Integration of Toxicogenomics and Physiologically Based Pharmacokinetic Modeling in Human Health Risk Assessment of Perfluorooctane Sulfonate

Identification of sanguinarine as a novel antagonist for perfluorooctanoate/perfluorooctane sulfonate-induced senescence of hepatocytes: An integrated computational and experimental analysis

Perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS), two prominent per- and polyfluoroalkyl substances (PFASs), are potentially harmful to many human organs. However, there only exist limited methods to mitigate their health hazards. The aim of this study is to combine a bioinformatics analysis with in vitro experiments to discover small molecules that can alleviate liver damage caused by PFOA/PFOS. We identified 192 and 82 key genes related to hepatocytes exposed to PFOA and PFOS, respectively. The functional enrichment analysis of key genes suggested cellular senescence may be important in PFOA/PFOS-induced hepatotoxicity. The in vitro models revealed that PFOA/PFOS led to hepatocyte senescence by increasing the activity of SA-β-gal, inducing mitochondrial dysfunction, impacting cell cycle arrest, and elevating the expressions of p21, p53, IL-1β, and SASP-related cytokines. The drug-target gene set enrichment analysis method was employed to compare the transcriptome data from the Gene Expression Omnibus database (GEO), Comparative Toxicogenomics Database (CTD), and the high-throughput experiment- and reference-guided database (HERB), and 21 traditional Chinese medicines (TCMs) were identified that may alleviate PFOA/PFOS-induced liver aging. The experimental results of co-exposure to PFOA/PFOS and TCMs showed that sanguinarine has particular promise in alleviating cellular senescence caused by PFOA/PFOS. Further investigations revealed that the mTOR-p53 signaling pathway was involved in PFOA/PFOS-mediated hepatic senescence and can be blocked using sanguinarine.

Ulgen K. Advances in Physiologically Based Pharmacokinetic (PBPK) Modeling of Nanomaterials

Nanoparticles (NPs) have been widely used to improve the pharmacokinetic properties and tissue distribution of small molecules such as targeting to a specific tissue of interest, enhancing their systemic circulation, and enlarging their therapeutic properties. NPs have unique and complicated in vivo disposition properties compared to small molecule drugs due to their complex multifunctionality. Physiologically based pharmacokinetic (PBPK) modeling has been a powerful tool in the simulation of the absorption, distribution, metabolism, and elimination (ADME) characteristics of the materials, and it can be used in the characterization and prediction of the systemic disposition, toxicity, efficacy, and target exposure of various types of nanoparticles. In this review, recent advances in PBPK model applications related to the nanoparticles with unique properties, and dispositional features in the biological systems, ADME characteristics, the description of transport processes of nanoparticles in the PBPK model, and the challenges in PBPK model development of nanoparticles are delineated and juxtaposed with those encountered in small molecule models. Nanoparticle related, non-nanoparticle-related, and interspecies-scaling methods applied in PBPK modeling are reviewed. In vitro to in vivo extrapolation (IVIVE) methods being a promising computational tool to provide in vivo predictions from the results of in vitro and in silico studies are discussed. Finally, as a recent advancement ML/AI-based approaches and challenges in PBPK modeling in the estimation of ADME parameters and pharmacokinetic (PK) analysis results are introduced.

In Vitro Toxicity and Modeling Reveal Nanoplastic Effects on Marine Bivalves

Nanoplastics (NPs) represent a growing concern for global environmental health, particularly in marine ecosystems where they predominantly accumulate. The impact of NPs on marine benthic organisms, such as bivalves, raises critical questions regarding ecological integrity and food safety. Traditional methods for assessing NP toxicity are often limited by their time-intensive nature and ethical considerations. Herein, we explore the toxicological effects of NPs on the marine bivalve Ruditapes philippinarum, employing a combination of in vitro cellular assays and advanced modeling techniques. Results indicate a range of adverse effects at the organismal level, including growth inhibition (69.5-108%), oxidative stress, lipid peroxidation, and DNA damage in bivalves, following exposure to NPs at concentrations in the range of 1.6 × 109-1.6 × 1011 particles/mL (p/mL). Interestingly, the growth inhibition predicted by models (54.7-104%), based on in vitro cellular proliferation assays, shows strong agreement with the in vivo outcomes of NP exposure. Furthermore, we establish a clear correlation between cytotoxicity observed in vitro and the toxicological responses at the organismal level. Taken together, this work suggests that the integration of computational modeling with in vitro toxicity assays can predict the detrimental effects of NPs on bivalves, offering insightful references for assessing the environmental risk assessment of NPs in marine benthic ecosystems.

