Identification of Optimal Machine Learning Algorithms and Molecular Fingerprints for Explainable Toxicity Prediction Models Using ToxCast/Tox21 Bioassay Data.

Recent studies have primarily focused on introducing novel frameworks to enhance the predictive power of toxicity prediction models by refining molecular representation methods and algorithms. However, these methods are inherently complex and often pose challenges in understanding and explaining, leading to barriers in their regulatory adoption and validation. Therefore, it is necessary to select the optimal model, considering not only model performance but also interpretability. This study aimed to identify the optimal combination of molecular fingerprints (pattern-based versus algorithm-based) and machine learning algorithms (simple versus complex) for developing explainable toxicity prediction models through an comprehensive investigation of the ToxCast/Tox21 bioassay data set. For 1092 ToxCast/Tox21 assays, five molecular fingerprints (MACCS, Morgan, RDKit, Layered, and Patterned) and six algorithms (MLP, GBT, Random Forest, kNN, Logistic Regression, and Naïve Bayes) were used to train the models. Results showed that 35 models revealed acceptable performance (F1 score or accuracy is 0.8 or higher). Among the combinations, either MACCS or Morgan, paired with Random Forest, demonstrated robust performance compared with other molecular fingerprints and algorithms. MACCS and Random Forest are valuable, even when prioritizing interpretability. Consequently, the MACCS-Random Forest combination model based on four assays, targeting G protein-coupled receptor and kinase, were identified and they can be used to discern specific structural features or patterns in chemical compounds, offering explainable insights into toxicity-related chemical structures. This study indicates the importance of not disregarding the utilization of simple models when assessing both predictivity and interpretability within the context of chemical feature-based Tox21 data analysis.

Inhalation Toxicity Screening of Consumer Products Chemicals using OECD Test Guideline Data-based Machine Learning Models.

This study aimed to screen the inhalation toxicity of chemicals found in consumer products such as air fresheners, fragrances, and anti-fogging agents submitted to K-REACH using machine learning models. We manually curated inhalation toxicity data based on OECD test guideline 403 (Acute inhalation), 412 (Sub-acute inhalation), and 413 (Sub-chronic inhalation) for 1709 chemicals from the OECD eChemPortal database. Machine learning models were trained using ten algorithms, along with four molecular fingerprints (MACCS, Morgan, Topo, RDKit) and molecular descriptors, achieving F1 scores ranging from 51 % to 91 % in test dataset. Leveraging the high-performing models, we conducted a virtual screening of chemicals, initially applying them to data-rich chemicals generally used in occupational settings to determine the prediction uncertainty. Results showed high sensitivity (75 %) but low specificity (23 %), suggesting that our models can contribute to conservative screening of chemicals. Subsequently, we applied the models to consumer product chemicals, identifying 79 as of high concern. Most of the prioritized chemicals lacked GHS classifications related to inhalation toxicity, even though they were predicted to be used in many consumer products. This study highlights a potential regulatory blind spot concerning the inhalation risk of consumer product chemicals while also indicating the potential of artificial intelligence (AI) models to aid in prioritizing chemicals at the screening level.

Bridging the Gap Between Human Toxicology and Ecotoxicology Under One Health Perspective by a Cross-Species Adverse Outcome Pathway Network for Reproductive Toxicity.

Although ecotoxicological and toxicological risk assessments are performed separately from each other, recent efforts have been made in both disciplines to reduce animal testing and develop predictive approaches instead, for example, via conserved molecular markers, and in vitro and in silico approaches. Among them, adverse outcome pathways (AOPs) have been proposed to facilitate the prediction of molecular toxic effects at larger biological scales. Thus, more toxicological data are used to inform on ecotoxicological risks and vice versa. An AOP has been previously developed to predict reproductive toxicity of silver nanoparticles via oxidative stress on the nematode Caenorhabditis elegans (AOPwiki ID 207). Following this previous study, our present study aims to extend the biologically plausible taxonomic domain of applicability (tDOA) of AOP 207. Various types of data, including in vitro human cells, in vivo, and molecular to individual, from previous studies have been collected and structured into a cross-species AOP network that can inform both human toxicology and ecotoxicology risk assessments. The first step was the collection and analysis of literature data to fit the AOP criteria and build a first AOP network. Then, key event relationships were assessed using a Bayesian network modeling approach, which gave more confidence in our overall AOP network. Finally, the biologically plausible tDOA was extended using in silico approaches (Genes-to-Pathways Species Conservation Analysis and Sequence Alignment to Predict Across Species Susceptibility), which led to the extrapolation of our AOP network across over 100 taxonomic groups. Our approach shows that various types of data can be integrated into an AOP framework, and thus facilitates access to knowledge and prediction of toxic mechanisms without the need for further animal testing. Environ Toxicol Chem 2024;00:1-14. © 2024 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Flavor characteristics of garlic fish cakes using electronic nose and tongue analyses.

