Data-Driven Quantitative Structure-Activity Relationship Modeling for Human Carcinogenicity by Chronic Oral Exposure

AI/ML modeling to enhance the capability of in vitro and in vivo tests in predicting human carcinogenicity

This study aimed to develop an in silico model for predicting human carcinogenicity using advanced deep learning techniques, specifically Graph Neural Networks (GNN), through a multitask learning (MTL) approach. The MTL framework leveraged auxiliary tasks, including mutagenicity, genotoxicity, animal carcinogenicity, androgen and estrogen receptor binding, to enhance the model's predictive capabilities for the primary task of human carcinogenicity. Three distinct GNN architectures were used alongside various combinations of auxiliary tasks to evaluate the variations in performance metrics. Results demonstrated that multitask learning significantly enhances the predictive performance of GNN models compared to single-task learning for predicting human carcinogenicity. The best performed MTL model achieved an area under the curve of 0.89, along with a balanced accuracy of 82 %, and sensitivity and specificity values of 0.75 and 0.89, respectively. The developed multitask learning (MTL) models function on tasks that represent assays for identifying both genotoxic and non-genotoxic carcinogens, thereby enhancing the model's capability to predict human carcinogenic risk with greater accuracy. The advanced GNN models demonstrated effectiveness in addressing data imbalance issues frequently observed in biological datasets, mitigating the bias that typically favors one class over another. Overall, these results underscore the promise of GNN-based MTL models for reliable chemical screening and prioritization, particularly in predicting human carcinogenicity.

QSAR modeling on aromatase inhibitory activity of 23 triazole fungicides by tritium-water release assay

The 1,2,4-triazole fungicides are extensively used in agriculture, and their impacts on aquatic organisms by continuous release are increasingly concerned. Aromatase, a rate-limiting enzyme for androgens converting to estrogens, is considered as a potential target for triazole fungicides. To reveal and predict the aromatase inhibition capacity of the existing and future developed triazole fungicides, 23 commonly used 1,2,4-triazole fungicides were used for the evaluation of their inhibitory effects (expressed as the 50% inhibitory concentration (IC50)) on human aromatase by 3H-H2O release assay in the present study. Result showed the IC50 values spanned four orders of magnitude from the strongest of 44 nM (flusilazole) to the lowest of 0.330 mM (bitertanol). The aromatase inhibitory activity of the triazoles was also verified in vivo by zebrafish use two triazoles with relatively weak inhibition. Subsequently, the Quantitative Structure-Activity Relationship (QSAR) modeling on the triazoles as aromatase inhibitors was constructed using stepwise regression analysis with the chemical structural descriptors including physicochemical, electronic and topological parameters. The optimal QSAR model was defined as pIC50 = -22.936-2.668 EHomo + 0.938 logD - 0.715 NHBD. The effectiveness and robustness of the model were evaluated by internal and external validation with residual assessment. The internal validation showed that the R2 and Radj2 were both higher than 0.700. The CCC and CCCExt were in acceptable levels as the cutoff value of 0.850. The cross-validation correlation coefficient Q2 and the external predictive correlation coefficients (Q2-F1, Q2-F2, and Q2-F3) were all greater than 0.600. The results of Y-Scrambling with 2000 iterations indicated the model had no accidental correlation as the average R2 of 0.166 and Q2 of -0.378. The findings offered data support for the potential risks associated with triazole fungicides in aquatic environment and provided theoretical guidance to expedite drug development and risk assessment.

Development of machine learning-based quantitative structure-activity relationship models for predicting plasma half-lives of drugs in six common food animal species

Plasma half-life is a crucial pharmacokinetic parameter for estimating extralabel withdrawal intervals of drugs to ensure the safety of food products derived from animals. This study focuses on developing a quantitative structure-activity relationship (QSAR) model incorporating multiple machine learning and artificial intelligence algorithms, and aims to predict the plasma half-lives of drugs in 6 food animals, including cattle, chickens, goats, sheep, swine, and turkeys. By integrating 4 machine learning algorithms with 5 molecular descriptor types, 20 QSAR models were developed using data from the Food Animal Residue Avoidance Databank (FARAD) Comparative Pharmacokinetic Database. The deep neural network (DNN) algorithm demonstrated the best prediction ability of plasma half-lives. The DNN model with all descriptors achieved superior performance with a high coefficient of determination (R2) of 0.82 ± 0.19 in 5-fold cross-validation on the training sets and an R2 of 0.67 on the independent test set, indicating accurate predictions and good generalizability. The final model was converted to a user-friendly web dashboard to facilitate its wide application by the scientific community. This machine learning-based QSAR model serves as a valuable tool for predicting drug plasma half-lives and extralabel withdrawal intervals in 6 common food animals based on physicochemical properties. It also provides a foundation to develop more advanced models to predict the tissue half-life of drugs in food animals.

