Natural food supplements, including bee pollen supplements, are increasingly popular, yet they may hide safety issues arising from concentrated chemicals, ranging from natural toxins to anthropogenic contaminants. To assess their safety, methods are needed to chemically characterize both their exposome and metabolome. Hence, we developed and validated a nontargeted screening method based on liquid chromatography high-resolution mass spectrometry (NTS LC-HRMS), combined with computational strategies to comprehensively annotate and simultaneously (semi)quantify contaminants and bioactive compounds in bee pollen. The method was validated for 15 target compounds spiked at 0.3-3 mg/kg in the bee pollen, including pesticides, pyrrolizidine alkaloids, and mycotoxins (log P range of -1.4 to 5). These compounds were accurately annotated using the developed nontargeted workflow (100% accuracy when MS2 data were acquired) and semiquantified with a mean prediction error of 2.42. Concurrently, the method successfully annotated more than 4000 compounds in bee pollen by integrating suspect screening and nontargeted screening approaches. These results demonstrate the robustness and added value of integrated targeted and nontargeted screening strategies applicable to complex food matrices, effectively replacing several matrix-specific targeted methods. Moreover, this comprehensive characterization provides crucial information for the next steps of the risk assessment.

The human toxicological risk assessment of per- and polyfluoroalkyl substances (PFAS) is challenging, due to their sheer number and structural diversity, but also the paucity of the toxicity data required to characterize them. The development of Next Generation Physiologically Based Kinetic (NG-PBK) models may assist in overcoming this challenge. The mechanistic nature of NG-PBK models allows for their extrapolation from data-rich PFAS, such as perfluorooctanoic acid (PFOA), to data-poor ones, facilitating their application in Next Generation Risk Assessment (NGRA). The present study proposes a NG-PBK model for PFOA in humans, parametrized exclusively using in vitro-, and in silico-derived data. The model describes the toxicokinetic processes of 1) partitioning to plasma and tissue proteins, 2) partitioning to cell membrane lipids, 3a) transporter-mediated entero-hepatic circulation and 3b) renal elimination and reabsorption, and 4) elimination via menstruation. Global sensitivity analysis indicated that the model was most sensitive to the fraction unbound in plasma, active-transport parameters, and tissue-plasma partition coefficients. The model was equivalent to already available validated human PFOA-PBK models, while compared to those, it is not calibrated to observed animal, nor human data, illustrating its strength in being mechanistic. The serum concentrations and half-lives predicted by the NG-PBK model were within the ranges of those reported in human volunteer and biomonitoring (HBM) studies, demonstrating the model's capacity to accurately predict PFOA toxicokinetics on exposure estimates. Extrapolation of the NG-PBK model to other PFAS, in conjunction with its integration with HBM data, will facilitate the NGRA of PFAS. This is particularly relevant given the paucity of in vivo data for most PFAS, ensuring compliance with the 3R principles.

Nephrotoxicity is a major concern in the safety assessment of chemicals and drugs. Computational modeling, particularly the use of quantitative adverse outcome pathways (qAOPs), offers a promising strategy to improve the translation from in vitro to in vivo, thereby facilitating reliable predictions of in vivo adverse outcomes and potentially reducing the need for animal testing. Platinum-based drugs are widely used in chemotherapy, yet their clinical application is frequently constrained by nephrotoxic effects. Here, we focus on the development of ordinary differential equation (ODE)-based qAOPs for platinum-induced nephrotoxicity by defining both an in vitro and an in vivo data-driven model. The in vitro model incorporates newly generated, time-course gene expression and propidium iodide (PI) staining data from RPTEC/TERT1 cells exposed to cisplatin. The in vivo model employs published rat data, including dose-response platinum kinetics as well as single-dose time-course platinum kinetics, gene expression, and histopathology data. Our quantitative approach shows that key processes in the AOP related to immune system activity are nonlinear. Specifically, clearance of necrotic kidney cells by immune system activity counters damage accumulation on a timescale of days, yet low-level inflammation still cumulatively affects kidney failure in the long run. Moreover, we perform quantitative in vitro to in vivo extrapolation (QIVIVE) to link the 2 models. With this approach, in vivo adverse outcome predictions can be made in the future not only for platinum-based compounds but also for the safety assessment of other chemicals and drugs, reducing the need for animal testing.

