Machine learning in the identification, prediction and exploration of environmental toxicology: Challenges and perspectives

Machine learning-derived dose-response relationships considering interactions in mixtures: Applications to the oxidative potential of particulate matter

Conventional environmental health research is primarily focused on isolated chemical exposures, neglecting the complex interactions between multiple pollutants that may synergistically or antagonistically influence toxicity, thereby posing unexpected health risks. In this study, we address this knowledge gap by introducing an explainable machine learning (ML) approach with Feature Localized Intercept Transformed-Shapley Additive Explanations (FLIT-SHAP) designed to extract the dose-response relationships of specific pollutants in mixtures. In contrast to traditional SHAP, FLIT-SHAP can localize the global intercept to elucidate mixture effects, which is crucial for understanding the oxidative potential (OP) of ambient particulate matter (PM). Assessing multi-pollutant OP using FLIT-SHAP revealed both synergistic (55-63 %) and antagonistic (25-42 %) effects in laboratory-controlled OP data, but an antagonistic (33-66 %; lower OP) effect in ambient PM. Notably, the FLIT-SHAP approach demonstrated higher prediction accuracy (R2 = 0.99) compared to the additive model (R2 = 0.89) when evaluated against real-world PM samples. Quinones, such as phenanthrenequinone, play a more significant role in PM2.5 than previously recognized. Through this study, we highlighted the potential of FLIT-SHAP to enhance toxicity predictions and aid decision-making in the field of environmental health.

Application of machine learning in the study of development, behavior, nerve, and genotoxicity of zebrafish

Machine learning (ML) as a novel model-based approach has been used in studying aquatic toxicology in the environmental field. Zebrafish, as an ideal model organism in aquatic toxicology research, has been widely used to study the toxic effects of various pollutants. However, toxicity testing on organisms may cause significant harm, consume considerable time and resources, and raise ethical concerns. Therefore, ML is used in related research to reduce animal experiments and assist researchers in conducting toxicological research. Although ML techniques have matured in various fields, research on ML-based aquatic toxicology is still in its infancy due to the lack of comprehensive large-scale toxicity databases for environmental pollutants and model organisms. Therefore, to better understand the recent research progress of ML in studying the development, behavior, nerve, and genotoxicity of zebrafish, this review mainly focuses on using ML modeling to assess and predict the toxic effects of zebrafish exposure to different toxic chemicals. Meanwhile, the opportunities and challenges faced by ML in the field of toxicology were analyzed. Finally, suggestions and perspectives were proposed for the toxicity studies of ML on zebrafish in future applications.

Advancing toxicity studies of per- and poly-fluoroalkyl substances (pfass) through machine learning: Models, mechanisms, and future directions

Perfluorinated and perfluoroalkyl substances (PFASs), encompassing a vast array of isomeric chemicals, are recognized as typical emerging contaminants with direct or potential impacts on human health and the ecological environment. With the complex and elusive toxicological profiles of PFASs, machine learning (ML) has been increasingly employed in their toxicity studies due to its proficiency in prediction and data analytics. This integration is poised to become a predominant trend in environmental toxicology, propelled by the swift advancements in computational technology. This review diligently examines the literature to encapsulate the varied objectives of employing ML in the toxicity studies of PFASs: (1) Utilizing ML to establish Quantitative Structure-Activity Relationship (QSAR) models for PFASs with diverse toxicity endpoints, facilitating the targeted toxicity prediction of unidentified PFASs; (2) Investigating and substantiating the Adverse Outcome Pathway (AOP) through the synergy of ML and traditional toxicological methods, with this refining the toxicity assessment framework for PFASs; (3) Dissecting and elucidating the features of established ML models to advance Open Research into the toxicity of PFASs, with a primary focus on determinants and mechanisms. The discourse extends to an in-depth examination of ML studies, segregating findings based on their distinct application trajectories. Given that ML represents a nascent paradigm within PFASs research, this review delineates the collective challenges encountered in the ML-mediated study of PFAS toxicity and proffers strategic guidance for ensuing investigations.

