer incidence in rodent bioassays indicate that the LTL acceptable intake (AI) as derived using the ICH M7 framework would not exceed a negligible additional risk of cancer. Therefore, controlling N-nitrosamines to an LTL AI based on the ICH M7 framework is thus demonstrated to be protective for potential carcinogenic risk to patients over the exposure durations typical of clinical trials and many prescribed medicines.

In vivo toxicity of nitroaromatic compounds to rats: QSTR modelling and interspecies toxicity relationship with mouse

Nitroaromatic compounds (NACs) in the environment can cause serious public health and environmental problems due to their potential toxicity. This study established quantitative structure-toxicity relationship (QSTR) models for the acute oral toxicity of NACs towards rats following the stringent OECD principles for QSTR modelling. All models were assessed by various internationally accepted validation metrics and the OECD criteria. The best QSTR model contains seven simple and interpretable 2D descriptors with defined physicochemical meaning. Mechanistic interpretation indicated that van der Waals surface area, presence of C-F at topological distance 6, heteroatom content and frequency of C-N at topological distance 9 are main factors responsible for the toxicity of NACs. This proposed model was successfully applied to a true external set (295 compounds), and prediction reliability was analysed and discussed. Moreover, the rat-mouse and mouse-rat interspecies quantitative toxicity-toxicity relationship (iQTTR) models were also constructed, validated and employed in toxicity prediction for true external sets consisting of 67 and 265 compounds, respectively. These models showed good external predictivity that can be used to rapidly predict the rat oral acute toxicity of new or untested NACs falling within the applicability domain of the models, thus being beneficial in environmental risk assessment and regulatory purposes.

Deep Learning Based Regression and Multiclass Models for Acute Oral Toxicity Prediction with Automatic Chemical Feature Extraction

Median lethal death, LD₅₀, is a general indicator of compound acute oral toxicity (AOT). Various in silico methods were developed for AOT prediction to reduce costs and time. In this study, we developed an improved molecular graph encoding convolutional neural networks (MGE-CNN) architecture to construct three types of high-quality AOT models: regression model (deepAOT-R), multiclassification model (deepAOT-C), and multitask model (deepAOT-CR). These predictive models highly outperformed previously reported models. For the two external data sets containing 1673 (test set I) and 375 (test set II) compounds, the R² and mean absolute errors (MAEs) of deepAOT-R on the test set I were 0.864 and 0.195, and the prediction accuracies of deepAOT-C were 95.5% and 96.3% on test sets I and II, respectively. The two external prediction accuracies of deepAOT-CR are 95.0% and 94.1%, while the R² and MAE are 0.861 and 0.204 for test set I, respectively. We then performed forward and backward exploration of deepAOT models for deep fingerprints, which could support shallow machine learning methods more efficiently than traditional fingerprints or descriptors. We further performed automatic feature learning, a key essence of deep learning, to map the corresponding activation values into fragment space and derive AOT-related chemical substructures by reverse mining of the features. Our deep learning architecture for AOT is generally applicable in predicting and exploring other toxicity or property end points of chemical compounds. The two deepAOT models are freely available at http://repharma.pku.edu.cn/DLAOT/DLAOThome.php or http://www.pkumdl.cn/DLAOT/DLAOThome.php .

Correlation between physicochemical properties of modified clinoptilolite and its performance in the removal of ammonia-nitrogen

The physicochemical properties of the 24 modified clinoptilolite samples and their ammonia-nitrogen removal rates were measured to investigate the correlation between them. The modified clinoptilolites obtained by acid modification, alkali modification, salt modification, and thermal modification were used to adsorb ammonia-nitrogen. The surface area, average pore width, macropore volume, mecropore volume, micropore volume, cation exchange capacity (CEC), zeta potential, silicon-aluminum ratios, and ammonia-nitrogen removal rate of the 24 modified clinoptilolite samples were measured. Subsequently, the linear regression analysis method was used to research the correlation between the physicochemical property of the different modified clinoptilolite samples and the ammonia-nitrogen removal rate. Results showed that the CEC was the major physicochemical property affecting the ammonia-nitrogen removal performance. According to the impacts from strong to weak, the order was CEC > silicon-aluminum ratios > mesopore volume > micropore volume > surface area. On the contrary, the macropore volume, average pore width, and zeta potential had a negligible effect on the ammonia-nitrogen removal rate. The relational model of physicochemical property and ammonia-nitrogen removal rate of the modified clinoptilolite was established, which was ammonia-nitrogen removal rate = 1.415[CEC] + 173.533 [macropore volume] + 0.683 [surface area] + 4.789[Si/Al] - 201.248. The correlation coefficient of this model was 0.982, which passed the validation of regression equation and regression coefficients. The results of the significance test showed a good fit to the correlation model.

