Deep learning for predicting toxicity of chemicals: a mini review

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.

Artificial intelligence (AI)-it's the end of the tox as we know it (and I feel fine)

The rapid progress of AI impacts diverse scientific disciplines, including toxicology, and has the potential to transform chemical safety evaluation. Toxicology has evolved from an empirical science focused on observing apical outcomes of chemical exposure, to a data-rich field ripe for AI integration. The volume, variety and velocity of toxicological data from legacy studies, literature, high-throughput assays, sensor technologies and omics approaches create opportunities but also complexities that AI can help address. In particular, machine learning is well suited to handle and integrate large, heterogeneous datasets that are both structured and unstructured-a key challenge in modern toxicology. AI methods like deep neural networks, large language models, and natural language processing have successfully predicted toxicity endpoints, analyzed high-throughput data, extracted facts from literature, and generated synthetic data. Beyond automating data capture, analysis, and prediction, AI techniques show promise for accelerating quantitative risk assessment by providing probabilistic outputs to capture uncertainties. AI also enables explanation methods to unravel mechanisms and increase trust in modeled predictions. However, issues like model interpretability, data biases, and transparency currently limit regulatory endorsement of AI. Multidisciplinary collaboration is needed to ensure development of interpretable, robust, and human-centered AI systems. Rather than just automating human tasks at scale, transformative AI can catalyze innovation in how evidence is gathered, data are generated, hypotheses are formed and tested, and tasks are performed to usher new paradigms in chemical safety assessment. Used judiciously, AI has immense potential to advance toxicology into a more predictive, mechanism-based, and evidence-integrated scientific discipline to better safeguard human and environmental wellbeing across diverse populations.

Interpretable attention-based multi-encoder transformer based QSPR model for assessing toxicity and environmental impact of chemicals

The rising demand from consumer goods and pharmaceutical industry is driving a fast expansion of newly developed chemicals. The conventional toxicity testing of unknown chemicals is expensive, time-consuming, and raises ethical concerns. The quantitative structure-property relationship (QSPR) is an efficient computational method because it saves time, resources, and animal experimentation. Advances in machine learning have improved chemical analysis in QSPR studies, but the real-world application of machine learning-based QSPR studies was limited by the unexplainable 'black box' feature of the machine learnings. In this study, multi-encoder structure-to-toxicity (S2T)-transformer based QSPR model was developed to estimate the properties of polychlorinated biphenyls (PCBs) and endocrine disrupting chemicals (EDCs). Simplified molecular input line entry systems (SMILES) and molecular descriptors calculated by the Dragon 6 software, were simultaneously considered as input of QSPR model. Furthermore, an attention-based framework is proposed to describe the relationship between the molecular structure and toxicity of hazardous chemicals. The S2T-transformer model achieved the highest R2 scores of 0.918, 0.856, and 0.907 for logarithm of octanol-water partition coefficient (Log KOW), octanol-air partition coefficient (Log KOA), and bioconcentration factor (Log BCF) estimation of PCBs, respectively. Moreover, the attention weights were able to properly interpret the lateral (meta, para) chlorination associated with PCBs toxicity and environmental impact.

Comparison of photoinduced and electrochemically induced degradation of venlafaxine

The European Union requires environmental monitoring of the antidepressant drug venlafaxine. Advanced oxidation processes provide a remedy against the spread of micropollutants. In this study, the photoinduced and electrochemical decompositions of venlafaxine were investigated in terms of mechanism and efficacy using high-performance liquid chromatography coupled to high-resolution multifragmentation mass spectrometry. Kinetic analysis, structure elucidation, matrix variation, and radical scavenging indicated the dominance of a hydroxyl-mediated indirect mechanism during photodegradation and hydroxyl and direct electrochemical oxidation for electrochemical degradation. Oxidants, sulfate, and chloride ions acted as accelerants, which reduced venlafaxine half-lives from 62 to 25 min. Humic acid decelerated degradation during ultra-violet irradiation up to 50%, but accelerated during electrochemical oxidation up to 56%. In silico quantitative structure activity relationship analysis predicted decreased environmental hazard after advanced oxidation process treatment. In general, photoirradiation proved more efficient due to faster decomposition and slightly less toxic transformation products. Yet, matrix effects would have to be carefully evaluated when potential applications as a fourth purification stage were to be considered.

Computational Nanotoxicology Models for Environmental Risk Assessment of Engineered Nanomaterials

Although engineered nanomaterials (ENMs) have tremendous potential to generate technological benefits in numerous sectors, uncertainty on the risks of ENMs for human health and the environment may impede the advancement of novel materials. Traditionally, the risks of ENMs can be evaluated by experimental methods such as environmental field monitoring and animal-based toxicity testing. However, it is time-consuming, expensive, and impractical to evaluate the risk of the increasingly large number of ENMs with the experimental methods. On the contrary, with the advancement of artificial intelligence and machine learning, in silico methods have recently received more attention in the risk assessment of ENMs. This review discusses the key progress of computational nanotoxicology models for assessing the risks of ENMs, including material flow analysis models, multimedia environmental models, physiologically based toxicokinetics models, quantitative nanostructure-activity relationships, and meta-analysis. Several challenges are identified and a perspective is provided regarding how the challenges can be addressed.