Predictions of tissue concentrations of myclobutanil, oxyfluorfen, and pronamide in rat and human after oral exposures via GastroPlusTM physiologically based pharmacokinetic modelling

Heritage agrochemicals like myclobutanil, oxyfluorfen, and pronamide, are extensively used in agriculture, with well-established studies on their animal toxicity. Yet, human toxicity assessment relies on conventional human risk assessment approaches including the utilization of animal-based ADME (Absorption, Distribution, Metabolism, and Excretion) data. In recent years, Physiologically Based Pharmacokinetic (PBPK) modelling approaches have played an increasing role in human risk assessment of many chemicals including agrochemicals. This study addresses the absence of PBPK-type data for myclobutanil, oxyfluorfen, and pronamide by generating in vitro data for key input PBPK parameters (Caco-2 permeability, rat plasma binding, rat blood to plasma ratio, and rat liver microsomal half-life), followed by generation of PBPK models for these three chemicals via the GastroPlusTM software. Incorporating these experimental input parameters into PBPK models, the prediction accuracy of plasma AUC (area under curve) was significantly improved. Validation against rat oral administration data demonstrated substantial enhancement. Steady-state plasma concentrations (Css) of pronamide aligned well with published data using measured PBPK parameters. Following validation, parent-based tissue concentrations for these agrochemicals were predicted in humans and rats after single or 30-day repeat exposure of 10 mg/kg/day. These predicted concentrations contribute valuable information for future human toxicity risk assessments of these agrochemicals.

A systems toxicology approach for identification of disruptions in cholesterol homeostasis after aggregated exposure to mixtures of perfluorinated compounds in humans

Per- and polyfluoroalkyl substances (PFAS) are used in various household and industrial products. In humans, positive associations were reported between PFAS, including perfluorsulfonic acid and perfluorooctanoic acid, and cholesterol, a cardiometabolic risk factor. Animal studies show the opposite. Human-centered approaches are needed to better understand the effects of PFAS mixtures on cholesterol. Here, a systems toxicology approach is described, using a gene-centered cholesterol biokinetic model. PFAS exposure-gene expression relations from published data were introduced into the model. An existing PFAS physiologically based kinetic model was augmented with lung and dermal compartments and integrated with the cholesterol model to enable exposure-effect modeling. The final model was populated with data reflecting lifetime mixture exposure from: tolerable weekly intake values; the environment; high occupational exposures (ski waxing, PFAS industry). Results indicate that low level exposures (tolerable weekly intake, environmental) did not change cholesterol. In contrast, occupational exposures clearly resulted in internal PFAS exposure and disruption of cholesterol homeostasis, largely in line with epidemiological observations. Despite model limitations (eg, dynamic range, directionality), changes in cholesterol homeostasis were predicted for ski waxers, hitherto unknown from epidemiological studies. Here, future studies involving lipid metabolism could improve risk assessment.

A novel method to derive a human safety limit for PFOA by gene expression profiling and modelling