New approach methodologies (NAMs), such as in vitro and in silico methods, provide opportunities to replace, reduce, and refine animal studies for chemical risk assessment, decrease experi­mental lead time, and increase predictivity for human health. However, starting from or including NAMs introduces uncertainties that need to be addressed and quantified to work towards regu­latory implementation of NAMs within risk assessment. A theoretical risk assessment based on the developmental neurotoxicity (DNT) of chlorpyrifos and its active metabolite chlorpyrifos-oxon was performed to highlight important extrapolations and identify and quantify uncertainties when extrapolating NAM-based data to estimate human health risk. Cell viability of differentiating human neuronal progenitor cells was used as a proxy for DNT during early pregnancy. A physiologically based kinetic (PBK) model describing early pregnancy was developed to facilitate quantitative in-vitro-to-in-vivo extrapolation (qIVIVE) to estimate the external dose leading to DNT onset. An in vitro-based health-based threshold value (HBTV) was derived for the onset of DNT caused by chlorpyrifos and chlorpyrifos-oxon following chlorpyrifos exposure. This value was higher than the maternal exposure associated with the onset of DNT based on epidemiological data, showing the need for improvements in the NAM-based risk assessment of chemicals suspected of DNT. Improve­ments relate to the selection and refinement of the in vitro endpoint and assay and defining the quantitative relationship between the selected key event and the adverse outcome and provide a starting point for future research to aid regulatory implementation.

The Virtual Human Platform for Safety Assessment (VHP4Safety) project aims to build a virtual human platform (VHP) to protect human health and revolutionize the safety assessment of chemicals and pharmaceuticals by transitioning from animal-based to human-based approaches. The goal of this article is to introduce the project and its interdisciplinary approach to co-creation with multiple academic, regulatory, industrial and societal partners covering the entire safety assessment knowledge chain. Three research lines drive the project: 1) building the VHP; 2) feeding the VHP with human data; and 3) implementing the VHP. The project focusses on three case studies that incorporate human-relevant scenarios not included in current animal-based safety assessment strategies. The VHP is built on tools and services, including pharmacokinetic and computational models, and integrates several data sources within each case study, including data on human physiology, epidemiology, toxicokinetic and-dynamic parameters, as well as data on chemical characteristics and exposures. In addition, the VHP integrates new data generated within the project using new approach methodologies representing key events within adverse outcome pathways. Implementation of the VHP is investigated using an innovation systems approach, engaging stakeholders, and organizing training and education. Central to the VHP4Safety project is our co-creative approach, which is facilitated by biannual designathons and hackathons that foster active involvement of all project participants from over 30 partner organizations. By integrating technological innovations with transparency and stakeholder collaboration, the VHP4Safety project will help shape the transition to next generation safety assessment in which animal testing becomes redundant.

Animal-based toxicity tests inadequately predict human nephrotoxicity, hence the call for human-based new approach methodologies. Nephrotoxic chemicals often accumulate in the proximal tubule as transporters on the polarized membrane of these cells can actively take-up chemicals. The aim of the present study was to develop next generation human physiologically-based kinetic (PBK) models to simulate plasma and kidney concentrations and predict the nephrotoxicity of i.v.-administered cisplatin as the model compound. The PBK models were used to predict nephrotoxic doses of this chemotherapeutic by quantitative in vitro to in vivo extrapolation (QIVIVE) of literature available concentration-response relationships of cytotoxicity in renal cell lines. Two forms of the PBK model were developed: one with a simple single kidney compartment and another including a multi-compartmental kidney that includes organic cation transporter 2 (OCT2)-mediated active renal secretion. Observed single dose of 50 mg/m2 lead to nephrotoxicity in 30 % of patients. Using only in vitro and in silico-derived pharmacokinetic parameters, human benchmark dose levels (BMDL)30 values were calculated for all QIVIVE obtained dose-response curves using the maximum concentration in kidney blood or proximal tubule and the area under the concentration-time curve (AUC) in the proximal tubule. Based on the AUC in proximal tubule, the complex form of the model predicts nephrotoxicity in 30 % of patients best, between 23.1 and 65 mg/m2. This study provides a case study for next generation risk assessment using PBK models that incorporate active renal secretion to better predict plasma concentration-time profiles are well as in vivo nephrotoxic dose levels.

