Thousands of chemicals have not been assessed for developmental neurotoxicity (DNT) despite potential human exposure, prompting efforts to screen chemicals for possible DNT hazard using in vitro approaches. Ninety-five chemicals were screened in in a battery including in vitro assays covering the key neurodevelopmental processes (KNDPs) of proliferation, apoptosis, cell migration, neuronal differentiation, neurite outgrowth, neuronal maturation and synaptogenesis, oligodendrocyte differentiation, and network formation using human and rat (primary cortical) cells. An in vivo assay of zebrafish neurobehavior was also included. This assay battery was sensitive enough to detect effects in the neurodevelopmental processes modeled. The chemicals comprised negative controls and several use classes: drugs, flame retardants, fungicides, herbicides, industrial chemicals (including PFAS), and insecticides. A data analysis pipeline was developed using a benchmark concentration approach and included multi-objective optimization (Pareto) and ToxPi analysis to integrate data from the individual assays to evaluate DNT hazard and prioritize chemicals, ranking them for further evaluation. Although the zebrafish assay was the most sensitive for detecting potency, selectivity varies significantly by assay and chemical class with neurite outgrowth assays and fungicides exhibiting the highest percentage of selective hits. Chemicals from the fungicide, drug, and insecticide use classes had higher priority in the Pareto and ToxPi analysis based on having the most potent and selective activity in the assays. These prioritization methods represent a novel and useful approach for comparative evaluation of potential developmental neurotoxicity hazard in large groups of chemicals.

Artificial intelligence (AI) is increasingly used to support predictive, mechanistic, and human-relevant toxicology at scale. However, its integration into regulatory science - particularly in drug development - remains uneven, because encouraging technical performance has not yet translated automatically into regulatory trust. Representative AI toxicology studies now span datasets from roughly 103 chemicals to >3 × 104 peptide or chemical records and report performance ranging from modest in prospective screening settings to strong on narrower, well-curated endpoints. This manuscript presents a critical analysis of the dual nature of AI in toxicology. We review the state of the art in AI-enabled applications, ranging from Green Toxicology and the Human Exposome to specific challenges in safety assessment for biologics and synthetic peptides. Particular attention is given to the gap between rapid model development and regulatory acceptance, highlighted by the challenges of model interpretability, dataset bias, insufficient external validation, and the assessment of complex endpoints like immunogenicity. To navigate these complexities, we discuss the next-generation “e-validation” framework and emphasize the TREAT principle - Trustworthiness, Reproducibility, Explainability, Applicability, and Transparency - as a foundation for building regulatory trust. We hypothesize that AI-based methods in toxicology can achieve regulatory acceptance when they satisfy the TREAT criteria and undergo continuous e-validation within a clearly defined context of use. This framework distinguishes credible AI applications from “snake oil” by establishing measurable criteria for trust-building, including dataset provenance, external validation, uncertainty characterization, and life-cycle monitoring. We argue that AI is neither a miracle cure nor a technological illusion, but a powerful evidence engine that can contribute to a more predictive and ethical toxicological science when it is rigorously validated.

Central nervous system (CNS) toxicities remain a major cause of drug attrition and represent a persistent challenge in predicting neurological risk during drug development. Limitations in the predictive resolution and translational relevance of conventional nonclinical paradigms contribute to uncertainty in identifying and interpreting neurotoxicity signals. This manuscript examines key challenges in CNS safety assessment and highlights emerging strategies to improve early detection and prediction of neurological risk. Through a series of case studies, we demonstrate practical approaches for interpreting CNS safety signals and integrating emerging methodologies into nonclinical safety assessment. Examples include sensory and seizure-related endpoints in nonclinical studies and the use of electroencephalography (EEG) to improve detection and characterization of seizure liability. We also highlight the expanding role of advanced sensor technologies and artificial intelligence (AI) in enabling continuous, noninvasive monitoring of animal behavior. In addition, an Integrated Approach to Testing and Assessment (IATA) case study demonstrates how systematic integration of mechanistic data, traditional toxicology findings, and exposure modeling can support regulatory decision-making while aligning with the 3Rs principles (replace, reduce, refine animal testing). Finally, we present regulatory CNS case studies in drug development. Collectively, these approaches enable quantitative assessment of neurological function across circadian cycles, reduce reliance on episodic observer-dependent measurements, and illustrate how integrating refined in vivo methods with New Approach Methodologies (NAMs) and digital technologies can improve prediction of neurological risk and strengthen translation from nonclinical findings to human outcomes in CNS drug development.

