Hasil untuk "Toxicology. Poisons"

Yueyan Dong, Minghui Xu, Qin Hu et al.

Low-Rank Adaptation (LoRA) has become a popular solution for fine-tuning large language models (LLMs) in federated settings, dramatically reducing update costs by introducing trainable low-rank matrices. However, when integrated with frameworks like FedIT, LoRA introduces a critical vulnerability: clients submit $A$ and $B$ matrices separately, while only their product $AB$ determines the model update, yet this composite is never directly verified. We propose Gradient Assembly Poisoning (GAP), a novel attack that exploits this blind spot by crafting individually benign $A$ and $B$ matrices whose product yields malicious updates. GAP operates without access to training data or inter-client coordination and remains undetected by standard anomaly detectors. We identify four systemic vulnerabilities in LoRA-based federated systems and validate GAP across LLaMA, ChatGLM, and GPT-2. GAP consistently induces degraded or biased outputs while preserving surface fluency, reducing BLEU by up to 14.5\%, increasing factual and grammatical errors by over 800\%, and maintaining 92.6\% long-form response length. These results reveal a new class of stealthy, persistent threats in distributed LoRA fine-tuning.

Hongyi Miao, Jun Jia, Xincheng Wang et al.

Recent advances in visual-language alignment have endowed vision-language models (VLMs) with fine-grained image understanding capabilities. However, this progress also introduces new privacy risks. This paper first proposes a novel privacy threat model named identity-affiliation learning: an attacker fine-tunes a VLM using only a few private photos of a target individual, thereby embedding associations between the target facial identity and their private property and social relationships into the model's internal representations. Once deployed via public APIs, this model enables unauthorized exposure of the target user's private information upon input of their photos. To benchmark VLMs' susceptibility to such identity-affiliation leakage, we introduce the first identity-affiliation dataset comprising seven typical scenarios appearing in private photos. Each scenario is instantiated with multiple identity-centered photo-description pairs. Experimental results demonstrate that mainstream VLMs like LLaVA, Qwen-VL, and MiniGPT-v2, can recognize facial identities and infer identity-affiliation relationships by fine-tuning on small-scale private photographic dataset, and even on synthetically generated datasets. To mitigate this privacy risk, we propose DP2-VL, the first Dataset Protection framework for private photos that leverages Data Poisoning. Though optimizing imperceptible perturbations by pushing the original representations toward an antithetical region, DP2-VL induces a dataset-level shift in the embedding space of VLMs'encoders. This shift separates protected images from clean inference images, causing fine-tuning on the protected set to overfit. Extensive experiments demonstrate that DP2-VL achieves strong generalization across models, robustness to diverse post-processing operations, and consistent effectiveness across varying protection ratios.

Anirban Mukherjee, Hannah Hanwen Chang

Copyright enforcement rests on an evidentiary bargain: a plaintiff must show both the defendant's access to the work and substantial similarity in the challenged output. That bargain comes under strain when AI systems are trained through multi-generational pipelines with recursive synthetic data. As successive models are tuned on the outputs of its predecessors, any copyrighted material absorbed by an early model is diffused into deeper statistical abstractions. The result is an evidentiary blind spot where overlaps that emerge look coincidental, while the chain of provenance is too attenuated to trace. These conditions are ripe for "copyright laundering"--the use of multi-generational synthetic pipelines, an "AI Ouroboros," to render traditional proof of infringement impracticable. This Article adapts the "fruit of the poisonous tree" (FOPT) principle to propose a AI-FOPT standard: if a foundational AI model's training is adjudged infringing (either for unlawful sourcing or for non-transformative ingestion that fails fair-use), then subsequent AI models principally derived from the foundational model's outputs or distilled weights carry a rebuttable presumption of taint. The burden shifts to downstream developers--those who control the evidence of provenance--to restore the evidentiary bargain by affirmatively demonstrating a verifiably independent and lawfully sourced lineage or a curative rebuild, without displacing fair-use analysis at the initial ingestion stage. Absent such proof, commercial deployment of tainted models and their outputs is actionable. This Article develops the standard by specifying its trigger, presumption, and concrete rebuttal paths (e.g., independent lineage or verifiable unlearning); addresses counterarguments concerning chilling innovation and fair use; and demonstrates why this lineage-focused approach is both administrable and essential.

Deemah H. Tashman, Soumaya Cherkaoui

Cognitive radio networks (CRNs) are a key mechanism for alleviating spectrum scarcity by enabling secondary users (SUs) to opportunistically access licensed frequency bands without harmful interference to primary users (PUs). To address unreliable direct SU links and energy constraints common in next-generation wireless networks, this work introduces an adaptive, energy-aware hybrid reconfigurable intelligent surface (RIS) for underlay multiple-input single-output (MISO) CRNs. Distinct from prior approaches relying on static RIS architectures, our proposed RIS dynamically alternates between passive and active operation modes in real time according to harvested energy availability. We also model our scenario under practical hardware impairments and cascaded fading channels. We formulate and solve a joint transmit beamforming and RIS phase optimization problem via the soft actor-critic (SAC) deep reinforcement learning (DRL) method, leveraging its robustness in continuous and highly dynamic environments. Notably, we conduct the first systematic study of reward poisoning attacks on DRL agents in RIS-enhanced CRNs, and propose a lightweight, real-time defense based on reward clipping and statistical anomaly filtering. Numerical results demonstrate that the SAC-based approach consistently outperforms established DRL baselines, and that the dynamic hybrid RIS strikes a superior trade-off between throughput and energy consumption compared to fully passive and fully active alternatives. We further show the effectiveness of our defense in maintaining SU performance even under adversarial conditions. Our results advance the practical and secure deployment of RIS-assisted CRNs, and highlight crucial design insights for energy-constrained wireless systems.

