Chemical rules for optimization of chemical mutagenicity via matched molecular pairs analysis and machine learning methods.

Chemical mutagenicity is a serious issue that needs to be addressed in early drug discovery. Over a long period of time, medicinal chemists have manually summarized a series of empirical rules for the optimization of chemical mutagenicity. However, given the rising amount of data, it is getting more difficult for medicinal chemists to identify more comprehensive chemical rules behind the biochemical data. Herein, we integrated a large Ames mutagenicity data set with 8576 compounds to derive mutagenicity transformation rules for reversing Ames mutagenicity via matched molecular pairs analysis. A well-trained consensus model with a reasonable applicability domain was constructed, which showed favorable performance in the external validation set with an accuracy of 0.815. The model was used to assess the generalizability and validity of these mutagenicity transformation rules. The results demonstrated that these rules were of great value and could provide inspiration for the structural modifications of compounds with potential mutagenic effects. We also found that the local chemical environment of the attachment points of rules was critical for successful transformation. To facilitate the use of these mutagenicity transformation rules, we integrated them into ADMETopt2 ( http://lmmd.ecust.edu.cn/admetsar2/admetopt2/ ), a free web server for optimization of chemical ADMET properties. The above-mentioned approach would be extended to the optimization of other toxicity endpoints.

In silico prediction of UGT-mediated metabolism in drug-like molecules via graph neural network.

UDP-glucuronosyltransferases (UGTs) have gained increasing attention as they play important roles in the phase II metabolism of drugs. Due to the time-consuming process and high cost of experimental approaches to identify the metabolic fate of UGT enzymes, in silico methods have been developed to predict the UGT-mediated metabolism of drug-like molecules. We developed consensus models with the combination of machine learning (ML) and graph neural network (GNN) methods to predict if a drug-like molecule is a potential UGT substrate, and then we applied the Weisfeiler-Lehman Network (WLN) model to identify the sites of metabolism (SOMs) of UGT-catalyzed substrates. For the substrate model, the accuracy of the single substrate prediction model on the test set could reach to 0.835. Compared with the single estimators, the consensus models are more stable and have better generalization ability, and the accuracy on the test set reached to 0.851. For the SOM model, the top-1 accuracy of the SOM model on the test set reached to 0.898, outperforming existing works. Thus, in this study, we proposed a computational framework, named Meta-UGT, which would provide a useful tool for the prediction and optimization of metabolic profiles and drug design.

IDL-PPBopt: A Strategy for Prediction and Optimization of Human Plasma Protein Binding of Compounds via an Interpretable Deep Learning Method.

The prediction and optimization of pharmacokinetic properties are essential in lead optimization. Traditional strategies mainly depend on the empirical chemical rules from medicinal chemists. However, with the rising amount of data, it is getting more difficult to manually extract useful medicinal chemistry knowledge. To this end, we introduced IDL-PPBopt, a computational strategy for predicting and optimizing the plasma protein binding (PPB) property based on an interpretable deep learning method. At first, a curated PPB data set was used to construct an interpretable deep learning model, which showed excellent predictive performance with a root mean squared error of 0.112 for the entire test set. Then, we designed a detection protocol based on the model and Wilcoxon test to identify the PPB-related substructures (named privileged substructures, PSubs) for each molecule. In total, 22 general privileged substructures (GPSubs) were identified, which shared some common features such as nitrogen-containing groups, diamines with two carbon units, and azetidine. Furthermore, a series of second-level chemical rules for each GPSub were derived through a statistical test and then summarized into substructure pairs. We demonstrated that these substructure pairs were equally applicable outside the training set and accordingly customized the structural modification schemes for each GPSub, which provided alternatives for the optimization of the PPB property. Therefore, IDL-PPBopt provides a promising scheme for the prediction and optimization of the PPB property and would be helpful for lead optimization of other pharmacokinetic properties.

In silico prediction of chemical aquatic toxicity for marine crustaceans via machine learning.