This study investigated the utility of garlic powder as a functional ingredient. The aim was to develop fish cakes with improved functionality and sensory preference based on the antioxidant activity and quality characteristics. Increasing amounts of garlic powder in the prepared fish cakes were associated with increasing total polyphenol and flavonoid contents, 2,2-diphenyl-1-picrylhydrazyl (DPPH) and 2,2'-azino-bis(3-ethylbenzothiazolin-6-sulfonic acid) (ABTS+) radical scavenging activity, and reducing power. Furthermore, electronic tongue and electronic nose analyses showed an increased the intensity of umami and sourness and increased the levels of volatile compounds. The lowest trimethylamine peak corresponded to the highest amount of garlic powder. Sensory evaluation indicated that 3% garlic powder had the highest score for all criteria. Fishy odor decreased as the proportion of garlic powder increased. These findings suggest that the addition of 3% garlic powder improves quality characteristics, sensory preference, and antioxidant activity of fish cakes.

Integrating aggregate exposure pathway and adverse outcome pathway for micro/nanoplastics: A review on exposure, toxicokinetics, and toxicity studies.

Micro/nanoplastics (MNPs) have emerged as a significant environmental concern due to their widespread distribution and potential adverse effects on human health and the environment. In this study, to integrate exposure and toxicity pathways of MNPs, a comprehensive review of the occurrence, toxicokinetics (absorption, distribution, and excretion [ADE]), and toxicity of MNPs were investigated using the aggregate exposure pathway (AEP) and adverse outcome pathway (AOP) frameworks. Eighty-five papers were selected: 34 papers were on detecting MNPs in environmental samples, 38 papers were on the ADE of MNPs in humans and fish, and 36 papers were related to MNPs toxicity using experimental models. This review not only summarizes individual studies but also presents a preliminary AEP-AOP framework. This framework offers a comprehensive overview of pathways, enabling a clearer visualization of intricate processes spanning from environmental media, absorption, distribution, and molecular effects to adverse outcomes. Overall, this review emphasizes the importance of integrating exposure and toxicity pathways of MNPs by utilizing AEP-AOP to comprehensively understand their impacts on human and ecological organisms. The findings contribute to highlighting the need for further research to fill the existing knowledge gaps in this field and the development of more effective strategies for the safe management of MNPs.

Quality characteristics and antioxidant activity of meringue Jeung-pyun with different amounts of cacao bean husk.

In this study, we aimed to investigate the quality characteristics, antioxidant activity, and sensory traits of meringue Jeung-pyun with different amounts of cacao bean husk powder. Based on our analyses, high cacao bean husk content resulted in an increase in certain Jeung-pyun qualities, such as the L values, b values, hardness, gumminess, and number of pores, whereas the moisture content, pH, pore size, adhesiveness, cohesiveness, and chewiness significantly decreased. Electronic tongue analysis showed that the intensity of sourness, saltiness, and umami increased with the amount of cacao bean husk added. For the sensory characteristics, C6 demonstrated the highest ranking for all test items. Furthermore, it was found that the addition of cacao bean husks increased the antioxidant activity of the Jeung-pyun (p < 0.001). Therefore, these results suggest that Jeung-pyun produced with a mixing ratio of C6 has excellent qualities, antioxidant activities, and sensory characteristics.

Differential DNA methylation and metabolite profiling of Atlantic killifish (Fundulus heteroclitus) from the New Bedford Harbor Superfund site.

Atlantic killifish (Fundulus heteroclitus) is a valuable model in evolutionary toxicology to study how the interactions between genetic and environmental factors serve the adaptive ability of organisms to resist chemical pollution. Killifish populations inhabiting environmental toxicant-contaminated New Bedford Harbor (NBH) show phenotypes tolerant to polychlorinated biphenyls (PCBs) and differences at the transcriptional and genomic levels. However, limited research has explored epigenetic alterations and metabolic effects in NBH killifish. To identify the involvement of epigenetic and metabolic regulation in the adaptive response of killifish, we investigated tissue- and sex-specific differences in global DNA methylation and metabolomic profiles of NBH killifish populations, compared to sensitive populations from a non-polluted site, Scorton Creek (SC). The results revealed that liver-specific global DNA hypomethylation and differential metabolites were evident in fish from NBH compared with those from SC. The sex-specific differences were not greater than the tissue-specific differences. We demonstrated liver-specific enriched metabolic pathways (e.g., amino acid metabolic pathways converged into the urea cycle and glutathione metabolism), suggesting possible crosstalk between differential metabolites and DNA hypomethylation in the livers of NBH killifish. Additional investigation of methylated gene regions is necessary to understand the functional role of DNA hypomethylation in the regulation of enzyme-encoding genes associated with metabolic processes and physiological changes in NBH populations.