Understanding the phytotoxic effects of organic contaminants on rice through predictive modeling with molecular descriptors: A data-driven analysis

The widespread introduction of organic compounds into environments poses significant risks to ecosystems. Assessing the adverse effects of organic contaminants on crops is crucial for ensuring food safety. However, laboratory research is often time-consuming and costly, and machine learning (ML) methods can offer a viable solution to address these challenges. This study aimed at developing a ML model that incorporates chemical descriptors to predict the phytotoxicity of organic contaminants on rice. A dataset was compiled by gathering published experimental data on the phytotoxicity of 60 organic compounds, with a focus on morphological inhibition, photosynthesis perturbation, and oxidative stress. Four ML models (RF, SVM, GBM, ANN) were developed using chemical molecular descriptors (CMD) and the Molecular ACCess System (MACCS) keys. RF-MACCS model demonstrated the highest fitness, achieving an R2 value of 0.79 and an RMSE of 0.14. Feature importance analysis highlighted nAtom, HBA, logKow, and TPSA as the most influential CMDs in our model. Additionally, substructures containing oxygen atoms, carbonyl group and carbon chains with nitrogen and oxygen atoms were identified as significant factors associated with phytotoxicity. This data-driven study could aid in predicting the phytotoxicity of organic contaminants on crops and evaluating the potential risks of emerging contaminants in agroecosystems.

Assessment of estrogenic potential from exudates of microcystin-producing and non-microcystin-producing Microcystis by metabolomics, machine learning and E-screen assay

Cyanobacterial blooms, often dominated by Microcystis aeruginosa, are capable of producing estrogenic effects. It is important to identify specific estrogenic compounds produced by cyanobacteria, though this can prove challenging owing to the complexity of exudate mixtures. In this study, we used untargeted metabolomics to compare components of exudates from microcystin-producing and non-microcystin-producing M. aeruginosa strains that differed with respect to their ability to produce microcystins, and across two growth phases. We identified 416 chemicals and found that the two strains produced similar components, mainly organoheterocyclic compounds (20.2%), organic acids and derivatives (17.3%), phenylpropanoids and polyketides (12.7%), benzenoids (12.0%), lipids and lipid-like molecules (11.5%), and organic oxygen compounds (10.1%). We then predicted estrogenic compounds from this group using random forest machine learning. Six compounds (daidzin, biochanin A, phenylethylamine, rhein, o-Cresol, and arbutin) belonging to phenylpropanoids and polyketides (3), benzenoids (2), and organic oxygen compound (1) were tested and exhibited estrogenic potency based upon the E-screen assay. This study confirmed that both Microcystis strains produce exudates that contain compounds with estrogenic properties, a growing concern in cyanobacteria management.

Exploring pollutant joint effects in disease through interpretable machine learning

Identifying the impact of pollutants on diseases is crucial. However, assessing the health risks posed by the interplay of multiple pollutants is challenging. This study introduces the concept of Pollutants Outcome Disease, integrating multidisciplinary knowledge and employing explainable artificial intelligence (AI) to explore the joint effects of industrial pollutants on diseases. Using lung cancer as a representative case study, an extreme gradient boosting predictive model that integrates meteorological, socio-economic, pollutants, and lung cancer statistical data is developed. The joint effects of industrial pollutants on lung cancer are identified and analyzed by employing the SHAP (Shapley Additive exPlanations) interpretable machine learning technique. Results reveal substantial spatial heterogeneity in emissions from CPG and ILC, highlighting pronounced nonlinear relationships among variables. The model yielded strong predictions (an R of 0.954, an RMSE of 4283, and an R2 of 0.911) and emphasized the impact of pollutant emission amounts on lung cancer responses. Diverse joint effects patterns were observed, varying in terms of patterns, regions (frequency), and the extent of antagonistic and synergistic effects among pollutants. The study provides a new perspective for exploring the joint effects of pollutants on diseases and demonstrates the potential of AI technology to assist scientific discovery.

Predicting thermal desorption efficiency of PAHs in contaminated sites based on an optimized machine learning approach

Thermal desorption (TD) remediation of polycyclic aromatic hydrocarbon (PAH)-contaminated sites is known for its high energy consumption and cost implications. The key to solving this issue lies in analyzing the PAHs desorption process, defining remediation endpoints, and developing prediction models to prevent excessive remediation. Establishing an accurate prediction model for remediation efficiency, which involves a systematic consideration of soil properties, TD parameters, and PAH characteristics, poses a significant challenge. This study employed a machine learning approach for predicting the remediation efficiency based on batch experiment results. The results revealed that the extreme gradient boosting (XGB) model yielded the most accurate predictions (R2 = 0.9832). The importance of features in the prediction process was quantified. A model optimization scheme was proposed, which involved integrating features based on their relevance, importance, and partial dependence. This integration not only reduced the number of input features but also enhanced prediction accuracy (R2 = 0.9867) without eliminating any features. The optimized XGB model was validated using soils from sites, demonstrating a prediction error of less than 30%. The optimized XGB model aids in identifying the most optimal conditions for thermal desorption to maximize the remediation efficiency of PAH-contaminated sites under relative cost and energy-saving conditions.