Organophosphate (OP) pesticide residues are frequently found in the environment and food products, with their acute exposures to humans posing a risk of neurotoxicity through acetylcholinesterase (AChE) inhibition. The aim of the present study is to develop a New Approach Methodology (a population-based physiologically based kinetic and toxicodynamic (PBK-TD) model) to define a health-based guidance value (HBGV) for acute exposure to diazinon as the model OP, taking into account human interindividual variability in physiology, toxicokinetics and toxicodynamics. Physiological and chemical-specific inputs for the PBK-TD model were obtained from literature or by in silico-in vitro strategies. Using this population model and Monte Carlo simulations, the dosedependent response for DZN-induced erythrocyte AChE inhibition was generated to provide a point of departure (POD) for defining an acute reference dose (ARfD). The model simulates the toxicokinetic and toxicity data observed in humans well, and results reveal that toxicokinetic and not toxicodynamic variations are the main driver of the overall interindividual variability in susceptibility towards acute DZN exposure. The POD predicted for the sensitive adults is in agreement with a previously reported human no-observed-adverse-effect level (NOAEL). It is concluded that the population PBK-TD modeling defines a novel way to derive a POD for human health risk assessment with the incorporation of interindividual differences. In the next step, the inclusion of correlations between certain model parameters as well as cholinesterase inhibition in tissues other than the blood is expected to be a further refinement.

Adverse outcome pathways (AOPs) describe toxicological processes from a dynamic perspective by linking a molecular initiating event to a specific adverse outcome via a series of key events and key event relationships. In the field of computational toxicology, AOPs can potentially facilitate the design and development of in silico prediction models for hazard identification. Various AOPs have been introduced for several types of hepatotoxicity, such as steatosis, cholestasis, fibrosis, and liver cancer. This chapter provides an overview of AOPs on hepatotoxicity, including their development, assessment, and applications in toxicology.

Organophosphate (OP) pesticides are common environmental contaminants, of which the resulting acetylcholinesterase (AChE) inhibition and concomitant neurotoxic effects following exposure remain a global concern. To evaluate the safety upon acute exposure to OP pesticides, the Dietary Comparator Ratio (DCR) approach was used for the first time for this class of chemicals. Six OPs including chlorpyrifos, diazinon, fenitrothion, methyl parathion, profenofos, and chlorfenvinphos were selected as model compounds. Seventy-four reports of human exposures were collected, and a DCR value at each defined exposure level was calculated with in vitro determined AChE inhibition potency and in silico simulated internal exposures. Results indicate that the DCR outcomes are comparable to the actual knowledge on the presence or absence of in vivo AChE inhibition and adverse effects for the respective exposure scenarios. Of all collected scenarios, only four false positives but no false negatives were obtained. No safety concern on acute neurotoxicity appears to be raised for the evaluated environmental exposure scenarios to OPs. To conclude, the described DCR approach provides an adequate evaluation of the OP-induced adverse outcomes for humans, shedding light on its utility for 3Rs-compliant safety assessment of chemicals with different toxicity mechanisms especially for which in vitro bioassays are available.

To underpin scientific evaluations of chemical risks, agencies such as the European Food Safety Authority (EFSA) heavily rely on the outcome of systematic reviews, which currently require extensive manual effort. One specific challenge constitutes the meaningful use of vast amounts of valuable data from new approach methodologies (NAMs) which are mostly reported in an unstructured way in the scientific literature. In the EFSA-initiated project ‘AI4NAMS’, the potential of large language models (LLMs) was explored. Models from the GPT family, where GPT refers to Generative Pre-trained Transformer, were used for searching, extracting, and integrating data from scientific publications for NAM-based risk assessment. A case study on bisphenol A (BPA), a substance of very high concern due to its adverse effects on human health, focused on the structured extraction of information on test systems measuring biologic activities of BPA. Fine-tuning of a GPT-3 model (Curie base model) for extraction tasks was tested and the performance of the fine-tuned model was compared to the performance of a ready-to-use model (text-davinci-002). To update findings from the AI4NAMS project and to check for technical progress, the fine-tuning exercise was repeated and a newer ready-to-use model (text-davinci-003) served as comparison. In both cases, the fine-tuned Curie model was found to be superior to the ready-to-use model. Performance improvement was also obvious between text-davinci-002 and the newer text-davinci-003. Our findings demonstrate how fine-tuning and the swift general technical development improve model performance and contribute to the growing number of investigations on the use of AI in scientific and regulatory tasks.