Perfluorooctanoic acid (PFOA) is a persistent environmental contaminant that can accumulate in the human body due to its long half-life. This substance has been associated with liver, pancreatic, testicular and breast cancers, liver steatosis and endocrine disruption. PFOA is a member of a large group of substances also known as "forever chemicals" and the vast majority of substances of this group lack toxicological data that would enable their effective risk assessment in terms of human health hazards. This study aimed to derive a health-based guidance value for PFOA intake (ng/kg BW/day) from in vitro transcriptomics data. To this end, we developed an in silico workflow comprising five components: (i) sourcing in vitro hepatic transcriptomics concentration-response data; (ii) deriving molecular points of departure using BMDExpress3 and performing pathway analysis using gene set enrichment analysis (GSEA) to identify the most sensitive molecular pathways to PFOA exposure; (iii) estimating freely-dissolved PFOA concentrations in vitro using a mass balance model; (iv) estimating in vivo doses by reverse dosimetry using a PBK model for PFOA as part of a quantitative in vitro to in vivo extrapolation (QIVIVE) algorithm; and (v) calculating a tolerable daily intake (TDI) for PFOA. Fourteen percent of interrogated genes exhibited in vitro concentration-response relationships. GSEA pathway enrichment analysis revealed that "fatty acid metabolism" was the most sensitive pathway to PFOA exposure. In vitro free PFOA concentrations were calculated to be 2.9% of the nominal applied concentrations, and these free concentrations were input into the QIVIVE workflow. Exposure doses for a virtual population of 3,000 individuals were estimated, from which a TDI of 0.15 ng/kg BW/day for PFOA was calculated using the benchmark dose modelling software, PROAST. This TDI is comparable to previously published values of 1.16, 0.69, and 0.86 ng/kg BW/day by the European Food Safety Authority. In conclusion, this study demonstrates the combined utility of an "omics"-derived molecular point of departure and in silico QIVIVE workflow for setting health-based guidance values in anticipation of the acceptance of in vitro concentration-response molecular measurements in chemical risk assessment.

IVIVE-PBPK based new approach methodology for addressing early life toxicity induced by Bisphenol A

Bisphenol A (BPA) is a known endocrine disruptor mimicking natural estrogens with the potential to affect human health, especially during prenatal and postnatal exposure at or below current acceptable daily intake levels. Different adverse effects of BPA are still under investigation, and multiple mechanisms of action remain unexplored. This may be one of the reasons for the continuously changing tolerable daily intake (TDI) of BPA with the emergence of new adverse health effects over time. In addition, translational modelling through in vitro-in vivo extrapolation (IVIVE) can act as prerequisite bridge for translating in-vitro finding into human risk assessment. The objective of this study was to conduct in-vitro experiments and utilize an IVIVE-pregnancy physiologically based pharmacokinetic (P-PBPK) modeling to investigate developmental neurotoxicity and embryotoxicity in humans. The data obtained from human embryonic stem cells-based assays (study conducted between October 2020-2021) were used for the IVIVE-P-PBPK models to obtain the human equivalent doses (HEDs) which were further extrapolated to reference doses (RfDs). The results showed that simulated mean RfDs (μg/kg/day) derived from the HSD3B1 and NFATC2 gene of embryotoxicity and neurodevelopmental toxicity tests, respectively, were 4.94 and 5.18. The simulated RfDs were close to the temporary-tolerable daily intake (t-TDI) recommended by European Food Safety Authority (EFSA) in 2015 (t-TDI: 4 μg/kg·bw) and higher than the TDI of 2023 (0.2 ng/kg·bw). In conclusion, in-vitro toxicogenomics dose-response data combined with PBPK modeling can become a promising alternative new approach methodology (NAM) to support decision-making in chemical risk assessment. Based on the simulated RfDs derived from this NAM, the t-TDI set by EFSA in 2015 may be considered a safe exposure limit for mothers and fetuses at the current BPA intake levels in Chinese mothers. This study provided an animal-free new strategy for NAMs based risk assessment by combining toxicogenomics and computational toxicology.

Acetylcholinesterase Inhibition in Rats and Humans Following Acute Fenitrothion Exposure Predicted by Physiologically Based Kinetic Modeling-Facilitated Quantitative In Vitro to In Vivo Extrapolation