SolPredictor: Predicting Solubility with Residual Gated Graph Neural Network

Computational methods play a pivotal role in the pursuit of efficient drug discovery, enabling the rapid assessment of compound properties before costly and time-consuming laboratory experiments. With the advent of technology and large data availability, machine and deep learning methods have proven efficient in predicting molecular solubility. High-precision in silico solubility prediction has revolutionized drug development by enhancing formulation design, guiding lead optimization, and predicting pharmacokinetic parameters. These benefits result in considerable cost and time savings, resulting in a more efficient and shortened drug development process. The proposed SolPredictor is designed with the aim of developing a computational model for solubility prediction. The model is based on residual graph neural network convolution (RGNN). The RGNNs were designed to capture long-range dependencies in graph-structured data. Residual connections enable information to be utilized over various layers, allowing the model to capture and preserve essential features and patterns scattered throughout the network. The two largest datasets available to date are compiled, and the model uses a simplified molecular-input line-entry system (SMILES) representation. SolPredictor uses the ten-fold split cross-validation Pearson correlation coefficient R2 0.79±0.02 and root mean square error (RMSE) 1.03±0.04. The proposed model was evaluated using five independent datasets. Error analysis, hyperparameter optimization analysis, and model explainability were used to determine the molecular features that were most valuable for prediction.

An overview on the protective effects of ellagic acid against heavy metals, drugs, and chemicals

Ellagic acid (EA) is a polyphenol extracted from many plants. EA modulates inflammatory mediators via antioxidant mechanisms, such as catalase (CAT) activities, superoxide dismutase (SOD), enhancement, increase in glutathione (GSH), and lipid peroxidation (LPO) suppression. EA has anti-apoptotic properties that are thought to be mediated by regulating the expression of B-cell lymphoma 2 (Bcl-2), Bcl-2-associated X protein (Bax), and caspase-3. In this article, we surveyed the literature dealing with the protective effects of EA against different heavy metals, drugs, and natural toxins. The findings indicated that EA has remarkable protective properties against various toxicants. Its protective effects were mostly mediated via normalizing lipid metabolism, oxidative stress, and inflammatory mediators, for example, tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), and IL-1β. The results of this study showed that EA has significant protective effects against a varied range of compounds, either chemical or natural. These effects are mainly mediated via intensifying the antioxidant defense system. However, other mechanisms such as inhibition of inflammatory responses and suppression of apoptosis are important.

Machine learning and deep learning for brain tumor MRI image segmentation

Brain tumors are often fatal. Therefore, accurate brain tumor image segmentation is critical for the diagnosis, treatment, and monitoring of patients with these tumors. Magnetic resonance imaging (MRI) is a commonly used imaging technique for capturing brain images. Both machine learning and deep learning techniques are popular in analyzing MRI images. This article reviews some commonly used machine learning and deep learning techniques for brain tumor MRI image segmentation. The limitations and advantages of the reviewed machine learning and deep learning methods are discussed. Even though each of these methods has a well-established status in their individual domains, the combination of two or more techniques is currently an emerging trend.

Review of machine learning and deep learning models for toxicity prediction

The ever-increasing number of chemicals has raised public concerns due to their adverse effects on human health and the environment. To protect public health and the environment, it is critical to assess the toxicity of these chemicals. Traditional in vitro and in vivo toxicity assays are complicated, costly, and time-consuming and may face ethical issues. These constraints raise the need for alternative methods for assessing the toxicity of chemicals. Recently, due to the advancement of machine learning algorithms and the increase in computational power, many toxicity prediction models have been developed using various machine learning and deep learning algorithms such as support vector machine, random forest, k-nearest neighbors, ensemble learning, and deep neural network. This review summarizes the machine learning- and deep learning-based toxicity prediction models developed in recent years. Support vector machine and random forest are the most popular machine learning algorithms, and hepatotoxicity, cardiotoxicity, and carcinogenicity are the frequently modeled toxicity endpoints in predictive toxicology. It is known that datasets impact model performance. The quality of datasets used in the development of toxicity prediction models using machine learning and deep learning is vital to the performance of the developed models. The different toxicity assignments for the same chemicals among different datasets of the same type of toxicity have been observed, indicating benchmarking datasets is needed for developing reliable toxicity prediction models using machine learning and deep learning algorithms. This review provides insights into current machine learning models in predictive toxicology, which are expected to promote the development and application of toxicity prediction models in the future.

What is the ecotoxicity of a given chemical for a given aquatic species? Predicting interactions between species and chemicals using recommender system techniques

Ecotoxicological safety assessment of chemicals requires toxicity data on multiple species, despite the general desire of minimizing animal testing. Predictive models, specifically machine learning (ML) methods, are one of the tools capable of solving this apparent contradiction as they allow to generalize toxicity patterns across chemicals and species. However, despite the availability of large public toxicity datasets, the data is highly sparse, complicating model development. The aim of this study is to provide insights into how ML can predict toxicity using a large but sparse dataset. We developed models to predict LC50-values, based on experimental LC50-data covering 2431 organic chemicals and 1506 aquatic species from the ECOTOX-database. Several well-known ML techniques were evaluated and a new ML model was developed, inspired by recommender systems. This new model involves a simple linear model that learns low-rank interactions between species and chemicals using factorization machines. We evaluated the predictive performances of the developed models based on two validation settings: 1) predicting unseen chemical-species pairs, and 2) predicting unseen chemicals. The results of this study show that ML models can accurately predict LC50-values in both validation settings. Moreover, we show that the novel factorization machine approach can match well-tuned, complex, ML approaches.