Animal studies are commonly used in drug development and in chemical and environmental toxicology to predict human toxicity, but their reliability, particularly in the central nervous system (CNS), is limited. For example, animal models often fail to predict drug-induced seizures, leading to unforeseen convulsions in clinical trials. Evaluating environmental compounds, such as pesticides, also poses challenges due to time and resource constraints, resulting in compounds remaining untested. To address these limitations, a government-industry collaboration identified 27 biological target families linked to seizure mechanisms by combining key events from adverse outcome pathways (AOPs) with drug discovery data. Over a hundred in vitro assay endpoints were identified, covering 26 of the target families, including neurotransmitter receptors, transporters, and voltage-gated calcium channels. A review of reference compounds identified 196 seizure-inducing and 34 seizure-negative chemicals, with 80% being tested in the in vitro assays. However, some target familes were more data-poor than others, highlighting significant data gaps. This proof-of-concept study demonstrates how mechanistic seizure liability can be assessed using an AOP framework and in vitro data. It underscores the need for expanded screening panels to include additional seizure-relevant targets. By integrating mechanistic insights into early drug development and environmental risk assessment, this approach enhances compound prioritization, complements animal studies, and optimizes resource use. Ultimately, this strategy refines CNS safety evaluation in drug development, improves public health protection to neurotoxicants, and bridges knowledge gaps.

Seizures are the hallmark of treatment-resistant epilepsy and are also a frequent adverse event in drug safety testing. Here we present a critical evaluation of new approach methodologies (NAMs) in epilepsy research and seizure liability. Specifically, recent data have shown that seizure can be profiled using an in vitro seizure liability assay (iSLA) comprising a microelectrode array (MEA) of human induced pluripotent stem cell (iPSC) neurons coupled with a human ion channel panel. Adverse outcome pathways (AOPs) also offer a powerful NAM framework to map how key events can lead to an adverse outcome such as seizure. These NAMs offers an innovative approach to antiseizure medication research as well as to seizure risk testing in drug discovery and development.

Organophosphorus flame retardants (OPFRs) are abundant and persistent in the environment but have limited toxicity information. Their similarity in structure to organophosphate pesticides presents great concern for developmental neurotoxicity (DNT). However, current in vivo testing is not suitable to provide DNT information on the amount of OPFRs that lack data. Over the past decade, an in vitro battery was developed to enhance DNT assessment, consisting of assays that evaluate cellular processes in neurodevelopment and function. In this study, behavioral data of small model organisms were also included. To assess if these assays provide sufficient mechanistic coverage to prioritize chemicals for further testing and/or identify hazards, an integrated approach to testing and assessment (IATA) was developed with additional information from the Integrated Chemical Environment (ICE) and the literature. Human biomonitoring and exposure data were identified and physiologically-based toxicokinetic models were applied to relate in vitro toxicity data to human exposure based on maximum plasma concentration. Eight OPFRs were evaluated, including aromatic OPFRs (triphenyl phosphate (TPHP), isopropylated phenyl phosphate (IPP), 2-ethylhexyl diphenyl phosphate (EHDP), tricresyl phosphate (TMPP), isodecyl diphenyl phosphate (IDDP), tert-butylphenyl diphenyl phosphate (BPDP)) and halogenated FRs ((Tris(1,3-dichloro-2-propyl) phosphate (TDCIPP), tris(2-chloroethyl) phosphate (TCEP)). Two representative brominated flame retardants (BFRs) (2,2′4,4′-tetrabromodiphenyl ether (BDE-47) and 3,3′,5,5′-tetrabromobisphenol A (TBBPA)) with known DNT potential were selected for toxicity benchmarking. Data from the DNT battery indicate that the aromatic OPFRs have activity at similar concentrations as the BFRs and should therefore be evaluated further. However, these assays provide limited information on the mechanism of the compounds. By integrating information from ICE and the literature, endocrine disruption was identified as a potential mechanism. This IATA case study indicates that human exposure to some OPFRs could lead to a plasma concentration similar to those exerting in vitro activities, indicating potential concern for human health.