Worldwide use of organophosphate pesticides as agricultural chemicals aims to maintain a stable food supply, while their toxicity remains a major public health concern. A common mechanism of acute neurotoxicity following organophosphate pesticide exposure is the inhibition of acetylcholinesterase (AChE). To support Next Generation Risk Assessment for public health upon acute neurotoxicity induced by organophosphate pesticides, physiologically based kinetic (PBK) modeling-facilitated quantitative in vitro to in vivo extrapolation (QIVIVE) approach was employed in this study, with fenitrothion (FNT) as an exemplary organophosphate pesticide. Rat and human PBK models were parametrized with data derived from in silico predictions and in vitro incubations. Then, PBK model-based QIVIVE was performed to convert species-specific concentration-dependent AChE inhibition obtained from in vitro blood assays to corresponding in vivo dose-response curves, from which points of departure (PODs) were derived. The obtained values for rats and humans were comparable with reported no-observed-adverse-effect levels (NOAELs). Humans were found to be more susceptible than rats toward erythrocyte AChE inhibition induced by acute FNT exposure due to interspecies differences in toxicokinetics and toxicodynamics. The described approach adequately predicts toxicokinetics and acute toxicity of FNT, providing a proof-of-principle for applying this approach in a 3R-based chemical risk assessment paradigm.

Development and application of an interactive generic physiologically based pharmacokinetic (igPBPK) model for adult beef cattle and lactating dairy cows to estimate tissue distribution and edible tissue and milk withdrawal intervals for per- and polyfluoroalkyl substances (PFAS)

Humans can be exposed to per- and polyfluoroalkyl substances (PFAS) through dietary intake from milk and edible tissues from food animals. This study developed a physiologically based pharmacokinetic (PBPK) model to predict tissue and milk residues and estimate withdrawal intervals (WDIs) for multiple PFAS including PFOA, PFOS and PFHxS in beef cattle and lactating dairy cows. Results showed that model predictions were mostly within a two-fold factor of experimental data for plasma, tissues, and milk with an estimated coefficient of determination (R2) of >0.95. The predicted muscle WDIs for beef cattle were <1 day for PFOA, 449 days for PFOS, and 69 days for PFHxS, while the predicted milk WDIs in dairy cows were <1 day for PFOA, 1345 days for PFOS, and zero day for PFHxS following a high environmental exposure scenario (e.g., 49.3, 193, and 161 ng/kg/day for PFOA, PFOS, and PFHxS, respectively, for beef cattle for 2 years). The model was converted to a web-based interactive generic PBPK (igPBPK) platform to provide a user-friendly dashboard for predictions of tissue and milk WDIs for PFAS in cattle. This model serves as a foundation for extrapolation to other PFAS compounds to improve safety assessment of cattle-derived food products.

PPARα/ACOX1 as a novel target for hepatic lipid metabolism disorders induced by per- and polyfluoroalkyl substances: An integrated approach

BACKGROUND

Per- and polyfluoroalkyl substances (PFAS) are persistent and ubiquitous environmental contaminants with well-documented hepatotoxicity. However, the mechanistic linkage between PFAS exposure and non-alcoholic fatty liver disease (NAFLD) remains largely elusive.

OBJECTIVES

This study aimed to explore PFAS-to-NAFLD link and the relevant molecular mechanisms.

METHODS

The cross-sectional analyses using National Health and Nutrition Examination Survey (NHANES) data were conducted to investigate the association between PFAS exposure and NAFLD. A combination of in silico toxicological analyses, bioinformatics approaches, animal experiments, and in vitro assays was used to explore the molecular initiating events (MIEs) and key events (KEs) in PFAS-induced hepatic lipid metabolism disorders.

RESULTS

The cross-sectional analyses with NHANES data revealed the significant association between PFAS exposure and hepatic steatosis/NAFLD. The in silico toxicological analyses showed that PPARα activation induced by perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS), prototypical representatives of PFAS, is the critical MIE associated with NAFLD-predominant liver diseases. Transcriptome-based bioinformatic annotation and analyses identified that transcriptional upregulation of hepatic acyl-CoA oxidase 1 (ACOX1) in PPARα-regulated peroxisomal β-oxidation pathway was the KE involved with PFOA/PFOS-perturbed hepatic lipid metabolic pathways in humans, mice and rats. The in vivo and in vitro assays further verified that ACOX1-mediated oxidative stress contributed to mitochondrial compromise and lipid accumulation in PFOA/PFOS-exposed mouse hepatocytes, which could be mitigated by co-treatment with ACOX1 inhibitor and mitochondria ROS scavenger. Additionally, we observed that besides PFOA and PFOS, hepatic ACOX1 exhibited good-fit response to short-term exposures of long-chain (C7-C10) perfluoroalkyl carboxylic acids (PFHpA, PFNA, PFDA) and perfluoroalkyl sulfonic acids (PFHpS, PFDS) in human hepatocyte spheroids through benchmark dose (BMD) modeling.