AI and ML-based risk assessment of chemicals: predicting carcinogenic risk from chemical-induced genomic instability

Chemical risk assessment plays a pivotal role in safeguarding public health and environmental safety by evaluating the potential hazards and risks associated with chemical exposures. In recent years, the convergence of artificial intelligence (AI), machine learning (ML), and omics technologies has revolutionized the field of chemical risk assessment, offering new insights into toxicity mechanisms, predictive modeling, and risk management strategies. This perspective review explores the synergistic potential of AI/ML and omics in deciphering clastogen-induced genomic instability for carcinogenic risk prediction. We provide an overview of key findings, challenges, and opportunities in integrating AI/ML and omics technologies for chemical risk assessment, highlighting successful applications and case studies across diverse sectors. From predicting genotoxicity and mutagenicity to elucidating molecular pathways underlying carcinogenesis, integrative approaches offer a comprehensive framework for understanding chemical exposures and mitigating associated health risks. Future perspectives for advancing chemical risk assessment and cancer prevention through data integration, advanced machine learning techniques, translational research, and policy implementation are discussed. By implementing the predictive capabilities of AI/ML and omics technologies, researchers and policymakers can enhance public health protection, inform regulatory decisions, and promote sustainable development for a healthier future.

Prediction of Acrylonitrile-Butadiene-Styrene Mechanical Properties through Compressed-Sensing Techniques

One of the challenges in the plastic industry is the cost and time spent on the characterization of different grades of polymers. Compressed sensing is a data reconstruction method that combines linear algebra with optimization schemes to retrieve a signal from a limited set of measurements of that signal. Using a data set of signal examples, a tailored basis can be constructed, allowing for the optimization of the measurements that should be conducted to provide the highest and most robust signal reconstruction accuracy. In this work, compressed sensing was used to predict the values of numerous properties based on measurements for a small subset of those properties. A data set of 21 fully characterized acrylonitrile-butadiene-styrene samples was used to construct a tailored basis to determine the minimal subset of properties to measure to achieve high reconstruction accuracy for the remaining nonmeasured properties. The analysis showed that using only six measured properties, an average reconstruction error of less than 5% can be achieved. In addition, by increasing the number of measured properties to nine, an average error of less than 3% was achieved. Compressed sensing enables experts in academia and industry to substantially reduce the number of properties that must be measured to fully and accurately characterize plastics, ultimately saving both costs and time. In future work, the method should be expanded to optimize not only individual properties but also entire tests used to simultaneously measure multiple properties. Furthermore, this approach can also be applied to recycled materials, of which the properties are more difficult to predict.

Unlocking the potential of AI: Machine learning and deep learning models for predicting carcinogenicity of chemicals

The escalating apprehension surrounding the carcinogenic potential of chemicals emphasizes the imperative need for efficient methods of assessing carcinogenicity. Conventional experimental approaches such as in vitro and in vivo assays, albeit effective, suffer from being costly and time-consuming. In response to this challenge, new alternative methodologies, notably machine learning and deep learning techniques, have attracted attention for their potential in developing carcinogenicity prediction models. This article reviews the progress in predicting carcinogenicity using various machine learning and deep learning algorithms. A comparative analysis on these developed models reveals that support vector machine, random forest, and ensemble learning are commonly preferred for their robustness and effectiveness in predicting chemical carcinogenicity. Conversely, models based on deep learning algorithms, such as feedforward neural network, convolutional neural network, graph convolutional neural network, capsule neural network, and hybrid neural networks, exhibit promising capabilities but are limited by the size of available carcinogenicity datasets. This review provides a comprehensive analysis of current machine learning and deep learning models for carcinogenicity prediction, underscoring the importance of high-quality and large datasets. These observations are anticipated to catalyze future advancements in developing effective and generalizable machine learning and deep learning models for predicting chemical carcinogenicity.