Deep learning in drug discovery: a futuristic modality to materialize the large datasets for cheminformatics

Artificial intelligence (AI) development imitates the workings of the human brain to comprehend modern problems. The traditional approaches such as high throughput screening (HTS) and combinatorial chemistry are lengthy and expensive to the pharmaceutical industry as they can only handle a smaller dataset. Deep learning (DL) is a sophisticated AI method that uses a thorough comprehension of particular systems. The pharmaceutical industry is now adopting DL techniques to enhance the research and development process. Multi-oriented algorithms play a crucial role in the processing of QSAR analysis, de novo drug design, ADME evaluation, physicochemical analysis, preclinical development, followed by clinical trial data precision. In this study, we investigated the performance of several algorithms, including deep neural networks (DNN), convolutional neural networks (CNN) and multi-task learning (MTL), with the aim of generating high-quality, interpretable big and diverse databases for drug design and development. Studies have demonstrated that CNN, recurrent neural network and deep belief network are compatible, accurate and effective for the molecular description of pharmacodynamic properties. In Covid-19, existing pharmacological compounds has also been repurposed using DL models. In the absence of the Covid-19 vaccine, remdesivir and oseltamivir have been widely employed to treat severe SARS-CoV-2 infections. In conclusion, the results indicate the potential benefits of employing the DL strategies in the drug discovery process.Communicated by Ramaswamy H. Sarma.

A Review on the Recent Applications of Deep Learning in Predictive Drug Toxicological Studies

Drug toxicity prediction is an important step in ensuring patient safety during drug design studies. While traditional preclinical studies have historically relied on animal models to evaluate toxicity, recent advances in deep-learning approaches have shown great promise in advancing drug safety science and reducing animal use in preclinical studies. However, deep-learning-based approaches also face challenges in handling large biological data sets, model interpretability, and regulatory acceptance. In this review, we provide an overview of recent developments in deep-learning-based approaches for predicting drug toxicity, highlighting their potential advantages over traditional methods and the need to address their limitations. Deep-learning models have demonstrated excellent performance in predicting toxicity outcomes from various data sources such as chemical structures, genomic data, and high-throughput screening assays. The potential of deep learning for automated feature engineering is also discussed. This review emphasizes the need to address ethical concerns related to the use of deep learning in drug toxicity studies, including the reduction of animal use and ensuring regulatory acceptance. Furthermore, emerging applications of deep learning in drug toxicity prediction, such as predicting drug-drug interactions and toxicity in rare subpopulations, are highlighted. The integration of deep-learning-based approaches with traditional methods is discussed as a way to develop more reliable and efficient predictive models for drug safety assessment, paving the way for safer and more effective drug discovery and development. Overall, this review highlights the critical role of deep learning in predictive toxicology and drug safety evaluation, emphasizing the need for continued research and development in this rapidly evolving field. By addressing the limitations of traditional methods, leveraging the potential of deep learning for automated feature engineering, and addressing ethical concerns, deep-learning-based approaches have the potential to revolutionize drug toxicity prediction and improve patient safety in drug discovery and development.

Interpretable Deep-Learning Unveils Structure-Property Relationships in Polybenzenoid Hydrocarbons

In this work, interpretable deep learning was used to identify structure-property relationships governing the HOMO-LUMO gap and the relative stability of polybenzenoid hydrocarbons (PBHs) using a ring-based graph representation. This representation was combined with a subunit-based perception of PBHs, allowing chemical insights to be presented in terms of intuitive and simple structural motifs. The resulting insights agree with conventional organic chemistry knowledge and electronic structure-based analyses and also reveal new behaviors and identify influential structural motifs. In particular, we evaluated and compared the effects of linear, angular, and branching motifs on these two molecular properties and explored the role of dispersion in mitigating the torsional strain inherent in nonplanar PBHs. Hence, the observed regularities and the proposed analysis contribute to a deeper understanding of the behavior of PBHs and form the foundation for design strategies for new functional PBHs.

Mitochondrial dynamics when mitochondrial toxic chemicals exposed in 3D cultured mouse embryonic stem cell

Mitochondria need to use considerable energy for the intracellular organelles that produce ATP. They are abundant in the cells of organs, such as muscles, liver, and kidneys. The heart, which requires a lot of energy, is also rich in mitochondria. Mitochondrial damage can induce cell death. Doxorubicin, acetaminophen, valproic acid, amiodarone, and hydroxytamoxifen are representative substances that induce mitochondrial damage. On the other hand, the effects of this substance on the progress of cardiomyocyte-differentiating stem cells have not been investigated. Therefore, a 3D cultured embryonic body toxicity test was performed. The results confirmed that the cytotoxic effects on cardiomyocytes were due to mitochondrial damage in the stage of cardiomyocyte differentiation. After drug treatment, the cells were raised in the embryoid body state for four days to obtain the ID50 values, and the levels of mRNA expression associated with the mitochondrial complex were examined. The mitochondrial DNA copy numbers were also compared to prove that the substance affects the number of mitochondria in EB-state cardiomyocytes.

Supplementary Information

The online version contains supplementary material available at 10.1007/s43188-022-00161-1.

Accurate clinical toxicity prediction using multi-task deep neural nets and contrastive molecular explanations

Explainable machine learning for molecular toxicity prediction is a promising approach for efficient drug development and chemical safety. A predictive ML model of toxicity can reduce experimental cost and time while mitigating ethical concerns by significantly reducing animal and clinical testing. Herein, we use a deep learning framework for simultaneously modeling in vitro, in vivo, and clinical toxicity data. Two different molecular input representations are used; Morgan fingerprints and pre-trained SMILES embeddings. A multi-task deep learning model accurately predicts toxicity for all endpoints, including clinical, as indicated by the area under the Receiver Operator Characteristic curve and balanced accuracy. In particular, pre-trained molecular SMILES embeddings as input to the multi-task model improved clinical toxicity predictions compared to existing models in MoleculeNet benchmark. Additionally, our multitask approach is comprehensive in the sense that it is comparable to state-of-the-art approaches for specific endpoints in in vitro, in vivo and clinical platforms. Through both the multi-task model and transfer learning, we were able to indicate the minimal need of in vivo data for clinical toxicity predictions. To provide confidence and explain the model's predictions, we adapt a post-hoc contrastive explanation method that returns pertinent positive and negative features, which correspond well to known mutagenic and reactive toxicophores, such as unsubstituted bonded heteroatoms, aromatic amines, and Michael receptors. Furthermore, toxicophore recovery by pertinent feature analysis captures more of the in vitro (53%) and in vivo (56%), rather than of the clinical (8%), endpoints, and indeed uncovers a preference in known toxicophore data towards in vitro and in vivo experimental data. To our knowledge, this is the first contrastive explanation, using both present and absent substructures, for predictions of clinical and in vivo molecular toxicity.