CONCLUSION

Our study unveils a novel molecular target for PFAS-induced hepatic lipid metabolic disorders, shedding new light on prediction, assessment, and mitigation of PFAS hepatotoxicity.

Integrating In Vitro Data and Physiologically Based Kinetic Modeling to Predict and Compare Acute Neurotoxic Doses of Saxitoxin in Rats, Mice, and Humans

Current climate trends are likely to expand the geographic distribution of the toxigenic microalgae and concomitant phycotoxins, making intoxications by such toxins a global phenomenon. Among various phycotoxins, saxitoxin (STX) acts as a neurotoxin that might cause severe neurological symptoms in mammals following consumptions of contaminated seafood. To derive a point of departure (POD) for human health risk assessment upon acute neurotoxicity induced by oral STX exposure, a physiologically based kinetic (PBK) modeling-facilitated quantitative in vitro to in vivo extrapolation (QIVIVE) approach was employed. The PBK models for rats, mice, and humans were built using parameters from the literature, in vitro experiments, and in silico predictions. Available in vitro toxicity data for STX were converted to in vivo dose-response curves via the PBK models established for these three species, and POD values were derived from the predicted curves and compared to reported in vivo toxicity data. Interspecies differences in acute STX toxicity between rodents and humans were found, and they appeared to be mainly due to differences in toxicokinetics. The described approach resulted in adequate predictions for acute oral STX exposure, indicating that new approach methodologies, when appropriately integrated, can be used in a 3R-based chemical risk assessment paradigm.

A State-of-the-Science Review of Interactions of Per- and Polyfluoroalkyl Substances (PFAS) with Renal Transporters in Health and Disease: Implications for Population Variability in PFAS Toxicokinetics

BACKGROUND

Per- and polyfluoroalkyl substances (PFAS) are ubiquitous in the environment and have been shown to cause various adverse health impacts. In animals, sex- and species-specific differences in PFAS elimination half-lives have been linked to the activity of kidney transporters. However, PFAS molecular interactions with kidney transporters are still not fully understood. Moreover, the impact of kidney disease on PFAS elimination remains unclear.

OBJECTIVES

This state-of-the-science review integrated current knowledge to assess how changes in kidney function and transporter expression from health to disease could affect PFAS toxicokinetics and identified priority research gaps that should be addressed to advance knowledge.

METHODS

We searched for studies that measured PFAS uptake by kidney transporters, quantified transporter-level changes associated with kidney disease status, and developed PFAS pharmacokinetic models. We then used two databases to identify untested kidney transporters that have the potential for PFAS transport based on their endogenous substrates. Finally, we used an existing pharmacokinetic model for perfluorooctanoic acid (PFOA) in male rats to explore the influence of transporter expression levels, glomerular filtration rate (GFR), and serum albumin on serum half-lives.

RESULTS

The literature search identified nine human and eight rat kidney transporters that were previously investigated for their ability to transport PFAS, as well as seven human and three rat transporters that were confirmed to transport specific PFAS. We proposed a candidate list of seven untested kidney transporters with the potential for PFAS transport. Model results indicated PFOA toxicokinetics were more influenced by changes in GFR than in transporter expression.

DISCUSSION

Studies on additional transporters, particularly efflux transporters, and on more PFAS, especially current-use PFAS, are needed to better cover the role of transporters across the PFAS class. Remaining research gaps in transporter expression changes in specific kidney disease states could limit the effectiveness of risk assessment and prevent identification of vulnerable populations. https://doi.org/10.1289/EHP11885.