Aquatic toxicity is a crucial endpoint for evaluating chemically adverse effects on ecosystems. Therefore, we developed in silico methods for the prediction of chemical aquatic toxicity in marine environment. At first, a diverse data set including different crustacean species was constructed. We then built local binary models using Mysidae data and global binary models using Mysidae, Palaemonidae, and Penaeidae data. Molecular fingerprints and descriptors were employed to represent chemical structures separately. All the models were built by six machine learning methods. The AUC (area under the receiver operating characteristic curve) values of the better local and global models were around 0.8 and 0.9 for the test sets, respectively. We also identified several chemicals with selective toxicity on different species. The analysis of selective toxicity would promote to design greener chemicals in a specific environment. Finally, to understand and interpret the models, we explored the relationships between chemical aquatic toxicity and the molecular descriptors. Our study would be helpful in gaining further insights into marine organisms, prediction of chemical aquatic toxicity and prioritization of environmental hazard assessment.

Computational Prediction of Site of Metabolism for UGT-Catalyzed Reactions.

The investigation of metabolically liable sites of xenobiotics mediated by UDP-glucuronosyltransferases (UGTs) is important for lead optimization in early drug discovery. However, it is time-consuming and costly to identify potentially susceptible sites experimentally. Hence, in silico approaches have been developed to predict the site of metabolism (SOM) of UGT-catalyzed substrates. In the present work, four major types of reactions catalyzed by UGTs were collected from the Handbook of Metabolic Pathways of Xenobiotics along with their corresponding glucuronidation products. These observed and nonobserved SOMs were identified and encoded by atom environment fingerprints. Four resampling methods were performed to treat data imbalance, while four feature selection methods were utilized to choose appropriate features. Three tree-form machine learning algorithms were conducted to build discriminating models, and optimal models were then obtained for the different types of reaction. The results indicated that all of the chosen best models showed area under the curve values ranging from 0.713 to 0.869 for two independent external validation sets. Our study could supply valuable information for optimization of pharmacokinetic profiles and contribute to metabolism prediction.

In silico estimation of chemical aquatic toxicity on crustaceans using chemical category methods.

With industrial development and eventual commercial use, environmental chemicals through accidental spills and effluents appear more frequently in aquatic ecosystems and may produce an enormous effect on water, soil, wildlife and human health. Therefore, aquatic toxicity becomes an increasingly important endpoint in the evaluation of the environmental impact of chemicals. In this study, based on ECOTOX database, a large data set containing 824 diverse compounds with experimental 48 h EC50 values on crustaceans was compiled. A series of in silico models were then developed using six machine learning methods combined with seven types of molecular fingerprints. Performance of these models was measured by an external validation set, involving 246 molecules. The best model proposed is MACCS fingerprint and SVM algorithm with high accuracy of 0.87 for external validation set. Additionally, we proposed five structural alerts identified by information gain and substructure frequency analysis for mechanistic interpretation. The models and structural alerts can provide critical information and useful tools for a priori evaluation of chemical aquatic toxicity in environmental hazard assessment.

Multiclassification Prediction of Enzymatic Reactions for Oxidoreductases and Hydrolases Using Reaction Fingerprints and Machine Learning Methods.

Drug metabolism is a complex procedure in the human body, including a series of enzymatically catalyzed reactions. However, it is costly and time consuming to investigate drug metabolism experimentally; computational methods are hence developed to predict drug metabolism and have shown great advantages. As the first step, classification of metabolic reactions and enzymes is highly desirable for drug metabolism prediction. In this study, we developed multiclassification models for prediction of reaction types catalyzed by oxidoreductases and hydrolases, in which three reaction fingerprints were used to describe the reactions and seven machine learnings algorithms were employed for model building. Data retrieved from KEGG containing 1055 hydrolysis and 2510 redox reactions were used to build the models, respectively. The external validation data consisted of 213 hydrolysis and 512 redox reactions extracted from the Rhea database. The best models were built by neural network or logistic regression with a 2048-bit transformation reaction fingerprint. The predictive accuracies of the main class, subclass, and superclass classification models on external validation sets were all above 90%. This study will be very helpful for enzymatic reaction annotation and further study on metabolism prediction.

In silico Prediction of Drug Induced Liver Toxicity Using Substructure Pattern Recognition Method.