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.

Effects of Various Heavy Metal Exposures on Insulin Resistance in Non-diabetic Populations: Interpretability Analysis from Machine Learning Modeling Perspective

Increasing and compelling evidence has been proved that heavy metal exposure is involved in the development of insulin resistance (IR). We trained an interpretable predictive machine learning (ML) model for IR in the non-diabetic populations based on levels of heavy metal exposure. A total of 4354 participants from the NHANES (2003-2020) with complete information were randomly divided into a training set and a test set. Twelve ML algorithms, including random forest (RF), XGBoost (XGB), logistic regression (LR), GaussianNB (GNB), ridge regression (RR), support vector machine (SVM), multilayer perceptron (MLP), decision tree (DT), AdaBoost (AB), Gradient Boosting Decision Tree (GBDT), Voting Classifier (VC), and K-Nearest Neighbour (KNN), were constructed for IR prediction using the training set. Among these models, the RF algorithm had the best predictive performance, showing an accuracy of 80.14%, an AUC of 0.856, and an F1 score of 0.74 in the test set. We embedded three interpretable methods, the permutation feature importance analysis, partial dependence plot (PDP), and Shapley additive explanations (SHAP) in RF model for model interpretation. Urinary Ba, urinary Mo, blood Pb, and blood Cd levels were identified as the main influencers of IR. Within a specific range, urinary Ba (0.56-3.56 µg/L) and urinary Mo (1.06-20.25 µg/L) levels exhibited the most pronounced upwards trend with the risk of IR, while blood Pb (0.05-2.81 µg/dL) and blood Cd (0.24-0.65 µg/L) levels showed a declining trend with IR. The findings on the synergistic effects demonstrated that controlling urinary Ba levels might be more crucial for the management of IR. The SHAP decision plot offered personalized care for IR based on heavy metal control. In conclusion, by utilizing interpretable ML approaches, we emphasize the predictive value of heavy metals for IR, especially Ba, Mo, Pb, and Cd.

Advancing Computational Toxicology by Interpretable Machine Learning

Chemical toxicity evaluations for drugs, consumer products, and environmental chemicals have a critical impact on human health. Traditional animal models to evaluate chemical toxicity are expensive, time-consuming, and often fail to detect toxicants in humans. Computational toxicology is a promising alternative approach that utilizes machine learning (ML) and deep learning (DL) techniques to predict the toxicity potentials of chemicals. Although the applications of ML- and DL-based computational models in chemical toxicity predictions are attractive, many toxicity models are "black boxes" in nature and difficult to interpret by toxicologists, which hampers the chemical risk assessments using these models. The recent progress of interpretable ML (IML) in the computer science field meets this urgent need to unveil the underlying toxicity mechanisms and elucidate the domain knowledge of toxicity models. In this review, we focused on the applications of IML in computational toxicology, including toxicity feature data, model interpretation methods, use of knowledge base frameworks in IML development, and recent applications. The challenges and future directions of IML modeling in toxicology are also discussed. We hope this review can encourage efforts in developing interpretable models with new IML algorithms that can assist new chemical assessments by illustrating toxicity mechanisms in humans.

A benchmark dataset for machine learning in ecotoxicology

The use of machine learning for predicting ecotoxicological outcomes is promising, but underutilized. The curation of data with informative features requires both expertise in machine learning as well as a strong biological and ecotoxicological background, which we consider a barrier of entry for this kind of research. Additionally, model performances can only be compared across studies when the same dataset, cleaning, and splittings were used. Therefore, we provide ADORE, an extensive and well-described dataset on acute aquatic toxicity in three relevant taxonomic groups (fish, crustaceans, and algae). The core dataset describes ecotoxicological experiments and is expanded with phylogenetic and species-specific data on the species as well as chemical properties and molecular representations. Apart from challenging other researchers to try and achieve the best model performances across the whole dataset, we propose specific relevant challenges on subsets of the data and include datasets and splittings corresponding to each of these challenge as well as in-depth characterization and discussion of train-test splitting approaches.