Changes in the concept of time among North Korean refugees: A socio-cultural perspective.

Refugees from diverse socio-cultural backgrounds and geopolitical locations face temporal challenges during their transition, yet more research needs to be conducted to understand their time conception in the host society. This study explores how North Korean refugees from remote rural areas in North Korea and China adjusted their time conception in a metropolitan city in South Korea. Data were collected through a pilot study (January-March 2017) and ethnographic fieldwork (February-July 2018), entailing semi-structured interviews, informal conversations, and field notes. Vygotsky's psychological tool and Ricoeur's interpretive approach were employed to solicit the mediation of time conception through organizational activities and narrative meanings. The analysis reveals themes that challenge refugees' time conception reflecting North Korean values and daily activities-Kimilsungism, collectivism, and task-oriented attitude-and the changes in time conception on the sense of the self. Refugees' learning in time conception is a holistic spiritual, social, and personal process.

Exploring the potential of ToxCast™ data for mechanism-based prioritization of chemicals in regulatory context: Case study with priority existing chemicals (PECs) under K-REACH.

Recent studies have highlighted the potential of the ToxCast™ database for mechanism-based prioritization of chemicals. To explore the applicability of ToxCast data in the context of regulatory inventory chemicals, we screened 510 priority existing chemicals (PECs) regulated under the Act on the Registration and Evaluation, etc. of Chemical Substances (K-REACH) using ToxCast bioassays. In our analysis, a hit-call data matrix containing 298984 chemical-gene interactions was computed for 949 bioassays with the intended target genes, which enabled the identification of the putative toxicity mechanisms. Based on the reactivity to the chemicals, we analyzed 412 bioassays whose intended target gene families were cytochrome P450, oxidoreductase, transporter, nuclear receptor, steroid hormone, and DNA-binding. We also identified 141 chemicals based on their reactivity in the bioassays. These chemicals are mainly in consumer products including colorants, preservatives, air fresheners, and detergents. Our analysis revealed that in vitro bioactivities were involved in the relevant mechanisms inducing in vivo toxicity; however, this was not sufficient to predict more hazardous chemicals. Overall, the current results point to a potential and limitation in using ToxCast data for chemical prioritization in regulatory context in the absence of suitable in vivo data.

Development of a Machine Learning Model to Estimate the Biotic Ligand Model-Based Predicted No-Effect Concentrations for Copper in Freshwater.

The copper (Cu) biotic ligand model (BLM) has been used for ecological risk assessment by taking into account the bioavailability of Cu in freshwater. The Cu BLM requires data for many water chemistry variables, such as pH, major cations, and dissolved organic carbon, which can be difficult to obtain from water quality monitoring programs. To develop an optimized predicted no-effect concentration (PNEC) estimation model based on an available monitoring dataset, we proposed an initial model that considers all BLM variables, a second model that requires variables excluding alkalinity, and a third model using electrical conductivity as a surrogate for the major cations and alkalinity. Furthermore, deep neural network (DNN) models have been used to predict the nonlinear relationships between the PNEC (outcome variable) and the required input variables (explanatory variables). The predictive capacity of DNN models was compared with the results of other existing PNEC estimation tools using a look-up table and multiple linear and multivariate polynomial regression methods. Three DNN models, using different input variables, provided better predictions of the Cu PNECs compared with the existing tools for the following four test datasets: Korean, United States, Swedish, and Belgian freshwaters. Consequently, it is expected that Cu BLM-based risk assessment can be applied to various monitoring datasets, and that the most applicable model among the three different types of DNN models could be selected according to data availability for a given monitoring database. Environ Toxicol Chem 2023;42:2271-2283. © 2023 SETAC.

Alleviation of neurotoxicity induced by polystyrene nanoplastics by increased exocytosis from neurons.