Pharmacokinetics and tumor delivery of nanoparticles

Nanoparticles (NPs) have been widely used in different areas, including consumer products and medicine. In terms of biomedical applications, NPs or NP-based drug formulations have been extensively investigated for cancer diagnostics and therapy in preclinical studies, but the clinical translation rate is low. Therefore, a thorough and comprehensive understanding of the pharmacokinetics of NPs, especially in drug delivery efficiency to the target therapeutic tissue tumor, is important to design more effective nanomedicines and for proper assessment of the safety and risk of NPs. This review article focuses on the pharmacokinetics of both organic and inorganic NPs and their tumor delivery efficiencies, as well as the associated mechanisms involved. We discuss the absorption, distribution, metabolism, and excretion (ADME) processes following different routes of exposure and the mechanisms involved. Many physicochemical properties and experimental factors, including particle type, size, surface charge, zeta potential, surface coating, protein binding, dose, exposure route, species, cancer type, and tumor size can affect NP pharmacokinetics and tumor delivery efficiency. NPs can be absorbed with varying degrees following different exposure routes and mainly accumulate in liver and spleen, but also distribute to other tissues such as heart, lung, kidney and tumor tissues; and subsequently get metabolized and/or excreted mainly through hepatobiliary and renal elimination. Passive and active targeting strategies are the two major mechanisms of tumor delivery, while active targeting tends to have less toxicity and higher delivery efficiency through direct interaction between ligands and receptors. We also discuss challenges and perspectives remaining in the field of pharmacokinetics and tumor delivery efficiency of NPs.

Integrative toxicogenomics: Advancing precision medicine and toxicology through artificial intelligence and OMICs technology

More information about a person's genetic makeup, drug response, multi-omics response, and genomic response is now available leading to a gradual shift towards personalized treatment. Additionally, the promotion of non-animal testing has fueled the computational toxicogenomics as a pivotal part of the next-gen risk assessment paradigm. Artificial Intelligence (AI) has the potential to provid new ways analyzing the patient data and making predictions about treatment outcomes or toxicity. As personalized medicine and toxicogenomics involve huge data processing, AI can expedite this process by providing powerful data processing, analysis, and interpretation algorithms. AI can process and integrate a multitude of data including genome data, patient records, clinical data and identify patterns to derive predictive models anticipating clinical outcomes and assessing the risk of any personalized medicine approaches. In this article, we have studied the current trends and future perspectives in personalized medicine & toxicology, the role of toxicogenomics in connecting the two fields, and the impact of AI on personalized medicine & toxicology. In this work, we also study the key challenges and limitations in personalized medicine, toxicogenomics, and AI in order to fully realize their potential.

Human health risk assessment of 6:2 Cl-PFESA through quantitative in vitro to in vivo extrapolation by integrating cell-based assays, an epigenetic key event, and physiologically based pharmacokinetic modeling

Human health risk assessment of chemicals is essential but often relies on time-consuming and animal and labor-extensive procedures. Here, we develop a population-based, quantitative in vitro to in vivo extrapolation (QIVIVE) approach which depended on cellular effects monitored by in vitro assays, considered chemical internal concentration determined by LC-MS/MS, extrapolated into in vivo target tissue concentration through physiologically based pharmacokinetic (PBPK) modelling, and assessed populational health risk using in silico modelling. By applying this QIVIVE approach to 6:2 chlorinated polyfluorinated ether sulfonate (6:2 Cl-PFESA), as a representative of the emerging pollutants, we find that 6:2 Cl-PFESA disturbed lipid homeostasis in HepG2 cells through enhancement of lipid accumulation and fatty acid β-oxidation, during which miR-93-5p served as a key event towards toxicity and thus, could serve as an efficient toxicity marker for risk assessment; further, the disruption potency of lipid homeostasis of 6:2 Cl-PFESA for the most of studied populations in China might be of moderate concern. Together, our approach improved the reliability of QIVIVE during human health risk assessment, which can readily be used for other chemicals.