Max Menssen, Carsten Kneuer, Gyamfi Akyianu et al.

In pre- and non-clinical toxicology, the reduction of animal use is highly desireable. Although approaches for possible sample size reduction in the concurrent control group were suggested previously under the virtual control groups framework for continuous endpoints, methodology that is applicable to binary outcomes that occur in long-term carcinogenicity studies is currently missing. In order to augment animals in the current control group with historical control data, we propose approaches that rely on dynamic Bayesian borrowing and simultaneous credible intervals for risk ratios. Several operation characteristics such as familywise error rate (FWER) and power are assessed via Monte-Carlo simulations and compared to the ones of approaches that rely on pooling of historical and current observations. It turned out that under optimal conditions, Bayesian approaches based on robustified prior distributions enable a substantial reduction of the control groups sample size, while still controlling the FWER up to a satisfactory level. Furthermore, at least to some extend, these approaches were able to protect against possible drift. This hightlights the potential of Bayesian study designs to reduce animal use in toxicology through re-use of the large pool of existing control data.

Yesenia Lopez Hernandez, Horatiu Vinerean, Saurabh Aggarwal et al.

Isotonitazene is a highly potent benzimidazole-derived synthetic opioid increasingly implicated in fatal intoxications. Despite growing forensic detection, controlled in vivo toxicity data remain limited. We determined the median lethal dose (LD50) of isotonitazene in adult male and female Fischer 344 rats using the Dixon up-and-down procedure following single intraperitoneal administration. Mortality and clinical signs were monitored for 14 days. The estimated LD50 was 0.08 mg/kg (95% CI 0.02996–0.182 mg/kg) in males and 0.25 mg/kg (95% CI 0.04841–0.768 mg/kg) in females, indicating greater male sensitivity. Rapid respiratory depression and neuromuscular rigidity occurred within minutes of dosing and preceded lethality. The narrow separation between survival and death demonstrates a steep dose–response relationship and limited safety margin. These findings provide quantitative data for toxicological risk assessment of nitazene opioids and underscore the importance of sex as a biological variable in opioid toxicity studies.

Runqi Sui

Retrieval-Augmented Generation (RAG) systems enhance response credibility and traceability by displaying reference contexts, but this transparency simultaneously introduces a novel black-box attack vector. Existing document poisoning attacks, where adversaries inject malicious documents into the knowledge base to manipulate RAG outputs, rely primarily on unrealistic white-box or gray-box assumptions, limiting their practical applicability. To address this gap, we propose CtrlRAG, a two-stage black-box attack that (1) constructs malicious documents containing misinformation or emotion-inducing content and injects them into the knowledge base, and (2) iteratively optimizes them using a localization algorithm and Masked Language Model (MLM) guided on reference context feedback, ensuring their retrieval priority while preserving linguistic naturalness. With only five malicious documents per target question injected into the million-document MS MARCO dataset, CtrlRAG achieves up to 90% attack success rates on commercial LLMs (e.g., GPT-4o), a 30% improvement over optimal baselines, in both *Emotion Manipulation* and *Hallucination Amplification* tasks. Furthermore, we show that existing defenses fail to balance security and performance. To mitigate this challenge, we introduce a dynamic *Knowledge Expansion* defense strategy based on *Parametric/Non-parametric Memory Confrontation*, blocking 78% of attacks while maintaining 95.5% system accuracy. Our findings reveal critical vulnerabilities in RAG systems and provide effective defense strategies.

Eduard Popescu, Adrian Groza, Andreea Cernat

The task here is to predict the toxicological activity of chemical compounds based on the Tox21 dataset, a benchmark in computational toxicology. After a domain-specific overview of chemical toxicity, we discuss current computational strategies, focusing on machine learning and deep learning. Several architectures are compared in terms of performance, robustness, and interpretability. This research introduces a novel image-based pipeline based on DenseNet121, which processes 2D graphical representations of chemical structures. Additionally, we employ Grad-CAM visualizations, an explainable AI technique, to interpret the model's predictions and highlight molecular regions contributing to toxicity classification. The proposed architecture achieves competitive results compared to traditional models, demonstrating the potential of deep convolutional networks in cheminformatics. Our findings emphasize the value of combining image-based representations with explainable AI methods to improve both predictive accuracy and model transparency in toxicology.