Nanoplastics (NPs) are potentially toxic and pose a health risk as they can induce an inflammatory response and oxidative stress at cellular and organismal levels. Humans can be exposed to NPs through various routes, including ingestion, inhalation, and skin contact. Notably, uptake into the body via inhalation could result in brain accumulation, which may occur directly across the blood-brain barrier or via other routes. NPs that accumulate in the brain may be endocytosed into neurons, inducing neurotoxicity. Recently, we demonstrated that exposure to polystyrene (PS)-NPs reduces the viability of neurons. We have also reported that inhibiting the retrograde transport of PS-NPs by histone deacetylase 6 (HDAC6) prevents their intracellular accumulation and promotes their export in mouse embryonic fibroblasts. However, whether HDAC6 inhibition can improve neuronal viability by increasing exocytosis of PS-NPs from neurons remains unknown. In this study, mice were intranasally administered fluorescent PS-NPs (PS-YG), which accumulated in the brain and showed potential neurotoxic effects. In cultured neurons, the HDAC6 inhibitor ACY-1215 reduced the fluorescence signal detected from PS-YG, suggesting that the removal of PS-YG from neurons was promoted. Therefore, these results suggest that blocking the retrograde transport of PS-NPs using an HDAC6 inhibitor can alleviate the neurotoxic effects of PS-NPs that enter the brain.

Insights into the mechanisms of within-species variation in sensitivity to chemicals: A case study using daphnids exposed to CMIT/MIT biocide.

Living organisms adapt to their environment, and this adaptive response to environmental changes is influenced by both genomic and epigenomic components. As adaptation underpins tolerance to stressors, it is crucial to consider biological adaptation in evaluating the adverse outcomes of environmental chemicals, such as biocides. Daphnid studies have revealed differences in sensitivity to environmental chemicals between conspecific populations or clones, as well as between species. This study aimed to identify whether sensitivity to chemicals is subject to intraspecific variation, and whether this sensitivity depends on the genetic and epigenetic backgrounds of the daphnid population. We used an integrative approach to assess the comparative toxicity of a mixture of 5-chloro-2-methyl-4-isothiazoline-3-one and 2-methyl-4-isothiazolin-3-one (CMIT/MIT), a commonly used isothiazolinone biocide, by measuring mortality, reproduction, physiological traits, global DNA methylation, and proteomic expression at the species and strain levels. The results showed that the variation in sensitivity to CMIT/MIT between conspecific strains (Daphnia pulex; DPR vs. DPA strains) could exceed that observed between congeneric species (D. magna vs. D. pulex DPR strain). Under the control conditions, DPR (the strain most sensitive to CMIT/MIT) was characterized by a larger body size, a higher heart rate, and a higher level of global DNA methylation compared to its counterpart (DPA), and proteome profiles differed between the two strains. Particularly, the study identified strain-specific epigenetic and proteomic responses to LC20 of CMIT/MIT, demonstrating putative critical proteins and biological pathways associated with the observed differences in phenotype and sensitivity to CMIT/MIT. Downregulation of certain proteins (e.g., SAM synthase, GSTs, hemoglobin, and cuticle proteins) and DNA hypomethylation can be proposed as key events (KEs) of adverse outcome pathway (AOP) for isothiazolinone toxicity. Our findings indicate that both genetic variations and epigenetic modifications can lead to intraspecific variation in sensitivity to chemicals, and this variation should be considered in the ecological risk assessment framework for chemical substances. We suggest conducting further analysis on methylated gene regions and observing transgenerational effects to verify the role of crosstalk between genetic and epigenetic factors in phenotypic and protein expressions. DATA AVAILABILITY: Proteomic data is available in supplementary materials.

Trans- and Multigenerational Effects of Isothiazolinone Biocide CMIT/MIT on Genotoxicity and Epigenotoxicity in Daphnia magna.