Advances and applications of machine learning and deep learning in environmental ecology and health

Machine learning (ML) and deep learning (DL) possess excellent advantages in data analysis (e.g., feature extraction, clustering, classification, regression, image recognition and prediction) and risk assessment and management in environmental ecology and health (EEH). Considering the rapid growth and increasing complexity of data in EEH, it is of significance to summarize recent advances and applications of ML and DL in EEH. This review summarized the basic processes and fundamental algorithms of the ML and DL modeling, and indicated the urgent needs of ML and DL in EEH. Recent research hotspots such as environmental ecology and restoration, environmental fate of new pollutants, chemical exposures and risks, chemical hazard identification and control were highlighted. Various applications of ML and DL in EEH demonstrate their versatility and technological revolution, and present some challenges. The perspective of ML and DL in EEH were further outlined to promote the innovative analysis and cultivation of the ML-driven research paradigm.

Digital mapping of soil organic carbon density in China using an ensemble model

The soil organic carbon stock (SOCS) is considered as one of the largest carbon reservoirs in terrestrial ecosystems, and small changes in soil can cause significant changes in atmospheric CO2 concentration. Understanding organic carbon accumulation in soils is crucial if China is to meet its dual carbon target. In this study, the soil organic carbon density (SOCD) in China was digitally mapped using an ensemble machine learning (ML) model. First, based on SOCD data obtained at depths of 0-20 cm from 4356 sampling points (15 environmental covariates), we compared the performance of four ML models, namely random forest (RF), extreme gradient boosting (XGBoost), support vector machine (SVM), and artificial neural network (ANN) models, in terms of coefficient of determination (R2), mean absolute error (MAE), and root mean square error (RMSE) values. Then, we ensembled four models using Voting Regressor and the principle of stacking. The results showed that ensemble model (EM) accuracy was high (RMSE = 1.29, R2 = 0.85, MAE = 0.81), so that it could be a good choice for future research. Finally, the EM was used to predict the spatial distribution of SOCD in China, which ranged from 0.63 to 13.79 kg C/m2 (average = 4.09 (±1.90) kg C/m2). The SOC storage amount in surface soil (0-20 cm) was 39.40 Pg C. This study developed a novel, ensemble ML model for SOC prediction, and improved our understanding of the spatial distribution of SOC in China.

Effects of heavy metal exposure on hypertension: A machine learning modeling approach

Heavy metal exposure is a common risk factor for hypertension. To develop an interpretable predictive machine learning (ML) model for hypertension based on levels of heavy metal exposure, data from the NHANES (2003-2016) were employed. Random forest (RF), support vector machine (SVM), decision tree (DT), multilayer perceptron (MLP), ridge regression (RR), AdaBoost (AB), gradient boosting decision tree (GBDT), voting classifier (VC), and K-nearest neighbour (KNN) algorithms were utilized to generate an optimal predictive model for hypertension. Three interpretable methods, the permutation feature importance analysis, partial dependence plot (PDP), and Shapley additive explanations (SHAP) methods, were integrated into a pipeline and embedded in ML for model interpretation. A total of 9005 eligible individuals were randomly allocated into two distinct sets for predictive model training and validation. The results showed that among the predictive models, the RF model demonstrated the highest performance, achieving an accuracy rate of 77.40% in the validation set. The AUC and F1 score for the model were 0.84 and 0.76, respectively. Blood Pb, urinary Cd, urinary Tl, and urinary Co levels were identified as the main influencers of hypertension, and their contribution weights were 0.0504 ± 0.0482, 0.0389 ± 0.0256, 0.0307 ± 0.0179, and 0.0296 ± 0.0162, respectively. Blood Pb (0.55-2.93 μg/dL) and urinary Cd (0.06-0.15 μg/L) levels exhibited the most pronounced upwards trend with the risk of hypertension within a specific value range, while urinary Tl (0.06-0.26 μg/L) and urinary Co (0.02-0.32 μg/L) levels demonstrated a declining trend with hypertension. The findings on the synergistic effects indicated that Pb and Cd were the primary determinants of hypertension. Our findings underscore the predictive value of heavy metals for hypertension. By utilizing interpretable methods, we discerned that Pb, Cd, Tl, and Co emerged as noteworthy contributors within the predictive model.