Accurate Prediction of Rat Acute Oral Toxicity and Reference Dose for Thousands of Polycyclic Aromatic Hydrocarbon Derivatives Based on Chemometric QSAR and Machine Learning

Acute oral toxicity is currently not available for most polycyclic aromatic hydrocarbons (PAHs), especially their derivatives, because it is cost-prohibitive to experimentally determine all of them. Here, quantitative structure-activity relationship (QSAR) models using machine learning (ML) for predicting the toxicity of PAH derivatives were developed, based on oral toxicity data points of 788 individual substances of rats. Both the individual ML algorithm gradient boosting regression trees (GBRT) and the stacking ML algorithm (extreme gradient boosting + GBRT + random forest regression) provided the best prediction results with satisfactory determination coefficients for both cross-validation and the test set. It was found that those PAH derivatives with fewer polar hydrogens, more large-sized atoms, more branches, and lower polarizability have higher toxicity. Software based on the optimal ML-QSAR model was successfully developed to expand the application potential of the developed model, obtaining reliable prediction of pLD50 values and reference doses for 6893 external PAH derivatives. Among these chemicals, 472 were identified as moderately or highly toxic; 10 out of them had clear environment detection or use records. The findings provide valuable insights into the toxicity of PAHs and their derivatives, offering a standard platform for effectively evaluating chemical toxicity using ML-QSAR models.

Exploring the Chemical Space of the Exposome: How Far Have We Gone?

Around two-thirds of chronic human disease can not be explained by genetics alone. The Lancet Commission on Pollution and Health estimates that 16% of global premature deaths are linked to pollution. Additionally, it is now thought that humankind has surpassed the safe planetary operating space for introducing human-made chemicals into the Earth System. Direct and indirect exposure to a myriad of chemicals, known and unknown, poses a significant threat to biodiversity and human health, from vaccine efficacy to the rise of antimicrobial resistance as well as autoimmune diseases and mental health disorders. The exposome chemical space remains largely uncharted due to the sheer number of possible chemical structures, estimated at over 1060 unique forms. Conventional methods have cataloged only a fraction of the exposome, overlooking transformation products and often yielding uncertain results. In this Perspective, we have reviewed the latest efforts in mapping the exposome chemical space and its subspaces. We also provide our view on how the integration of data-driven approaches might be able to bridge the identified gaps.

Predicting Chemical Immunotoxicity through Data-Driven QSAR Modeling of Aryl Hydrocarbon Receptor Agonism and Related Toxicity Mechanisms

Computational modeling has emerged as a time-saving and cost-effective alternative to traditional animal testing for assessing chemicals for their potential hazards. However, few computational modeling studies for immunotoxicity were reported, with few models available for predicting toxicants due to the lack of training data and the complex mechanisms of immunotoxicity. In this study, we employed a data-driven quantitative structure-activity relationship (QSAR) modeling workflow to extensively enlarge the limited training data by revealing multiple targets involved in immunotoxicity. To this end, a probe data set of 6,341 chemicals was obtained from a high-throughput screening (HTS) assay testing for the activation of the aryl hydrocarbon receptor (AhR) signaling pathway, a key event leading to immunotoxicity. Searching this probe data set against PubChem yielded 3,183 assays with testing results for varying proportions of these 6,341 compounds. 100 assays were selected to develop QSAR models based on their correlations to AhR agonism. Twelve individual QSAR models were built for each assay using combinations of four machine-learning algorithms and three molecular fingerprints. 5-fold cross-validation of the resulting models showed good predictivity (average CCR = 0.73). A total of 20 assays were further selected based on QSAR model performance, and their resulting QSAR models showed good predictivity of potential immunotoxicants from external chemicals. This study provides a computational modeling strategy that can utilize large public toxicity data sets for modeling immunotoxicity and other toxicity endpoints, which have limited training data and complicated toxicity mechanisms.

Hybrid non-animal modeling: A mechanistic approach to predict chemical hepatotoxicity

Developing mechanistic non-animal testing methods based on the adverse outcome pathway (AOP) framework must incorporate molecular and cellular key events associated with target toxicity. Using data from an in vitro assay and chemical structures, we aimed to create a hybrid model to predict hepatotoxicants. We first curated a reference dataset of 869 compounds for hepatotoxicity modeling. Then, we profiled them against PubChem for existing in vitro toxicity data. Of the 2560 resulting assays, we selected the mitochondrial membrane potential (MMP) assay, a high-throughput screening (HTS) tool that can test chemical disruptors for mitochondrial function. Machine learning was applied to develop quantitative structure-activity relationship (QSAR) models with 2536 compounds tested in the MMP assay for screening new compounds. The MMP assay results, including QSAR model outputs, yielded hepatotoxicity predictions for reference set compounds with a Correct Classification Ratio (CCR) of 0.59. The predictivity improved by including 37 structural alerts (CCR = 0.8). We validated our model by testing 37 reference set compounds in human HepG2 hepatoma cells, and reliably predicting them for hepatotoxicity (CCR = 0.79). This study introduces a novel AOP modeling strategy that combines public HTS data, computational modeling, and experimental testing to predict chemical hepatotoxicity.