Machine learning and artificial intelligence in physiologically based pharmacokinetic modeling

Physiologically based pharmacokinetic (PBPK) models are useful tools in drug development and risk assessment of environmental chemicals. PBPK model development requires the collection of species-specific physiological, and chemical-specific absorption, distribution, metabolism, and excretion (ADME) parameters, which can be a time-consuming and expensive process. This raises a need to create computational models capable of predicting input parameter values for PBPK models, especially for new compounds. In this review, we summarize an emerging paradigm for integrating PBPK modeling with machine learning (ML) or artificial intelligence (AI)-based computational methods. This paradigm includes 3 steps (1) obtain time-concentration PK data and/or ADME parameters from publicly available databases, (2) develop ML/AI-based approaches to predict ADME parameters, and (3) incorporate the ML/AI models into PBPK models to predict PK summary statistics (eg, area under the curve and maximum plasma concentration). We also discuss a neural network architecture "neural ordinary differential equation (Neural-ODE)" that is capable of providing better predictive capabilities than other ML methods when used to directly predict time-series PK profiles. In order to support applications of ML/AI methods for PBPK model development, several challenges should be addressed (1) as more data become available, it is important to expand the training set by including the structural diversity of compounds to improve the prediction accuracy of ML/AI models; (2) due to the black box nature of many ML models, lack of sufficient interpretability is a limitation; (3) Neural-ODE has great potential to be used to generate time-series PK profiles for new compounds with limited ADME information, but its application remains to be explored. Despite existing challenges, ML/AI approaches will continue to facilitate the efficient development of robust PBPK models for a large number of chemicals.

Integrating Physiologically Based Pharmacokinetic Modeling-Based Forward Dosimetry and in Vitro Bioassays to Improve the Risk Assessment of Organophosphate Esters on Human Health

The ability to accurately assess the health risks of contaminants is limited by the shortcomings of toxicological standards. Using organophosphate esters (OPEs) as an example, this study attempted to integrate physiologically based pharmacokinetic (PBPK)-based forward dosimetry and in vitro bioassays to assess the likelihood of contaminants inducing biological effects in humans. The total exposure level of OPEs for Chinese residents was 19.5 ± 8.71 ng/kg/day with inhalation being the main exposure pathway. Then, human PBPK models were developed for individual OPEs to predict their steady-state concentrations in human tissues, and the predicted median levels in blood were close to the measurements. The reference doses (RfDs) of OPEs based on in vitro bioassays were comparable to in vivo animal-derived RfDs, demonstrating the reliability of in vitro bioassays. Therefore, the likelihood of OPEs inducing bioactivities in humans (RQ_in-vitro) was calculated using in vitro toxicity data and OPE levels in human tissues. The RQ_in-vitros of tris(2-chloroisopropyl) phosphate, tris(1,3-dichloropropyl) phosphate, and triphenyl phosphate (7.68 × 10^-5-3.18 × 10^-3) were comparable to the risks assessed using traditional RfDs (5.22 × 10^-5-1.94 × 10^-3), indicating the credibility of the method proposed in this study. This study establishes a new framework to improve the health risk assessment of contaminants without sufficient toxicity data and minimize the need for animal experimentation.

Machine learning and artificial intelligence in toxicological sciences

Machine learning and artificial intelligence approaches have revolutionized multiple disciplines, including toxicology. This review summarizes representative recent applications of machine learning and artificial intelligence approaches in different areas of toxicology, including physiologically based pharmacokinetic (PBPK) modeling, quantitative structure-activity relationship modeling for toxicity prediction, adverse outcome pathway analysis, high-throughput screening, toxicogenomics, big data and toxicological databases. By leveraging machine learning and artificial intelligence approaches, now it is possible to develop PBPK models for hundreds of chemicals efficiently, to create in silico models to predict toxicity for a large number of chemicals with similar accuracies compared to in vivo animal experiments, and to analyze a large amount of different types of data (toxicogenomics, high-content image data, etc.) to generate new insights into toxicity mechanisms rapidly, which was impossible by manual approaches in the past. To continue advancing the field of toxicological sciences, several challenges should be considered: (1) not all machine learning models are equally useful for a particular type of toxicology data, and thus it is important to test different methods to determine the optimal approach; (2) current toxicity prediction is mainly on bioactivity classification (yes/no), so additional studies are needed to predict the intensity of effect or dose-response relationship; (3) as more data become available, it is crucial to perform rigorous data quality check and develop infrastructure to store, share, analyze, evaluate, and manage big data; and (4) it is important to convert machine learning models to user-friendly interfaces to facilitate their applications by both computational and bench scientists.