Since their introduction into agriculture, the toxicity of organophosphate (OP) pesticides has been widely studied in animal models. However, next generation risk assessment (NGRA) intends to maximize the use of novel approach methodologies based on in vitro and in silico methods. Therefore, this study describes the development and evaluation of a generic physiologically based kinetic (PBK) model for acute exposure to OP pesticides in rats and humans using quantitative structure property relationships and data from published in vitro studies. The models were evaluated using in vivo studies from the literature for chlorpyrifos, diazinon, fenitrothion, methyl-parathion, ethyl-parathion, dimethoate, chlorfenvinphos, and profenofos. Evaluation was performed by comparing simulated and in vivo observed time profiles for blood, plasma, or urinary concentrations and other toxicokinetic parameters. Of simulated concentration-time profiles, 87 and 91% were within a 5-fold difference from observed toxicokinetic data from rat and human studies, respectively. Only for dimethyl-organophosphates further refinement of the model is required. It is concluded that the developed generic PBK model provides a new tool to assess species differences in rat and human kinetics of OP pesticides. This approach provides a means to perform NGRA for these compounds and could also be adopted for other classes of compounds.

The first Stakeholder Network Meeting of the EU Horizon 2020-funded ONTOX project was held on 13-14 March 2023, in Brussels, Belgium. The discussion centred around identifying specific challenges, barriers and drivers in relation to the implementation of non-animal new approach methodologies (NAMs) and probabilistic risk assessment (PRA), in order to help address the issues and rank them according to their associated level of difficulty. ONTOX aims to advance the assessment of chemical risk to humans, without the use of animal testing, by developing non-animal NAMs and PRA in line with 21st century toxicity testing principles. Stakeholder groups (regulatory authorities, companies, academia, non-governmental organisations) were identified and invited to participate in a meeting and a survey, by which their current position in relation to the implementation of NAMs and PRA was ascertained, as well as specific challenges and drivers highlighted. The survey analysis revealed areas of agreement and disagreement among stakeholders on topics such as capacity building, sustainability, regulatory acceptance, validation of adverse outcome pathways, acceptance of artificial intelligence (AI) in risk assessment, and guaranteeing consumer safety. The stakeholder network meeting resulted in the identification of barriers, drivers and specific challenges that need to be addressed. Breakout groups discussed topics such as hazard versus risk assessment, future reliance on AI and machine learning, regulatory requirements for industry and sustainability of the ONTOX Hub platform. The outputs from these discussions provided insights for overcoming barriers and leveraging drivers for implementing NAMs and PRA. It was concluded that there is a continued need for stakeholder engagement, including the organisation of a 'hackathon' to tackle challenges, to ensure the successful implementation of NAMs and PRA in chemical risk assessment.

The global burden of Inflammatory bowel disease (IBD) has been rising over the last decades. IBD is an intestinal disorder with a complex and largely unknown etiology. The disease is characterized by a chronically inflamed gastrointestinal tract, with intermittent phases of exacerbation and remission. This compromised intestinal barrier can contribute to, enhance, or even enable the toxicity of drugs, food-borne chemicals and particulate matter. This review discusses whether the rising prevalence of IBD in our society warrants the consideration of IBD patients as a specific population group in toxicological safety assessment. Various in vivo, ex vivo and in vitro models are discussed that can simulate hallmarks of IBD and may be used to study the effects of prevalent intestinal inflammation on the hazards of these various toxicants. In conclusion, risk assessments based on healthy individuals may not sufficiently cover IBD patient safety and it is suggested to consider this susceptible subgroup of the population in future toxicological assessments.