The mixture of 5-chloro-2-methylisothiazol-3(2H)-one and 2-methylisothiazol-3(2H)-one, CMIT/MIT, is an isothiazolinone biocide that is consistently detected in aquatic environments because of its broad-spectrum usage in industrial fields. Despite concerns about ecotoxicological risks and possible multigenerational exposure, toxicological information on CMIT/MIT is very limited to human health and within-generational toxicity. Furthermore, epigenetic markers altered by chemical exposure can be transmitted over generations, but the role of these changes in phenotypic responses and toxicity with respect to trans- and multigenerational effects is poorly understood. In this study, the toxicity of CMIT/MIT on Daphnia magna was evaluated by measuring various endpoints (mortality, reproduction, body size, swimming behavior, and proteomic expression), and its trans- and multigenerational effects were investigated over four consecutive generations. The genotoxicity and epigenotoxicity of CMIT/MIT were examined using a comet assay and global DNA methylation measurements. The results show deleterious effects on various endpoints and differences in response patterns according to different exposure histories. Parental effects were transgenerational or recovered after exposure termination, while multigenerational exposure led to acclimatory/defensive responses. Changes in DNA damage were closely associated with altered reproduction in daphnids, but their possible relationship with global DNA methylation was not found. Overall, this study provides ecotoxicological information on CMIT/MIT relative to multifaceted endpoints and aids in understanding multigenerational phenomena under CMIT/MIT exposure. It also emphasizes the consideration of exposure duration and multigenerational observations in evaluating ecotoxicity and the risk management of isothiazolinone biocides.

Identification of molecular initiating events (MIE) using chemical database analysis and nuclear receptor activity assays for screening potential inhalation toxicants.

An adverse outcome pathway (AOP) framework can facilitate the use of alternative assays in chemical regulations by providing scientific evidence. Previously, an AOP, peroxisome proliferative-activating receptor gamma (PPARγ) antagonism that leads to pulmonary fibrosis, was developed. Based on a literature search, PPARγ inactivation has been proposed as a molecular initiating event (MIE). In addition, a list of candidate chemicals that could be used in the experimental validation was proposed using toxicity database and deep learning models. In this study, the screening of environmental chemicals for MIE was conducted using in silico and in vitro tests to maximize the applicability of this AOP for screening inhalation toxicants. Initially, potential inhalation exposure chemicals that are active in three or more key events were selected, and in silico molecular docking was performed. Among the chemicals with low binding energy to PPARγ, nine chemicals were selected for validation of the AOP using in vitro PPARγ activity assay. As a result, rotenone, triorthocresyl phosphate, and castor oil were proposed as PPARγ antagonists and stressor chemicals of the AOP. Overall, the proposed tiered approach of the database-in silico-in vitro can help identify the regulatory applicability and assist in the development and experimental validation of AOP.

Integrative Data Mining Approach: Case Study with Adverse Outcome Pathway Network Leading to Pulmonary Fibrosis.

An adverse outcome pathway (AOP) framework can be applied as an efficient tool for the rapid screening of environmental chemicals. For the development of an AOP, a database mining approach can support an expert derivation approach by gathering a wider range of evidence than in a literature review. In this study, data from various databases were integrated and analyzed to supplement the AOP leading to pulmonary fibrosis by analyzing additional evidence using a data mining approach and establishing an application domain for chemicals. First, we collected chemicals, genes, and phenotypes that were studied and related to pulmonary fibrosis through the Comparative Toxicogenomics Database (CTD). CGPD-tetramers constructed by linking each related chemical, gene, phenotype, and disease can provide the basic components for the assembly of putative AOPs. Next, an AOP network was established by connecting eight existing AOPs for pulmonary fibrosis developed by expert derivation from the AOP Wiki. Finally, the pulmonary fibrosis AOP network was proposed by integrating the AOP network from AOP Wiki and the CGPD-tetramers from the CTD. To prioritize potential chemical stressors in the AOP network, 61 chemicals were ranked using the relevance of the chemical to the AOP and chemical exposure information from the CompTox Chemicals Dashboard. The approach proposed in this study can guide the utilization of available evidence from various databases as well as the literature in constructing AOP networks related to specific diseases.

Comparison of the Estimation Methods from Acute to Chronic Biotic Ligand Model-Based Predicted No-Effect Concentrations for Nickel in Freshwater Species.

Biotic ligand models (BLMs) and the sensitivities of indigenous species are used to assess the environmental risk considering the bioavailability of metals, such as nickel. However, the BLM-based acute-to-chronic ratio (ACR) is required if the predicted no-effect concentration (PNEC) cannot be derived from the chronic species sensitivity distribution (SSD). The applicability of the ACR approach for estimating BLM-based PNEC for nickel from acute toxicity data was evaluated in the present study. The BLM-based acute SSD for nickel was built using the sensitivities of 21 indigenous species and different taxon-specific BLMs for each taxonomic group. To predict the acute sensitivity of invertebrates, the chronic crustacean nickel BLM with pH effect term, which can account for nickel toxicity at high pH levels, was used. This was used instead of the existing acute BLM for crustacean, which has too narrow a pH range to cover the pH dependency of toxicity. The final BLM-based ACR of nickel, determined within a factor of 1.53 from the species-specific acute and chronic sensitivities of the six species, was more reliable than the typical ACR estimated within a factor of 1.84. A linear relationship (r2 = 0.95) was observed between the PNECs using BLM-based ACR and the PNECs derived from the BLM-based chronic SSD of the European Union Risk Assessment Reports. In conclusion, the BLM-based PNEC for nickel could be derived using the ACR approach, unlike when copper BLM was applied. The BLM-based ACR for nickel is the first result calculated by directly comparing acute and chronic species sensitivities, and will contribute to the application of BLM-based risk assessment in broader ecoregions. Environ Toxicol Chem 2023;42:914-927. © 2023 SETAC.