The prediction of molecular toxicity based on BiGRU and GraphSAGE

The prediction of molecules toxicity properties plays an crucial role in the realm of the drug discovery, since it can swiftly screen out the expected drug moleculars. The conventional method for predicting toxicity is to use some in vivo or in vitro biological experiments in the laboratory, which can easily pose a threat significant time and financial waste and even ethical issues. Therefore, using computational approaches to predict molecular toxicity has become a common strategy in modern drug discovery. In this article, we propose a novel model named MTBG, which primarily makes use of both SMILES (Simplified molecular input line entry system) strings and graph structures of molecules to extract drug molecular feature in the field of drug molecular toxicity prediction. To verify the performance of the MTBG model, we opt the Tox21 dataset and several widely used baseline models. Experimental results demonstrate that our model can perform better than these baseline models.

Multitask Deep Neural Networks for Ames Mutagenicity Prediction

The Ames mutagenicity test constitutes the most frequently used assay to estimate the mutagenic potential of drug candidates. While this test employs experimental results using various strains of Salmonella typhimurium, the vast majority of the published in silico models for predicting mutagenicity do not take into account the test results of the individual experiments conducted for each strain. Instead, such QSAR models are generally trained employing overall labels (i.e., mutagenic and nonmutagenic). Recently, neural-based models combined with multitask learning strategies have yielded interesting results in different domains, given their capabilities to model multitarget functions. In this scenario, we propose a novel neural-based QSAR model to predict mutagenicity that leverages experimental results from different strains involved in the Ames test by means of a multitask learning approach. To the best of our knowledge, the modeling strategy hereby proposed has not been applied to model Ames mutagenicity previously. The results yielded by our model surpass those obtained by single-task modeling strategies, such as models that predict the overall Ames label or ensemble models built from individual strains. For reproducibility and accessibility purposes, all source code and datasets used in our experiments are publicly available.

Generation of novel Diels-Alder reactions using a generative adversarial network

Deep learning has enormous potential in the chemical and pharmaceutical fields, and generative adversarial networks (GANs) in particular have exhibited remarkable performance in the field of molecular generation as generative models. However, their application in the field of organic chemistry has been limited; thus, in this study, we attempt to utilize a GAN as a generative model for the generation of Diels-Alder reactions. A MaskGAN model was trained with 14 092 Diels-Alder reactions, and 1441 novel Diels-Alder reactions were generated. Analysis of the generated reactions indicated that the model learned several reaction rules in-depth. Thus, the MaskGAN model can be used to generate organic reactions and aid chemists in the exploration of novel reactions.

MS2Tox Machine Learning Tool for Predicting the Ecotoxicity of Unidentified Chemicals in Water by Nontarget LC-HRMS

To achieve water quality objectives of the zero pollution action plan in Europe, rapid methods are needed to identify the presence of toxic substances in complex water samples. However, only a small fraction of chemicals detected with nontarget high-resolution mass spectrometry can be identified, and fewer have ecotoxicological data available. We hypothesized that ecotoxicological data could be predicted for unknown molecular features in data-rich high-resolution mass spectrometry (HRMS) spectra, thereby circumventing time-consuming steps of molecular identification and rapidly flagging molecules of potentially high toxicity in complex samples. Here, we present MS2Tox, a machine learning method, to predict the toxicity of unidentified chemicals based on high-resolution accurate mass tandem mass spectra (MS²). The MS2Tox model for fish toxicity was trained and tested on 647 lethal concentration (LC₅₀) values from the CompTox database and validated for 219 chemicals and 420 MS² spectra from MassBank. The root mean square error (RMSE) of MS2Tox predictions was below 0.89 log-mM, while the experimental repeatability of LC₅₀ values in CompTox was 0.44 log-mM. MS2Tox allowed accurate prediction of fish LC₅₀ values for 22 chemicals detected in water samples, and empirical evidence suggested the right directionality for another 68 chemicals. Moreover, by incorporating structural information, e.g., the presence of carbonyl-benzene, amide moieties, or hydroxyl groups, MS2Tox outperforms baseline models that use only the exact mass or log K_OW.

Medical deep learning-A systematic meta-review

Deep learning has remarkably impacted several different scientific disciplines over the last few years. For example, in image processing and analysis, deep learning algorithms were able to outperform other cutting-edge methods. Additionally, deep learning has delivered state-of-the-art results in tasks like autonomous driving, outclassing previous attempts. There are even instances where deep learning outperformed humans, for example with object recognition and gaming. Deep learning is also showing vast potential in the medical domain. With the collection of large quantities of patient records and data, and a trend towards personalized treatments, there is a great need for automated and reliable processing and analysis of health information. Patient data is not only collected in clinical centers, like hospitals and private practices, but also by mobile healthcare apps or online websites. The abundance of collected patient data and the recent growth in the deep learning field has resulted in a large increase in research efforts. In Q2/2020, the search engine PubMed returned already over 11,000 results for the search term 'deep learning', and around 90% of these publications are from the last three years. However, even though PubMed represents the largest search engine in the medical field, it does not cover all medical-related publications. Hence, a complete overview of the field of 'medical deep learning' is almost impossible to obtain and acquiring a full overview of medical sub-fields is becoming increasingly more difficult. Nevertheless, several review and survey articles about medical deep learning have been published within the last few years. They focus, in general, on specific medical scenarios, like the analysis of medical images containing specific pathologies. With these surveys as a foundation, the aim of this article is to provide the first high-level, systematic meta-review of medical deep learning surveys.