Two organophosphate esters used as flame retardants and plasticizers, triphenyl phosphate (TPHP) and isopropylated phenyl phosphate (IPP), have been detected in environmental samples around the world. Human exposure primarily occurs via oral ingestion with reported higher concentrations in children. Currently, there are no data to evaluate potential risk from exposure to either TPHP or IPP during fetal development. These short-term perinatal studies in rats provide preliminary toxicity data for TPHP and IPP, including information on transfer to fetus/offspring and across the pup blood-brain barrier. In separate experiments, TPHP or IPP were administered via dosed feed at concentrations 0, 1000, 3000, 10 000, 15 000, or 30 000ppm to time-mated Hsd:Sprague Dawley SD rats from gestation day (GD) 6 through postnatal day (PND) 28; offspring were provided dosed feed at the same concentration as their dam (PND 28-PND 56). TPHP- and IPP-related toxicity resulted in removal of both 30 000ppm groups on GD 12 and 15 000ppm IPP group after parturition. Body weight and organ weights were impacted with exposure in remaining dams. Reproductive performance was perturbed at >= 10 000 ppm TPHP and all IPP exposure groups. In offspring, both TPHP- and IPP-related toxicity was noted in pups at >= 10 000ppm as well as reduction in bodyweights, delays in pubertal endpoints, and/or reduced cholinesterase enzyme activity starting at 1000ppm TPHP or IPP. Preliminary internal dose assessment indicated gestational and lactational transfer following exposure to TPHP or IPP. These findings demonstrate that offspring development is sensitive to 1000ppm TPHP or IPP exposure.

Neurotoxicology is the study of adverse effects on the structure or function of the developing or mature adult nervous system following exposure to chemical, biological, or physical agents. The development of more informative alternative methods to assess developmental (DNT) and adult (NT) neurotoxicity induced by xenobiotics is critically needed. The use of such alternative methods including in silico approaches that predict DNT or NT from chemical structure (e.g., statistical-based and expert rule-based systems) is ideally based on a comprehensive understanding of the relevant biological mechanisms. This paper discusses known mechanisms alongside the current state of the art in DNT/NT testing. In silico approaches available today that support the assessment of neurotoxicity based on knowledge of chemical structure are reviewed, and a conceptual framework for the integration of in silico methods with experimental information is presented. Establishing this framework is essential for the development of protocols, namely standardized approaches, to ensure that assessments of NT and DNT based on chemical structures are generated in a transparent, consistent, and defendable manner.

Federal agencies are making significant investments to advance predictive approaches to evaluate chemical hazards and risks. Environmental Defense Fund (EDF) believes that engagement with the broader scientific community is critical to building and maintaining a strong biological foundation for these approaches.On June 18–19, 2015, EDF organized a meeting to 1) foster a conversation between federal scientists advancing predictive approaches and environmental health researchers investigating environmental exposures and neurological outcomes, and 2) explore opportunities and challenges for the use of federal chemical high-throughput in vitro screening (HTS) data in hypothesis-driven research toward, ultimately, improved data for public health decision-making.The meeting achieved its objectives. Government scientists showcased their chemical testing programs and vision for how emerging data may be used to meet agency missions. Environmental health researchers shared their experiences using federal HTS data, offered recommendations for strengthening federal HTS platforms, and expressed great interest in continued engagement with evolving federal chemical testing initiatives.The meeting provided an invaluable exchange between two scientific communities with a shared interest in protecting public health from harmful environmental exposures, but who have not sufficiently engaged with each other. Discussions identified opportunities and work ahead for the use of HTS data in hypothesis-driven research. Though the meeting focused on neurological outcomes, the purpose, objectives and experience of the meeting are broadly applicable. EDF strongly encourages more discourse and collaboration between federal and non-government scientists working to understand environmental influences on health outcomes.

Compound toxicity data obtained from independent zebrafish laboratories can vary vastly, complicating the use of zebrafish screening for regulatory decisions. Differences in the assay protocol parameters are the primary source of variability. We investigated this issue by utilizing data from the NTP DNT-DIVER database (https://doi.org/10.22427/NTP-DATA-002-00062-0001-0000-1, last accessed June 2, 2022), which consists of data from zebrafish developmental toxicity (devtox) and locomotor response (designated as "neurotox") screens from 3 independent laboratories, using the same set of 87 compounds. The data were analyzed using the benchmark concentration (BMC) modeling approach, which estimates the concentration of interest based on a predetermined response threshold. We compared the BMC results from 3 laboratories (A, B, C) in 3 toxicity outcome categories: mortality, cumulative devtox, and neurotox, in terms of activity calls and potency values. We found that for devtox screening, laboratories with similar/same protocol parameters (B vs C) had an active call concordance as high as 86% with negligible potency difference. For neurotox screening, active call concordances between paired laboratories are lower than devtox screening (highest 68%). When protocols with different protocol parameters were compared, the concordance dropped, and the potency shift was on average about 3.8-fold for the cumulative devtox outcome and 5.8-fold for the neurotox outcome. The potential contributing protocol parameters for potency shift are listed or ranked. This study provides a quantitative assessment of the source of variability in zebrafish screening protocols and sets the groundwork for the ongoing Systematic Evaluation of the Application of Zebrafish in Toxicology effort at the National Toxicology Program.