Mixture and individual effects of benzene, toluene, and formaldehyde in zebrafish (Danio rerio) development: Metabolomics, epigenetics, and behavioral approaches.

In this study, we aimed to investigate the potential hazards of volatile organic compounds (VOCs) on the development of zebrafish. To this end, zebrafish embryos were exposed in two different windows, either alone or in a mixture with VOCs (benzene, toluene, and formaldehyde) [EW1: 4 ± 2 h post-fertilization (hpf) to 24 hpf and EW2: 24 ± 2 hpf to 48 hpf]. Alterations in global DNA methylation and related gene expression, behavioral responses, and stress-related gene expression were observed. In addition to these endpoints, non-targeted NMR-based global metabolomics followed by pathway analysis showed significant changes in the metabolism of various amino acids during VOC exposure. Regardless of the analyzed endpoints, toluene was the most toxic chemical when exposed individually and possibly played the most pivotal role in the mixture treatment conditions. In conclusion, our data show that exposure to VOCs at embryonic developmental stages causes physiological perturbations and adverse outcomes at later life stages.

Effects of Class Imbalance and Data Scarcity on the Performance of Binary Classification Machine Learning Models Developed Based on ToxCast/Tox21 Assay Data.

The development of toxicity classification models using the ToxCast database has been extensively studied. Machine learning approaches are effective in identifying the bioactivity of untested chemicals. However, ToxCast assays differ in the amount of data and degree of class imbalance (CI). Therefore, the resampling algorithm employed should vary depending on the data distribution to achieve optimal classification performance. In this study, the effects of CI and data scarcity (DS) on the performance of binary classification models were investigated using ToxCast bioassay data. An assay matrix based on CI and DS was prepared for 335 assays with biologically intended target information, and 28 CI assays and 3 DS assays were selected. Thirty models established by combining five molecular fingerprints (i.e., Morgan, MACCS, RDKit, Pattern, and Layered) and six algorithms [i.e., gradient boosting tree, random forest (RF), multi-layered perceptron, k-nearest neighbor, logistic regression, and naive Bayes] were trained using the selected assay data set. Of the 30 trained models, MACCS-RF showed the best performance and thus was selected for analyses of the effects of CI and DS. Results showed that recall and F1 were significantly lower when training with the CI assays than with the DS assays. In addition, hyperparameter tuning of the RF algorithm significantly improved F1 on CI assays. This study provided a basis for developing a toxicity classification model with improved performance by evaluating the effects of data set characteristics. This study also emphasized the importance of using appropriate evaluation metrics and tuning hyperparameters in model development.

Quantitative adverse outcome pathway (qAOP) using bayesian network model on comparative toxicity of multi-walled carbon nanotubes (MWCNTs): safe-by-design approach.

While the various physicochemical properties of engineered nanomaterials influence their toxicities, their understanding is still incomplete. A predictive framework is required to develop safe nanomaterials, and a Bayesian network (BN) model based on adverse outcome pathway (AOP) can be utilized for this purpose. In this study, to explore the applicability of the AOP-based BN model in the development of safe nanomaterials, a comparative study was conducted on the change in the probability of toxicity pathways in response to changes in the dimensions and surface functionalization of multi-walled carbon nanotubes (MWCNTs). Based on the results of our previous study, we developed an AOP leading to cell death, and the experimental results were collected in human liver cells (HepG2) and bronchial epithelium cells (Beas-2B). The BN model was trained on these data to identify probabilistic causal relationships between key events. The results indicated that dimensions were the main influencing factor for lung cells, whereas -OH or -COOH surface functionalization and aspect ratio were the main influencing factors for liver cells. Endoplasmic reticulum stress was found to be a more sensitive pathway for dimensional changes, and oxidative stress was a more sensitive pathway for surface functionalization. Overall, our results suggest that the AOP-based BN model can be used to provide a scientific basis for the development of safe nanomaterials.