The future of risk assessment cannot neglect to consider the vast literature produced through the application of new approach methodologies (NAMs). This, however, constitutes a challenge for the risk assessor, as the availability of data in this context is huge and heterogeneous both for the methods applied and the standardisation and quality of the results. The integration of results generated from NAMs is hence only feasible under some degree of automation of the risk assessment workflow, specifically for searching, extracting and integrating such results in "AOP-like" knowledge networks (AOP – Adverse Outcome Pathway). Artificial intelligence (AI) with its state-of-the-art methods and tools is one of the most promising sources to support automation of manual tasks among modern technologies. The present paper illustrates the results of the exploration of possible applications of AI to achieve this goal. After the introduction of an evaluation framework to quantitatively assess these tools and methods, the results of the implementation of six selected case studies with the support of a selection of such tools in a dedicated workflow are presented. A qualitative survey of the state-of-the-art tools and methods, which also incorporates the experience gathered during the case study implementation, is then presented. Finally, recommendations are formulated which address the main aspects identified through the case study implementation that should, in the opinion of the authors, be pursued by EFSA in the context of its SPIDO NAMs and AI roadmaps. In summary, potentials for AI tool support could be identified throughout the workflow. Although many of the tasks can be supported by (semi-)automation, experience showed that subject matter experts need to be involved in all workflow steps.

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.

The disruption of thyroid hormone homeostasis by hexabromocyclododecane (HBCD) in rodents is hypothesized to be due to HBCD increasing the hepatic clearance of thyroxine (T4). The extent to which these effects are relevant to humans is unclear. To evaluate HBCD effects on humans, the activation of key hepatic nuclear receptors and the consequent disruption of thyroid hormone homeostasis were studied in different human hepatic cell models. The hepatoma cell line, HepaRG, cultured as two-dimensional (2D), sandwich (SW) and spheroid (3D) cultures, and primary human hepatocytes (PHH) cultured as sandwich were exposed to 1 and 10 µM HBCD and characterized for their transcriptome changes. Pathway enrichment analysis showed that 3D models, followed by SW, had a stronger transcriptome response to HBCD, which is explained by the higher expression of hepatic nuclear receptors but also greater accumulation of HBCD measured inside cells in these models. The Pregnane X receptor pathway is one of the pathways most upregulated across the three hepatic models, followed by the constitutive androstane receptor and general hepatic nuclear receptors pathways. Lipid metabolism pathways had a downregulation tendency in all exposures and in both PHH and the three cultivation modes of HepaRG. The activity of enzymes related to PXR/CAR induction and T4 metabolism were evaluated in the three different types of HepaRG cultures exposed to HBCD for 48 h. Reference inducers, rifampicin and PCB-153 did affect 2D and SW HepaRG cultures' enzymatic activity but not 3D. HBCD did not induce the activity of any of the studied enzymes in any of the cell models and culture methods. This study illustrates that for nuclear receptor-mediated T4 disruption, transcriptome changes might not be indicative of an actual adverse effect. Clarification of the reasons for the lack of translation is essential to evaluate new chemicals' potential to be thyroid hormone disruptors by altering thyroid hormone metabolism.

The future of risk assessment cannot neglect to consider the vast literature produced through the application of new approach methodologies (NAMs). This, though, constitutes a challenge for the risk assessor, as the production in this context is huge and heterogeneous both for the methods applied and the standardisation and quality of the results. The integration of results from NAMs is hence only feasible under some degree of automation of the risk assessment workflow, specifically for searching, extracting and integrating such results in “AOP-like” knowledge networks (AOP – Adverse Outcome Pathway). Artificial intelligence (AI) with its state-of-the-art methods and tools is the most promising source for help among modern technologies supporting automation of manual tasks. Within the broader context of investigating possible applications of AI to achieve this goal, the present paper illustrates an evaluation framework to quantitatively assess these tools and methods and, in turn, support the selection of the most promising among them for a (at least partial) automation of the overall workflow. It also provides a survey of the state-of-the-art tools and methods identified at present which can then be assessed for specific use cases with the help of the introduced evaluation framework.