Drug-induced liver injury (DILI) is a leading cause of acute liver failure in the US and less severe liver injury worldwide. It is also one of the major reasons of drug withdrawal from the market. Thus, DILI has become one of the most important concerns of drugs, and should be predicted in very early stage of drug discovery process. In this study, a comprehensive data set containing 1317 diverse compounds was collected from publications. Then, high accuracy classification models were built using five machine learning methods based on MACCS and FP4 fingerprints after evaluating by substructure pattern recognition method. The best model was built using SVM method together with FP4 fingerprint at the IG value threshold of 0.0005. Its overall predictive accuracies were 79.7 % and 64.5 % for the training and test sets, separately, which yielded overall accuracy of 75.0 % for the external validation dataset, consisting of 88 compounds collected from a benchmark DILI database - the Liver Toxicity Knowledge Base. This model could be used for drug-induced liver toxicity prediction. Moreover, some key substructure patterns correlated with drug-induced liver toxicity were also identified as structural alerts.

In silico prediction of hERG potassium channel blockage by chemical category approaches.

The human ether-a-go-go related gene (hERG) plays an important role in cardiac action potential. It encodes an ion channel protein named Kv11.1, which is related to long QT syndrome and may cause avoidable sudden cardiac death. Therefore, it is important to assess the hERG channel blockage of lead compounds in an early drug discovery process. In this study, we collected a large data set containing 1163 diverse compounds with IC50 values determined by the patch clamp method on mammalian cell lines. The whole data set was divided into 80% as the training set and 20% as the test set. Then, five machine learning methods were applied to build a series of binary classification models based on 13 molecular descriptors, five fingerprints and molecular descriptors combining fingerprints at four IC50 thresholds to discriminate hERG blockers from nonblockers, respectively. Models built by molecular descriptors combining fingerprints were validated by using an external validation set containing 407 compounds collected from the hERGCentral database. The performance indicated that the model built by molecular descriptors combining fingerprints yielded the best results and each threshold had its best suitable method, which means that hERG blockage assessment might depend on threshold values. Meanwhile, kNN and SVM methods were better than the others for model building. Furthermore, six privileged substructures were identified using information gain and frequency analysis methods, which could be regarded as structural alerts of cardiac toxicity mediated by hERG channel blockage.

In silico prediction of chemical toxicity on avian species using chemical category approaches.

Avian species are sensitive to pesticides and industrial chemicals, and hence used as model species in evaluation of chemical toxicity. In present study, we assessed the toxicity of more than 663 diverse chemicals on 17 avian species. All the chemicals were classified into three categories, i.e. highly toxic, slightly toxic and non-toxic, based on the toxicity classification criteria of the United States Environmental Protection Agency (EPA). To evaluate these chemicals, the toxicity prediction models were built using chemical category approaches with molecular descriptors and five commonly used fingerprints, in which five machine learning methods were performed on two standard test species: aquatic bird mallard duck and terrestrial bird northern bobwhite quail. The support vector machine (SVM) method with Pubchem fingerprint performed best as revealed by 5-fold cross-validation and the external validation set on Japanese quail. No species difference existed in our database despite several chemicals with different toxicity on some avian species. The best model had an overall accuracy at 0.851 for the prediction of toxicity on avian species, which outperformed the work of Mazzatorta et al. Furthermore, several representative substructures for characterizing avian toxicity were identified via information gain (IG) method. This study would provide a new tool for chemical safety assessment.

Prediction of human genes and diseases targeted by xenobiotics using predictive toxicogenomic-derived models (PTDMs).

New technologies for systems-level determinants of human exposure to drugs, industrial chemicals, pesticides, and other environmental agents provide an invaluable opportunity to extend the understanding of human health and potential environmental hazards. We report here the development of a new computational-systems toxicology framework, called predictive toxicogenomics-derived models (PTDMs). PTDMs integrate three networks of chemical-gene interactions (CGIs), chemical-disease associations (CDAs) and gene-disease associations (GDAs) to infer chemical hazard profiles, identify exposure data gaps and to incorporate genes and disease networks into chemical safety evaluations. Three comprehensive networks addressing CGI, CDA and GDA extracted from the comparative toxicogenomics database (CTD) were constructed. The areas under the receiver operating characteristics curve ranged from 0.85 to 0.97 and were yielded using our methodology using a 10-fold cross validation by a simulation carried out 100 times. As the illustrated examples show, we predicted new potential target genes and diseases for bisphenol A and aspirin. The molecular hypothesis and experimental evidence from published literature for these predictions were provided. The results demonstrated that our method has potential applications for chemical profiling in human health exposure and environmental hazard assessment.

In silico prediction of chemical Ames mutagenicity.