Identifying congeners and transformation products of organic contaminants within complex chemical mixtures in impacted surface waters with a top-down non-targeted screening workflow

Over 350,000 compounds are registered for production and use including a high number of congeners found in complex chemical mixtures (CCMs). With such a high number of chemicals being released in the environment and degraded into transformation products (TPs), the challenge of identifying contaminants by non-targeted screening (NTS) is massive. "Bottom-up" studies, where compounds are subjected to conditions simulating environmental degradation to identify new TPs, are time consuming and cannot be relied upon to study the TPs of hundreds of thousands of compounds. Therefore, the development of "top-down" workflows, where the structural elucidation of unknown compounds is carried directly on the sample, is of interest. In this study, a top-down NTS workflow was developed using molecular networking and clustering (MNC). A total of 438 compounds were identified including 176 congeners of consumer product additives and 106 TPs. Reference standards were used to confirm the identification of 53 contaminants among them lesser-known pharmaceuticals (aliskiren, sitagliptin) and consumer product additives (lauramidopropyl betaine, 2,2,4-trimethyl-1,2-dihydroquinoline). The MNC tools allowed to group similar TPs and congeners together. As such, several previously unknown TPs of pesticides (metolachlor) and pharmaceuticals (gliclazide, irbesartan) were identified as tentative candidates or probable structures. Moreover, some congeners that had no entry on global repositories (PubChem, ChemSpider) were identified as probable structures. The workflow worked efficiently with oligomers containing ethylene oxide moieties, and with TPs structurally related to their parent compounds. The top-down approach shown in this study addresses several issues with the identification of congeners of industrial compounds from CCMs. Furthermore, it allows elucidating the structure of TPs directly from samples without relying on bottom-up studies under conditions discussed herein. The top-down workflow and the MNC tools show great potential for data mining and retrospective analysis of previous NTS studies.

Machine learning models on chemical inhibitors of mitochondrial electron transport chain

Chemicals can induce adverse effects in humans by inhibiting mitochondrial electron transport chain (ETC) such as disrupting mitochondrial membrane potential, enhancing oxidative stress and causing some diseases. Thus, identifying ETC inhibitors (ETCi) is important to chemical risk assessment and protecting the public health. However, it is not feasible to identify all ETCi with experimental methods. Quantitative structure-activity relationship (QSAR) modeling is a promising method to rapidly and effectively identify ETCi. In this study, QSAR models for predicting ETCi were developed using machine learning methods. A clustering-based under-sampling (CBUS) method was developed to handle the imbalance issue in training sets. Structure-activity landscapes were generated and analyzed for training sets generated by the CBUS method. The consensus QSAR models constructed with CBUS achieved satisfactory performances (balanced accuracy = 0.852) in 100 iterations of five-fold cross validations, indicating the models can effectively classify ETCi. The classification model was further employed to screen chemicals in the Inventory of Existing Chemical Substances of China and 13 chemicals were identified as ETCi. Fifteen structural alerts for ETCi were identified in this study. These results demonstrated that the model and structural alerts are useful to screen ETCi.

Machine Learning Models for Predicting Cytotoxicity of Nanomaterials

The wide application of nanomaterials in consumer and medical products has raised concerns about their potential adverse effects on human health. Thus, more and more biological assessments regarding the toxicity of nanomaterials have been performed. However, the different ways the evaluations were performed, such as the utilized assays, cell lines, and the differences of the produced nanoparticles, make it difficult for scientists to analyze and effectively compare toxicities of nanomaterials. Fortunately, machine learning has emerged as a powerful tool for the prediction of nanotoxicity based on the available data. Among different types of toxicity assessments, nanomaterial cytotoxicity was the focus here because of the high sensitivity of cytotoxicity assessment to different treatments without the need for complicated and time-consuming procedures. In this review, we summarized recent studies that focused on the development of machine learning models for prediction of cytotoxicity of nanomaterials. The goal was to provide insight into predicting potential nanomaterial toxicity and promoting the development of safe nanomaterials.

Prediction Models for Agonists and Antagonists of Molecular Initiation Events for Toxicity Pathways Using an Improved Deep-Learning-Based Quantitative Structure-Activity Relationship System

In silico approaches have been studied intensively to assess the toxicological risk of various chemical compounds as alternatives to traditional in vivo animal tests. Among these approaches, quantitative structure-activity relationship (QSAR) analysis has the advantages that it is able to construct models to predict the biological properties of chemicals based on structural information. Previously, we reported a deep learning (DL) algorithm-based QSAR approach called DeepSnap-DL for high-performance prediction modeling of the agonist and antagonist activity of key molecules in molecular initiating events in toxicological pathways using optimized hyperparameters. In the present study, to achieve high throughput in the DeepSnap-DL system-which consists of the preparation of three-dimensional molecular structures of chemical compounds, the generation of snapshot images from the three-dimensional chemical structures, DL, and statistical calculations-we propose an improved DeepSnap-DL approach. Using this improved system, we constructed 59 prediction models for the agonist and antagonist activity of key molecules in the Tox21 10K library. The results indicate that modeling of the agonist and antagonist activity with high prediction performance and high throughput can be achieved by optimizing suitable parameters in the improved DeepSnap-DL system.