QSAR-Based Studies of Nanomaterials in the Environment

Nanotechnology is a newly emerging field, posing substantial impacts on society, economy, and the environment. In recent years, the development of nanotechnology has led to the design and large-scale production of many new materials and devices with a vast range of applications. However, along with the benefits, the use of nanomaterials raises many questions and generates concerns due to the possible health-risks and environmental impacts. This chapter provides an overview of the Quantitative Structure-Activity Relationships (QSAR) studies performed so far towards predicting nanoparticles' environmental toxicity. Recent progresses on the application of these modeling studies are additionally pointed out. Special emphasis is given to the setup of a QSAR perturbation-based model for the assessment of ecotoxic effects of nanoparticles in diverse conditions. Finally, ongoing challenges that may lead to new and exciting directions for QSAR modeling are discussed.

Mathematical modeling for local trans-round window membrane drug transport in the inner ear

The structure and composition of the round window membrane (RWM) make it a particularly effective pathway for drug delivery to the inner ear. Therefore, predicting the efficiency of RWM transport would provide useful information for enhancing local application. In the present study, a mathematical model was established to achieve this goal. A series of drugs with different physicochemical properties were introduced in the inner ear cavity of guinea pigs via RWM by intratympanic application. The perilymphatic drug concentration (C) data were used to calculate the permeability coefficient (K_p) of different drugs diffusing through the RWM. The experimental data were fitted using the Matlab software to set up the numerical model based on Fick's diffusion law and the single-compartment model following extravascular administration, which facilitated the prediction of the permeation profiles of different drugs while trans-RWM. In summary, this mathematical model is a contribution toward developing potentially useful RWM administration simulating tools.

Computational modeling in nanomedicine: prediction of multiple antibacterial profiles of nanoparticles using a quantitative structure-activity relationship perturbation model

AIMS

We introduce the first quantitative structure-activity relationship (QSAR) perturbation model for probing multiple antibacterial profiles of nanoparticles (NPs) under diverse experimental conditions.

MATERIALS & METHODS

The dataset is based on 300 nanoparticles containing dissimilar chemical compositions, sizes, shapes and surface coatings. In general terms, the NPs were tested against different bacteria, by considering several measures of antibacterial activity and diverse assay times. The QSAR perturbation model was created from 69,231 nanoparticle-nanoparticle (NP-NP) pairs, which were randomly generated using a recently reported perturbation theory approach.

RESULTS

The model displayed an accuracy rate of approximately 98% for classifying NPs as active or inactive, and a new copper-silver nanoalloy was correctly predicted by this model with consensus accuracy of 77.73%.

CONCLUSION

Our QSAR perturbation model can be used as an efficacious tool for the virtual screening of antibacterial nanomaterials.

Chemoinformatics for medicinal chemistry: in silico model to enable the discovery of potent and safer anti-cocci agents

Background: Gram-positive cocci are increasingly antibiotic-resistant bacteria responsible for causing serious diseases. Chemoinformatics can help to rationalize the discovery of more potent and safer antibacterial drugs. We have developed a chemoinformatic model for simultaneous prediction of anti-cocci activities, and profiles involving absorption, distribution, metabolism, elimination and toxicity (ADMET). Results: A dataset containing 48,874 cases from many different chemicals assayed under dissimilar experimental conditions was created. The best model displayed accuracies around 93% in both training and prediction (test) sets. Quantitative contributions of several fragments to the biological effects were calculated and analyzed. Multiple biological effects of the investigational drug JNJ-Q2 were correctly predicted. Conclusion: Our chemoinformatic model can be used as powerful tool for virtual screening of promising anti-cocci agents.

Predicting carcinogenicity and understanding the carcinogenic mechanism of N-nitroso compounds using a TOPS-MODE approach

A linear discriminant analysis (LDA) coupled with an enhanced replacement method (ERM) was used as an alternative method to predict the carcinogenicity of N-nitroso compounds (NOCs) in rats. This presented LDA based on the topological substructural molecular descriptors (TOPS-MODE) approach was developed to predict the carcinogenic and noncarcinogenic activity on a data set of 111 NOCs with a good classification value of 90.1%. The predictive power of the LDA model was validated through an external validation set (37 compounds) with a prediction accuracy of 94.6% and a leave-one-out cross-validation procedure (LOOCV) with a good prediction of 86.5%. This methodology showed that the TOPS-MODE descriptors weighted, respectively, by bond dipole moment and Abraham solute descriptor dipolarity/polarizability affected the NOC carcinogenicity. The contributions of certain bonds and fragments to carcinogenicity were used to assess biotransformation and carcinogenic mechanisms. The positive contribution of the carbon-nitrogen single bond (between the N-nitroso group and α-carbon to the N-nitroso group) indicated that the α-hydroxylation reaction could occur at the α-carbon or otherwise not occur. Similarly, the contributions from the molecular fragment could be applied to indicate whether the fragments generated an alkylating agent. These results suggested that this approach could discriminate between carcinogenic and noncarcinogenic NOCs, thereby providing insight into the structural features and chemical factors related to NOC carcinogenicity.