Mutagenicity is one of the most important end points of toxicity. Due to high cost and laboriousness in experimental tests, it is necessary to develop robust in silico methods to predict chemical mutagenicity. In this paper, a comprehensive database containing 7617 diverse compounds, including 4252 mutagens and 3365 nonmutagens, was constructed. On the basis of this data set, high predictive models were then built using five machine learning methods, namely support vector machine (SVM), C4.5 decision tree (C4.5 DT), artificial neural network (ANN), k-nearest neighbors (kNN), and naïve Bayes (NB), along with five fingerprints, namely CDK fingerprint (FP), Estate fingerprint (Estate), MACCS keys (MACCS), PubChem fingerprint (PubChem), and Substructure fingerprint (SubFP). Performances were measured by cross validation and an external test set containing 831 diverse chemicals. Information gain and substructure analysis were used to interpret the models. The accuracies of fivefold cross validation were from 0.808 to 0.841 for top five models. The range of accuracy for the external validation set was from 0.904 to 0.980, which outperformed that of Toxtree. Three models (PubChem-kNN, MACCS-kNN, and PubChem-SVM) showed high and reliable predictive accuracy for the mutagens and nonmutagens and, hence, could be used in prediction of chemical Ames mutagenicity.

admetSAR: a comprehensive source and free tool for assessment of chemical ADMET properties.

Absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties play key roles in the discovery/development of drugs, pesticides, food additives, consumer products, and industrial chemicals. This information is especially useful when to conduct environmental and human hazard assessment. The most critical rate limiting step in the chemical safety assessment workflow is the availability of high quality data. This paper describes an ADMET structure-activity relationship database, abbreviated as admetSAR. It is an open source, text and structure searchable, and continually updated database that collects, curates, and manages available ADMET-associated properties data from the published literature. In admetSAR, over 210,000 ADMET annotated data points for more than 96,000 unique compounds with 45 kinds of ADMET-associated properties, proteins, species, or organisms have been carefully curated from a large number of diverse literatures. The database provides a user-friendly interface to query a specific chemical profile, using either CAS registry number, common name, or structure similarity. In addition, the database includes 22 qualitative classification and 5 quantitative regression models with highly predictive accuracy, allowing to estimate ecological/mammalian ADMET properties for novel chemicals. AdmetSAR is accessible free of charge at http://www.admetexp.org.

In silico assessment of chemical biodegradability.

Biodegradation is the principal environmental dissipation process. Due to a lack of comprehensive experimental data, high study cost and time-consuming, in silico approaches for assessing the biodegradable profiles of chemicals are encouraged and is an active current research topic. Here we developed in silico methods to estimate chemical biodegradability in the environment. At first 1440 diverse compounds tested under the Japanese Ministry of International Trade and Industry (MITI) protocol were used. Four different methods, namely support vector machine, k-nearest neighbor, naïve Bayes, and C4.5 decision tree, were used to build the combinatorial classification probability models of ready versus not ready biodegradability using physicochemical descriptors and fingerprints separately. The overall predictive accuracies of the best models were more than 80% for the external test set of 164 diverse compounds. Some privileged substructures were further identified for ready or not ready biodegradable chemicals by combining information gain and substructure fragment analysis. Moreover, 27 new predicted chemicals were selected for experimental assay through the Japanese MITI test protocols, which validated that all 27 compounds were predicted correctly. The predictive accuracies of our models outperform the commonly used software of the EPI Suite. Our study provided critical tools for early assessment of biodegradability of new organic chemicals in environmental hazard assessment.

Insights into molecular basis of cytochrome p450 inhibitory promiscuity of compounds.

Cytochrome P450 inhibitory promiscuity of a drug has potential effects on the occurrence of clinical drug-drug interactions. Understanding how a molecular property is related to the P450 inhibitory promiscuity could help to avoid such adverse effects. In this study, an entropy-based index was defined to quantify the P450 inhibitory promiscuity of a compound based on a comprehensive data set, containing more than 11,500 drug-like compounds with inhibition against five major P450 isoforms, 1A2, 2C9, 2C19, 2D6, and 3A4. The results indicated that the P450 inhibitory promiscuity of a compound would have a moderate correlation with molecular aromaticity, a minor correlation with molecular lipophilicity, and no relations with molecular complexity, hydrogen bonding ability, and TopoPSA. We also applied an index to quantify the susceptibilities of different P450 isoforms to inhibition based on the same data set. The results showed that there was a surprising level of P450 inhibitory promiscuity even for substrate specific P450, susceptibility to inhibition follows the rank-order: 1A2 > 2C19 > 3A4 > 2C9 > 2D6. There was essentially no correlation between P450 inhibitory potency and specificity and minor negative trade-offs between P450 inhibitory promiscuity and catalytic promiscuity. In addition, classification models were built to predict the P450 inhibitory promiscuity of new chemicals using support vector machine algorithm with different fingerprints. The area under the receiver operating characteristic curve of the best model was about 0.9, evaluated by 5-fold cross-validation. These findings would be helpful for understanding the mechanism of P450 inhibitory promiscuity and improving the P450 inhibitory selectivity of new chemicals in drug discovery.