The therapeutic effects of Prunella vulgaris against fluoride-induced oxidative damage by using the metabolomics method

Fluoride is considered as one of the most ubiquitous environmental pollutants. Numerous studies have linked reactive oxygen species (ROS)-dependent oxidative damage with fluoride intoxication, which could be prevented by antioxidants. However, the metabolomic changes induced by ROS disruptions in fluoride intoxication are yet unknown. The present study aimed to provide novel mechanistic insights into the fluoride-induced oxidative damage and to investigate the potential protective effects of ethanolic extract of Prunella vulgaris (natural antioxidant, PV) against fluoride-induced oxidative damage. The serum biochemical indicators related to fluoride-induced oxidative damage, such as lipid peroxidation parameter, inflammation and marker enzymes in the liver increased significantly in the fluoride-treated group, while antioxidant enzymes were decreased. However, PV treatment restored the level of these biochemical indicators, indicating satisfactory antioxidant, anti-inflammatory, and hepatoprotective potential of PV. The metabolomics analysis in the serum was performed by liquid chromatography-mass spectroscopy, whereas the fluoride treatment caused severe metabolic disorders in rats, which could be improved by PV. The differential metabolites screened by multivariate analysis after fluoride and PV treatment, were organic acids, fatty acids, and lipids. These differential metabolites represented disorders of glyoxylate and dicarboxylate metabolism and the citrate cycle (TCA) according to metabolic pathway analysis in fluoride treatment rats. Interestingly, the result of metabolic pathway analysis of post-treatment with PV was consistent with that of fluoride treatment, indicating that the energy metabolism plays a major role in the progress of fluoride-induced oxidative damage, as well as the therapeutic effect of PV. These findings provided a theoretical basis for understanding the mechanism underlying metabolic disorders of fluoride toxicity and the effect of PV.

Recent trends in artificial intelligence-driven identification and development of anti-neurodegenerative therapeutic agents

Neurological disorders affect various aspects of life. Finding drugs for the central nervous system is a very challenging and complex task due to the involvement of the blood-brain barrier, P-glycoprotein, and the drug's high attrition rates. The availability of big data present in online databases and resources has enabled the emergence of artificial intelligence techniques including machine learning to analyze, process the data, and predict the unknown data with high efficiency. The use of these modern techniques has revolutionized the whole drug development paradigm, with an unprecedented acceleration in the central nervous system drug discovery programs. Also, the new deep learning architectures proposed in many recent works have given a better understanding of how artificial intelligence can tackle big complex problems that arose due to central nervous system disorders. Therefore, the present review provides comprehensive and up-to-date information on machine learning/artificial intelligence-triggered effort in the brain care domain. In addition, a brief overview is presented on machine learning algorithms and their uses in structure-based drug design, ligand-based drug design, ADMET prediction, de novo drug design, and drug repurposing. Lastly, we conclude by discussing the major challenges and limitations posed and how they can be tackled in the future by using these modern machine learning/artificial intelligence approaches.

Deep Learning-Based Conformal Prediction of Toxicity

Predictive modeling for toxicity can help reduce risks in a range of applications and potentially serve as the basis for regulatory decisions. However, the utility of these predictions can be limited if the associated uncertainty is not adequately quantified. With recent studies showing great promise for deep learning-based models also for toxicity predictions, we investigate the combination of deep learning-based predictors with the conformal prediction framework to generate highly predictive models with well-defined uncertainties. We use a range of deep feedforward neural networks and graph neural networks in a conformal prediction setting and evaluate their performance on data from the Tox21 challenge. We also compare the results from the conformal predictors to those of the underlying machine learning models. The results indicate that highly predictive models can be obtained that result in very efficient conformal predictors even at high confidence levels. Taken together, our results highlight the utility of conformal predictors as a convenient way to deliver toxicity predictions with confidence, adding both statistical guarantees on the model performance as well as better predictions of the minority class compared to the underlying models.

Deep Learning Model for Identifying Critical Structural Motifs in Potential Endocrine Disruptors

This paper aims to identify structural motifs within a molecule that contribute the most toward a chemical being an endocrine disruptor. We have developed a deep neural network-based toolkit toward this aim. The trained model can virtually assess a synthetic chemical's potential to be an endocrine disruptor using machine-readable molecular representation, simplified molecular input line entry system (SMILES). Our proposed toolkit is a multilabel or multioutput classification model that combines both convolution and long short-term memory (LSTM) architectures. The toolkit leverages the advantages of an active learning-based framework that combines multiple sources of data. Class activation maps (CAMs) generated from the feature-extraction layers can identify the structural alerts and the chemical environment that determines the specificity of the structural alerts.

Machine Learning in Predictive Toxicology: Recent Applications and Future Directions for Classification Models

In recent times, machine learning has become increasingly prominent in predictive toxicology as it has shifted from in vivo studies toward in silico studies. Currently, in vitro methods together with other computational methods such as quantitative structure-activity relationship modeling and absorption, distribution, metabolism, and excretion calculations are being used. An overview of machine learning and its applications in predictive toxicology is presented here, including support vector machines (SVMs), random forest (RF) and decision trees (DTs), neural networks, regression models, naïve Bayes, k-nearest neighbors, and ensemble learning. The recent successes of these machine learning methods in predictive toxicology are summarized, and a comparison of some models used in predictive toxicology is presented. In predictive toxicology, SVMs, RF, and DTs are the dominant machine learning methods due to the characteristics of the data available. Lastly, this review describes the current challenges facing the use of machine learning in predictive toxicology and offers insights into the possible areas of improvement in the field.