Developing a pragmatic consensus procedure supporting the ICH S1B(R1) weight of evidence carcinogenicity assessment

The ICH S1B carcinogenicity global testing guideline has been recently revised with a novel addendum that describes a comprehensive integrated Weight of Evidence (WoE) approach to determine the need for a 2-year rat carcinogenicity study. In the present work, experts from different organizations have joined efforts to standardize as much as possible a procedural framework for the integration of evidence associated with the different ICH S1B(R1) WoE criteria. The framework uses a pragmatic consensus procedure for carcinogenicity hazard assessment to facilitate transparent, consistent, and documented decision-making and it discusses best-practices both for the organization of studies and presentation of data in a format suitable for regulatory review. First, it is acknowledged that the six WoE factors described in the addendum form an integrated network of evidence within a holistic assessment framework that is used synergistically to analyze and explain safety signals. Second, the proposed standardized procedure builds upon different considerations related to the primary sources of evidence, mechanistic analysis, alternative methodologies and novel investigative approaches, metabolites, and reliability of the data and other acquired information. Each of the six WoE factors is described highlighting how they can contribute evidence for the overall WoE assessment. A suggested reporting format to summarize the cross-integration of evidence from the different WoE factors is also presented. This work also notes that even if a 2-year rat study is ultimately required, creating a WoE assessment is valuable in understanding the specific factors and levels of human carcinogenic risk better than have been identified previously with the 2-year rat bioassay alone.

Prediction of molecular-specific mutagenic alerts and related mechanisms of chemicals by a convolutional neural network (CNN) model based on SMILES split

Structural alerts (SAs) are essential to identify chemicals for toxicity evaluation and health risk assessment. We constructed a novel SMILES split-based deep learning model (SSDL) that was trained and verified with 5850 chemicals from the ISSSTY database and 384 external test chemicals from published papers. The training accuracy was above 0.90 and the evaluation metrics (precision, recall and F1-score) all reached 0.78 or above on both internal and external test chemicals. In this model, the molecular-specific fragment importance of chemicals was first quantified independently. Then, the SA identification method based on the importance of these fragments was statistically analyzed and verified with the ISSSTY test and external test chemicals containing one of 28 typical SAs, and most of the performances were better than that of expert rules. Furthermore, a mutagenicity mechanism prediction method was developed using 237 chemicals with four known mutagenic mechanisms based on molecular similarity calibrated by the SSDL method and fragment importance, which significantly improved accuracy in three mechanisms and had comparable accuracy in the other one compared to traditional methods. Overall, the SSDL model quantifying fragment toxicity within molecules would be a novel potentially powerful tool in the determination and visualization of molecular-specific SAs and the prediction of mutagenicity mechanisms for environmental or industrial compounds and drugs.

Stacked ensemble learning on HaCaT cytotoxicity for skin irritation prediction: A case study on dipterocarpol

Skin irritation is an adverse effect associated with various substances, including chemicals, drugs, or natural products. Dipterocarpol, extracted from Dipterocarpus alatus, contains several skin benefits notably anticancer, wound healing, and antibacterial properties. However, the skin irritation of dipterocarpol remains unassessed. Quantitative structure-activity relationship (QSAR) is a recommended tool for toxicity assessment involving less time, money, and animal testing to access unavailable acute toxicity data. Therefore, our study aimed to develop a highly accurate machine learning-based QSAR model for predicting skin irritation. We utilized a stacked ensemble learning model with 1064 chemicals. We also adhered to the recommendations from the OECD for QSAR validation. Subsequently, we used the proposed model to explore the cytotoxicity of dipterocarpol on keratinocytes. Our findings indicate that the model displayed promising statistical quality in terms of accuracy, precision, and recall in both 10-fold cross-validation and test datasets. Moreover, the model predicted that dipterocarpol does not have skin irritation, which was confirmed by the cell-based assay. In conclusion, our proposed model can be applied for the risk assessment of skin irritation in untested compounds that fall within its applicability domain. The web application of this model is available at https://qsarlabs.com/#stackhacat.