Addressing chemical pollution in biodiversity research

Climate change, biodiversity loss, and chemical pollution are planetary-scale emergencies requiring urgent mitigation actions. As these "triple crises" are deeply interlinked, they need to be tackled in an integrative manner. However, while climate change and biodiversity are often studied together, chemical pollution as a global change factor contributing to worldwide biodiversity loss has received much less attention in biodiversity research so far. Here, we review evidence showing that the multifaceted effects of anthropogenic chemicals in the environment are posing a growing threat to biodiversity and ecosystems. Therefore, failure to account for pollution effects may significantly undermine the success of biodiversity protection efforts. We argue that progress in understanding and counteracting the negative impact of chemical pollution on biodiversity requires collective efforts of scientists from different disciplines, including but not limited to ecology, ecotoxicology, and environmental chemistry. Importantly, recent developments in these fields have now enabled comprehensive studies that could efficiently address the manifold interactions between chemicals and ecosystems. Based on their experience with intricate studies of biodiversity, ecologists are well equipped to embrace the additional challenge of chemical complexity through interdisciplinary collaborations. This offers a unique opportunity to jointly advance a seminal frontier in pollution ecology and facilitate the development of innovative solutions for environmental protection.

Computational Exploration of Bio-Degradation Patterns of Various Plastic Types

Plastic materials are recalcitrant in the open environment, surviving for longer without complete remediation. The current disposal methods of used plastic material are inefficient; consequently, plastic wastes are infiltrating the natural resources of the biosphere. The mixed composition of urban domestic waste with different plastic types makes them unfavorable for recycling; however, natural assimilation in situ is still an option to explore. In this research work, we have utilized previously published reports on the biodegradation of various plastics types and analyzed the pattern of microbial degradation. Our results demonstrate that the biodegradation of plastic material follows the chemical classification of plastic types based on their main molecular backbone. The clustering analysis of various plastic types based on their biodegradation reports has grouped them into two broad categories of C-C (non-hydrolyzable) and C-X (hydrolyzable). The C-C and C-X groups show a statistically significant difference in their biodegradation pattern at the genus level. The Bacilli class of bacteria is found to be reported more often in the C-C category, which is challenging to degrade compared to C-X. Genus enrichment analysis suggests that Pseudomonas and Bacillus from bacteria and Aspergillus and Penicillium from fungi are potential genera for the bioremediation of mixed plastic waste. The lack of uniformity in reporting the results of microbial degradation of plastic also needs to be addressed to enable productive growth in the field. Overall, the result points towards the feasibility of a microbial-based biodegradation solution for mixed plastic waste.

Modeling and insights into the structural characteristics of drug-induced autoimmune diseases

The incidence and complexity of drug-induced autoimmune diseases (DIAD) have been on the rise in recent years, which may lead to serious or fatal consequences. Besides, many environmental and industrial chemicals can also cause DIAD. However, there are few effective approaches to estimate the DIAD potential of drugs and other chemicals currently, and the structural characteristics and mechanism of action of DIAD compounds have not been clarified. In this study, we developed the in silico models for chemical DIAD prediction and investigated the structural characteristics of DIAD chemicals based on the reliable drug data on human autoimmune diseases. We collected 148 medications which were reported can cause DIAD clinically and 450 medications that clearly do not cause DIAD. Several different machine learning algorithms and molecular fingerprints were combined to develop the in silico models. The best performed model provided the good overall accuracy on validation set with 76.26%. The model was made freely available on the website http://diad.sapredictor.cn/. To further investigate the differences in structural characteristics between DIAD chemicals and non-DIAD chemicals, several key physicochemical properties were analyzed. The results showed that AlogP, molecular polar surface area (MPSA), and the number of hydrogen bond donors (nHDon) were significantly different between the DIAD and non-DIAD structures. They may be related to the DIAD toxicity of chemicals. In addition, 14 structural alerts (SA) for DIAD toxicity were detected from predefined substructures. The SAs may be helpful to explain the mechanism of action of drug induced autoimmune disease, and can used to identify the chemicals with potential DIAD toxicity. The structural alerts have been integrated in a structural alert-based web server SApredictor (http://www.sapredictor.cn). We hope the results could provide useful information for the recognition of DIAD chemicals and the insights of structural characteristics for chemical DIAD toxicity.