Probing the Mechanism of Hepatotoxicity of Hexabromocyclododecanes through Toxicological Network Analysis

The prediction and mechanism analysis of hepatotoxicity of contaminants, because of their various phenotypes and complex mechanisms, is still a key problem in environmental research. We applied a toxicological network analysis method to predict the hepatotoxicity of three hexabromocyclododecane (HBCD) diastereoisomers (α-HBCD, β-HBCD, and γ-HBCD) and explore their potential mechanisms. First, we collected the hepatotoxicity related genes and found that those genes were significantly localized in the human interactome. Therefore, these genes form a disease module of hepatotoxicity. We also collected targets of α-, β-, and γ-HBCD and found that their targets overlap with the hepatotoxicity disease module. Then, we trained a model to predict hepatotoxicity of three HBCD diastereoisomers based on the relationship between the hepatotoxicity disease module and targets of compounds. We found that 593 genes were significantly located in the hepatotoxicity disease module (Z = 11.9, p < 0.001) involved in oxidative stress, cellular immunity, and proliferation, and the accuracy of hepatotoxicity prediction of HBCD was 0.7095 ± 0.0193 and the recall score was 0.8355 ± 0.0352. HBCD mainly affects the core disease module genes to mediate the adenosine monophosphate-activated kinase, p38MAPK, PI3K/Akt, and TNFα pathways to regulate the immune reaction and inflammation. HBCD also induces the secretion of IL6 and STAT3 to lead hepatotoxicity by regulating NR3C1. This approach is transferable to other toxicity research studies of environmental pollutants.

Development of classification models for predicting inhibition of mitochondrial fusion and fission using machine learning methods

Mitochondrial fusion and fission are processes to maintain mitochondrial function when cells respond to environment stresses. Disruption of mitochondrial fusion and fission influences cell health and can cause adverse events such as neurodegenerative disorders. It is critical to identify environmental chemicals that can disrupt mitochondrial fusion and fission. However, experimentally testing all the chemicals is not practical because experimental methods are time-consuming and costly. Quantitative structure-activity relationship (QSAR) modeling is an attractive approach for evaluation of chemicals disrupting potential on mitochondrial fusion and fission. In this study, QSAR models were developed for differentiating chemicals capable of inhibition of mitochondrial fusion and fission using machine learning algorithms (i.e. random forest, logistic regression, Bernoulli naive Bayes, and deep neural network). One hundred iterations of five-fold cross validations and external validations showed that the best model on mitochondrial fusion had area under the receiver operating characteristic curve (AUC) of 82.8% and 78.1%, respectively; and the best model for mitochondrial fission yielded AUC of 84.3% and 97.5%, respectively. Furthermore, 45 and 56 structural alerts were identified for inhibition of mitochondrial fusion and fission, respectively. The results demonstrated that the models and the structural alerts could be useful for screening chemicals that inhibit mitochondrial fusion and fission.

Discriminant models on mitochondrial toxicity improved by consensus modeling and resolving imbalance in training

Humans and animals may be exposed to tens of thousands of natural and synthetic chemicals during their lifespan. It is difficult to assess risk for all the chemicals with experimental toxicity tests. An alternative approach is to use computational toxicology methods such as quantitative structure-activity relationship (QSAR) modeling. Mitochondrial toxicity is involved in many diseases such as cancer, neurodegeneration, type 2 diabetes, cardiovascular diseases and autoimmune diseases. Thus, it is important to rapidly and efficiently identify chemicals with mitochondrial toxicity. In this study, five machine learning algorithms and twelve types of molecular fingerprints were employed to generate QSAR discriminant models for mitochondrial toxicity. A threshold moving method was adopted to resolve the imbalance issue in the training data. Consensus of the models by an averaging probability strategy improved prediction performance. The best model has correct classification rates of 81.8% and 88.3% in ten-fold cross validation and external validation, respectively. Substructures such as phenol, carboxylic acid, nitro and arylchloride were found informative through analysis of information gain and frequency of substructures. The results demonstrate that resolving imbalance in training and building consensus models can improve classification rates for mitochondrial toxicity prediction.

The power of deep learning to ligand-based novel drug discovery

INTRODUCTION

Deep discriminative and generative neural-network models are becoming an integral part of the modern approach to ligand-based novel drug discovery. The variety of different architectures of neural networks, the methods of their training, and the procedures of generating new molecules require expert knowledge to choose the most suitable approach.

AREAS COVERED

Three different approaches to deep learning use in ligand-based drug discovery are considered: virtual screening, neural generative models, and mutation-based structure generation. Several architectures of neural networks for building either discriminative or generative models are considered in this paper, including deep multilayer neural networks, different kinds of convolutional neural networks, recurrent neural networks, and several types of autoencoders. Several kinds of learning frameworks are also considered, including adversarial learning and reinforcement learning. Different types of representations for generating molecules, including SMILES, graphs, and several alternative string representations are also considered.

EXPERT OPINION

Two kinds of problem should be solved in order to make the models built using deep neural networks, especially generative models, a valuable option in ligand-based drug discovery: the issue of interpretability and explainability of deep-learning models and the issue of synthetic accessibility of novel compounds designed by deep-learning algorithms.

Binding interactions of halo-benzoic acids, halo-benzenesulfonic acids and halo-phenylboronic acids with human transthyretin

The anionic form-dependent binding interaction of halo-phenolic substances with human transthyretin (hTTR) has been observed previously. This indicates that ionizable compounds should be the primary focus in screening potential hTTR disruptors. Here, the potential binding potency of halo-benzoic acids, halo-benzenesulfonic acids/sulfates and halo-phenylboronic acids with hTTR was determined and analyzed by competitive fluorescence displacement assay integrated with computational methods. The laboratorial results indicated that the three test groups of model compounds exhibited a distinct binding affinity to hTTR. All the tested halo-phenylboronic acids, some of the tested halo-benzoic acids and halo-benzenesulfonic acids/sulfates were shown to be inactive with hTTR. Other halo-benzoic acids and halo-benzenesulfonic acids/sulfates were moderate and/or weak hTTR binders. The binding affinity of halo-benzoic acids and halo-benzenesulfonic acids/sulfates with hTTR was similar. The low distribution ability of the model compounds from water to hTTR may be the reason why they exhibited the binding potency observed with hTTR. By introducing other highly hydrophobic compounds, we observed that the binding affinity between compounds and hTTR increased with increasing molecular hydrophobicity. Those results indicated that the highly hydrophobic halo-benzoic acids and halo-benzenesulfonic acids/sulfates may be high-priority hTTR disruptors. Finally, a binary classification model was constructed employing three predictive variables. The sensitivity (S_n), specificity (S_p), predictive accuracy (Q) values of the training set and validation set were >0.83, indicating that the model had good classification performance. Thus, the binary classification model developed here could be used to distinguish whether a given ionizable compound is a potential hTTR binder or not.