Quantitative structure-activity relationship modelling of the carcinogenic risk of nitroso compounds using regression analysis and the TOPS-MODE approach

Worldwide, legislative and governmental efforts are focusing on establishing simple screening tools for identifying those chemicals most likely to cause adverse effects without experimentally testing all chemicals of regulatory concern. This is because even the most basic biological testing of compounds of concern, apart from requiring a huge number of test animals, would be neither resource nor time effective. Thus, alternative approaches such as the one proposed here, quantitative structure-activity relationship (QSAR) modelling, are increasingly being used for identifying the potential health hazards and subsequent regulation of new industrial chemicals. This paper follows up on our earlier work that demonstrated the use of the TOPological Substructural MOlecular DEsign (TOPS-MODE) approach to QSAR modelling for predictions of the carcinogenic potency of nitroso compounds. The data set comprises 56 nitroso compounds which have been bio-assayed in female rats and administered by the oral water route. The QSAR model was able to account for about 81% of the variance in the experimental activity and exhibited good cross-validation statistics. A reasonable interpretation of the TOPS-MODE descriptors was achieved by means of bond contributions, which in turn afforded the recognition of structural alerts (SAs) regarding carcinogenicity. A comparison of the SAs obtained from different data sets showed that experimental factors, such as the sex and the oral administration route, exert a major influence on the carcinogenicity of nitroso compounds. The present and previous QSAR models combined together provide a reliable tool for estimating the carcinogenic potency of yet untested nitroso compounds and they should allow the identification of SAs, which can be used as the basis of prediction systems for the rodent carcinogenicity of these compounds.

A topological substructural molecular design approach for predicting mutagenesis end-points of alpha, beta-unsaturated carbonyl compounds

Chemically reactive, alpha, beta-unsaturated carbonyl compounds are common environmental pollutants able to produce a wide range of adverse effects, including, e.g. mutagenicity. This toxic property can often be related to chemical structure, in particular to specific molecular substructures or fragments (alerts), which can then be used in specialized software or expert systems for predictive purposes. In the past, there have been many attempts to predict the mutagenicity of alpha, beta-unsaturated carbonyl compounds through quantitative structure activity relationships (QSAR) but considering only one exclusive endpoint: the Ames test. Besides, even though those studies give a comprehensive understanding of the phenomenon, they do not provide substructural information that could be useful forward improving expert systems based on structural alerts (SAs). This work reports an evaluation of classification models to probe the mutagenic activity of alpha, beta-unsaturated carbonyl compounds over two endpoints--the Ames and mammalian cell gene mutation tests--based on linear discriminant analysis along with the topological Substructure molecular design (TOPS-MODE) approach. The obtained results showed the better ability of the TOPS-MODE approach in flagging structural alerts for the mutagenicity of these compounds compared to the expert system TOXTREE. Thus, the application of the present QSAR models can aid toxicologists in risk assessment and in prioritizing testing, as well as in the improvement of expert systems, such as the TOXTREE software, where SAs are implemented.

Rapid toxicity prediction of organic chemicals to Chlorella vulgaris using quantitative structure-activity relationships methods

This paper presents the results of an optimization study on the toxicity of 91 aliphatic and aromatic compounds as well as a small subset of pesticides to algae Chlorella vulgaris, which was accomplished by using quantitative structure-activity relationships (QSAR). The linear (HM) and the nonlinear method radial basis function neural networks (RBFNN) were used to develop the QSAR models and both of them can give satisfactory prediction results. At the same time, by interpreting the descriptors, we can get some insight into structural features (molecular surface area, electrostatic repulsion, and hydrogen bonds) related to the toxic action. Finally, a detailed analysis on the model application domain defined the compounds, whose estimation can be accepted with confidence. The results of this study suggest that the proposed approaches could be successfully used as a general tool for the estimate of novel toxic compounds.

QSAR modelling of carcinogenicity by balance of correlations

Optimal descriptors based on the simplified molecular input line entry system (SMILES) have been utilized in modeling of carcinogenicity. Carcinogenicity of 401 compounds has been modeled by means of balance of correlations for the training (n = 170) and calibration (n = 170) sets. The obtained models were evaluated with an external test set (n = 61). Comparison of models based on the balance of correlations and models which were obtained on the basis of the total training set (i.e., both training and calibration sets as the united training set) has shown that the balance of correlations improves the statistical quality for the external test set.