Rodent alveolar/bronchiolar carcinomas (ABC) that arise either spontaneously or due to chemical exposure are similar to a subtype of lung adenocarcinomas in humans. B6C3F1/N mice and F344/NTac rats exposed to cobalt metal dust (CMD) by inhalation developed ABCs in a dose dependent manner. In CMD-exposed mice, the incidence of Kras mutations in ABCs was 67% with 80% of those being G to T transversions on codon 12 suggesting a role of oxidative stress in the pathogenesis. In vitro studies, such as DMPO (5,5-dimethyl-1-pyrroline N-oxide) immune-spin trapping assay, and dihydroethidium (DHE) fluorescence assay on A549 and BEAS-2B cells demonstrated increased oxidative stress due to cobalt exposure. In addition, significantly increased 8-oxo-dG adducts were demonstrated by immunohistochemistry in lungs from mice exposed to CMD for 90 days. Furthermore, transcriptomic analysis on ABCs arising spontaneously or due to chronic CMD-exposure demonstrated significant alterations in canonical pathways related to MAPK signaling (IL-8, ErbB, Integrin, and PAK pathway) and oxidative stress (PI3K/AKT and Melatonin pathway) in ABCs from CMD-exposed mice. Oxidative stress can stimulate PI3K/AKT and MAPK signaling pathways. Nox4 was significantly upregulated only in CMD-exposed ABCs and NOX4 activation of PI3K/AKT can lead to increased ROS levels in human cancer cells. The gene encoding Ereg was markedly up-regulated in CMD-exposed mice. Oncogenic KRAS mutations have been shown to induce EREG overexpression. Collectively, all these data suggest that oxidative stress plays a significant role in CMD-induced pulmonary carcinogenesis in rodents and these findings may also be relevant in the context of human lung cancers.

Due to their neurodevelopmental toxicity, flame retardants (FRs) like polybrominated diphenyl ethers are banned from the market and replaced by alternative FRs, like organophosphorus FRs, that have mostly unknown toxicological profiles. To study their neurodevelopmental toxicity, we evaluated the hazard of several FRs including phased-out polybrominated FRs and organophosphorus FRs: 2,2′,4,4′-tetrabromodiphenylether (BDE-47), 2,2′,4,4′,5-pentabromodiphenylether (BDE-99), tetrabromobisphenol A, triphenyl phosphate, tris(2-butoxyethyl) phosphate and its metabolite bis-(2-butoxyethyl) phosphate, isodecyl diphenyl phosphate, triphenyl isopropylated phosphate, tricresyl phosphate, tris(1,3-dichloro-2-propyl) phosphate, tert-butylphenyl diphenyl phosphate, 2-ethylhexyl diphenyl phosphate, tris(1-chloroisopropyl) phosphate, and tris(2-chloroethyl) phosphate. Therefore, we used a human cell–based developmental neurotoxicity (DNT) in vitro battery covering a large variety of neurodevelopmental endpoints. Potency according to the respective most sensitive benchmark concentration (BMC) across the battery ranked from <1 μM (5 FRs), 1<10 μM (7 FRs) to the >10 μM range (3 FRs). Evaluation of the data with the ToxPi tool revealed a distinct ranking (a) than with the BMC and (b) compared to the ToxCast data, suggesting that DNT hazard of these FRs is not well predicted by ToxCast assays. Extrapolating the DNT in vitro battery BMCs to human FR exposure via breast milk suggests low risk for individual compounds. However, it raises a potential concern for real-life mixture exposure, especially when different compounds converge through diverse modes-of-action on common endpoints, like oligodendrocyte differentiation in this study. This case study using FRs suggests that human cell–based DNT in vitro battery is a promising approach for neurodevelopmental hazard assessment and compound prioritization in risk assessment. Graphical abstract