Prediction of the inhibitory concentrations of chloroquine derivatives using deep neural networks models

In recent years, deep neural networks have begun to receive much attention, which has obvious advantages in feature extraction and modeling. However, in the using of deep neural networks for the QSAR modeling process, the selection of various parameters (number of neurons, hidden layers, transfer functions, data set partitioning, number of iterations, etc.) becomes difficult. Thus, we proposed a new and easy method for optimizing the model and selecting Deep Neural Networks (DNN) parameters through uniform design ideas and orthogonal design methods. By using this approach, 222 chloroquine (CQ) derivatives with half maximal inhibitory concentration values reported in different kinds of literature were selected to establish DNN models and a total number of 128,000 DNN models were built to determine the optimized parameters for selecting the better models. Comparing with linear and Artificial Neural Network (ANN) models, we found that DNN models showed better performance.Communicated by Ramaswamy H. Sarma.

Nano-(Q)SAR for Cytotoxicity Prediction of Engineered Nanomaterials

Although nanotechnology is a new and rapidly growing area of science, the impact of nanomaterials on living organisms is unknown in many aspects. In this regard, it is extremely important to perform toxicological tests, but complete characterization of all varying preparations is extremely laborious. The computational technique called quantitative structure–activity relationship, or QSAR, allows reducing the cost of time- and resource-consuming nanotoxicity tests. In this review, (Q)SAR cytotoxicity studies of the past decade are systematically considered. We regard here five classes of engineered nanomaterials (ENMs): Metal oxides, metal-containing nanoparticles, multi-walled carbon nanotubes, fullerenes, and silica nanoparticles. Some studies reveal that QSAR models are better than classification SAR models, while other reports conclude that SAR is more precise than QSAR. The quasi-QSAR method appears to be the most promising tool, as it allows accurately taking experimental conditions into account. However, experimental artifacts are a major concern in this case.

Prediction Model with High-Performance Constitutive Androstane Receptor (CAR) Using DeepSnap-Deep Learning Approach from the Tox21 10K Compound Library

The constitutive androstane receptor (CAR) plays pivotal roles in drug-induced liver injury through the transcriptional regulation of drug-metabolizing enzymes and transporters. Thus, identifying regulatory factors for CAR activation is important for understanding its mechanisms. Numerous studies conducted previously on CAR activation and its toxicity focused on in vivo or in vitro analyses, which are expensive, time consuming, and require many animals. We developed a computational model that predicts agonists for the CAR using the Toxicology in the 21st Century 10k library. Additionally, we evaluate the prediction performance of novel deep learning (DL)-based quantitative structure-activity relationship analysis called the DeepSnap-DL approach, which is a procedure of generating an omnidirectional snapshot portraying three-dimensional (3D) structures of chemical compounds. The CAR prediction model, which applies a 3D structure generator tool, called CORINA-generated and -optimized chemical structures, in the DeepSnap-DL demonstrated better performance than the existing methods using molecular descriptors. These results indicate that high performance in the prediction model using the DeepSnap-DL approach may be important to prepare suitable 3D chemical structures as input data and to enable the identification of modulators of the CAR.

Development of liposome/water partition coefficients predictive models for neutral and ionogenic organic chemicals

Membrane/water partition coefficient (K_m/w) is a vital parameter used to characterize the membrane permeability of compounds. Considering the K_m/w value is difficult to observe experimentally for real biological membranes, liposome/water partition coefficient (K_lip/w) is employed to approximate K_m/w. Here, quantitative structure property relationship (QSPR) models for logK_lip/w of the neutral organic chemicals and the neutral form of ionogenic organic chemicals (IOCs) (logK_{lip/w-neutral}), ionic form of IOCs (logK_lip/w-ionic), the speciation-corrected liposome-water distribution ratios at a pH = 7.40 (logD_{lip/w-(pH=7.40)}) were developed. In the modeling, two modeling methods (multiple linear regressions (MLR) and k-nearest neighbor (kNN)) were used. The predictive variables employed here could be calculated from the molecular structure directly. For logK_{lip/w-neutral} and logD_{lip/w-(pH=7.40)}, the logK_OW and logD_OW-based, non-logK_OW and non-logD_OW-based kNN-QSPR and MLR-QSPR models were developed, respectively. The evaluation results implied that the predictive performance of kNN-QSPR models is better than that of MLR-QSPR models. For logK_lip/w-ionic, only one acceptable MLR-QSPR model was developed for cation and anion, respectively. The model quality of the derived models was evaluated following the OECD QSPR models validation guideline. The determination coefficient (R²), leave-one-out cross validation Q² (Q²_LOO) and bootstrapping coefficient (Q²_BOOT), the external validation coefficient (Q²_EXT) of all the models met the acceptable criteria (Q² > 0.600, R² > 0.700); while the root-mean-square error (RMSE) range from 0.351 to 0.857. All the results implied that the models had good goodness-of-fit, robustness and predictive ability. Therefore, the developed models could be used to fill the data gap for substances within the applicability domain on their missing logK_{lip/w-neutral}, logK_lip/w-ionic, logD_{lip/w-(pH=7.40)} values.