Classification of cytochrome P450 inhibitors and noninhibitors using combined classifiers.

Adverse side effects of drug-drug interactions induced by human cytochrome P450 (CYP) inhibition is an important consideration, especially, during the research phase of drug discovery. It is highly desirable to develop computational models that can predict the inhibitive effect of a compound against a specific CYP isoform. In this study, inhibitor predicting models were developed for five major CYP isoforms, namely 1A2, 2C9, 2C19, 2D6, and 3A4, using a combined classifier algorithm on a large data set containing more than 24,700 unique compounds, extracted from PubChem. The combined classifiers algorithm is an ensemble of different independent machine learning classifiers including support vector machine, C4.5 decision tree, k-nearest neighbor, and naïve Bayes, fused by a back-propagation artificial neural network (BP-ANN). All developed models were validated by 5-fold cross-validation and a diverse validation set composed of about 9000 diverse unique compounds. The range of the area under the receiver operating characteristic curve (AUC) for the validation sets was 0.764 to 0.815 for CYP1A2, 0.837 to 0.861 for CYP2C9, 0.793 to 0.842 for CYP2C19, 0.839 to 0.886 for CYP2D6, and 0.754 to 0.790 for CYP3A4, respectively, using the new developed combined classifiers. The overall performance of the combined classifiers fused by BP-ANN was superior to that of three classic fusion techniques (Mean, Maximum, and Multiply). The chemical spaces of data sets were explored by multidimensional scaling plots, and the use of applicability domain improved the prediction accuracies of models. In addition, some representative substructure fragments differentiating CYP inhibitors and noninhibitors were characterized by the substructure fragment analysis. These classification models are applicable for virtual screening of the five major CYP isoforms inhibitors or can be used as simple filters of potential chemicals in drug discovery.

In silico prediction of Tetrahymena pyriformis toxicity for diverse industrial chemicals with substructure pattern recognition and machine learning methods.

There is an increasing need for the rapid safety assessment of chemicals by both industries and regulatory agencies throughout the world. In silico techniques are practical alternatives in the environmental hazard assessment. It is especially true to address the persistence, bioaccumulative and toxicity potentials of organic chemicals. Tetrahymena pyriformis toxicity is often used as a toxic endpoint. In this study, 1571 diverse unique chemicals were collected from the literature and composed of the largest diverse data set for T. pyriformis toxicity. Classification predictive models of T. pyriformis toxicity were developed by substructure pattern recognition and different machine learning methods, including support vector machine (SVM), C4.5 decision tree, k-nearest neighbors and random forest. The results of a 5-fold cross-validation showed that the SVM method performed better than other algorithms. The overall predictive accuracies of the SVM classification model with radial basis functions kernel was 92.2% for the 5-fold cross-validation and 92.6% for the external validation set, respectively. Furthermore, several representative substructure patterns for characterizing T. pyriformis toxicity were also identified via the information gain analysis methods.

cis-Configuration: a new tactic/rationale for neonicotinoid molecular design.

Resistance development and limited lepidopteran activities call for the discovery of "super-neonicotinoids" solving these problems. Compounds with the cis-configuration offer an opportunity for further optimization. Fixing the nitro group in the cis-configuration provided a new approach for neonicotinoid molecular design. Introductions of the heterocycle or a bulky group are two synthesis concepts to fix the cis-configuration of the nitro group. The design, synthesis, bioactivity, and preliminary modes of action of five types of cis-neonicotinoids are reviewed. cis- and trans-neonicotinoids have some differences in bioactivities and modes of action. This study focused, especially, on the reaction diversities of nitromethylene analogues of imidacloprid with various aldehydes.