Xingfu Zhou, Pengfei Wang

Large Language Model (LLM) agents relying on external retrieval are increasingly deployed in high-stakes environments. While existing adversarial attacks primarily focus on content falsification or instruction injection, we identify a novel, process-oriented attack surface: the agent's reasoning style. We propose Reasoning-Style Poisoning (RSP), a paradigm that manipulates how agents process information rather than what they process. We introduce Generative Style Injection (GSI), an attack method that rewrites retrieved documents into pathological tones--specifically "analysis paralysis" or "cognitive haste"--without altering underlying facts or using explicit triggers. To quantify these shifts, we develop the Reasoning Style Vector (RSV), a metric tracking Verification depth, Self-confidence, and Attention focus. Experiments on HotpotQA and FEVER using ReAct, Reflection, and Tree of Thoughts (ToT) architectures reveal that GSI significantly degrades performance. It increases reasoning steps by up to 4.4 times or induces premature errors, successfully bypassing state-of-the-art content filters. Finally, we propose RSP-M, a lightweight runtime monitor that calculates RSV metrics in real-time and triggers alerts when values exceed safety thresholds. Our work demonstrates that reasoning style is a distinct, exploitable vulnerability, necessitating process-level defenses beyond static content analysis.

C. L. Chang, Y. -Y. Chang, M. Garcia-Sciveres et al.

Solid state phonon detectors used in the search for dark matter and coherent neutrino nucleus interactions (CE$ν$NS) require excellent energy resolution (eV-scale or below) and low backgrounds. An unknown source of phonon bursts, the low energy excess (LEE), dominates other above-threshold backgrounds and generates excess shot noise from sub-threshold bursts. In this paper, we measure these phonon bursts for 12 days after cooldown in two nearly identical 1 cm$^2$ silicon detectors that differ only in the thickness of their substrate (1 mm vs. 4 mm thick). We find that both the channel-correlated shot noise and near-threshold shared LEE relax with time since cooldown. Additionally, both the correlated shot noise and LEE rates scale linearly with substrate thickness. When combined with previous measurements of other silicon phonon detectors with different substrate geometries and mechanical support strategies, these measurements strongly suggest that the dominant source of both above and below threshold LEE is the bulk substrate. By monitoring the relation between bias power and excess phonon shot noise, we estimate that the energy scale for sub-threshold noise events is $0.68 \pm 0.38$ meV. In our final dataset, we report a world-leading energy resolution of 258.5$\pm$0.4 meV in the 1 mm thick detector. Simple calculations suggest that these silicon substrate phonon bursts are likely a significant source of quasiparticle poisoning in superconducting qubits operated in well shielded and vibration free environments.

Xingyu Lyu, Ning Wang, Yang Xiao et al.

Federated Learning is a popular paradigm that enables remote clients to jointly train a global model without sharing their raw data. However, FL has been shown to be vulnerable towards model poisoning attacks due to its distributed nature. Particularly, attackers acting as participants can upload arbitrary model updates that effectively compromise the global model of FL. While extensive research has been focusing on fighting against these attacks, we find that most of them assume data at remote clients are under iid while in practice they are inevitably non-iid. Our benchmark evaluations reveal that existing defenses generally fail to live up to their reputation when applied to various non-iid scenarios. In this paper, we propose a novel approach, GeminiGuard, that aims to address such a significant gap. We design GeminiGuard to be lightweight, versatile, and unsupervised so that it aligns well with the practical requirements of deploying such defenses. The key challenge from non-iids is that they make benign model updates look more similar to malicious ones. GeminiGuard is mainly built on two fundamental observations: (1) existing defenses based on either model-weight analysis or latent-space analysis face limitations in covering different MPAs and non-iid scenarios, and (2) model-weight and latent-space analysis are sufficiently different yet potentially complementary methods as MPA defenses. We hence incorporate a novel model-weight analysis component as well as a custom latent-space analysis component in GeminiGuard, aiming to further enhance its defense performance. We conduct extensive experiments to evaluate our defense across various settings, demonstrating its effectiveness in countering multiple types of untargeted and targeted MPAs, including adaptive ones. Our comprehensive evaluations show that GeminiGuard consistently outperforms SOTA defenses under various settings.

Muhammad Umer Khan, Iqra Khurram, Maha Munir et al.

ABSTRACT Fucoidan is a naturally occurring, sulfated fucose‐based polymer found in many types of brown algae. Fucoidan has recently attracted interest in the field of oncology due to its biological and pharmacological features, such as antitumor, anti‐proliferation, and anti‐inflammatory effects. Fucoidan's attributes are mostly determined by its relative molecular mass and the processes of purification and extraction, which result in structural modifications and can alter outcomes. Several in vivo and in vitro research are being conducted to investigate fucoidan's anticancer potential so that it can be employed as a therapeutic agent; also, decreased toxicity and in vitro impacts of fucoidan make it feasible for cancer therapy. The current review will summarize various aspects of fucoidan including the source and structural activity of fucoidan along with the anticancer relationship of fucoidan on multiple cell lines. In this review, the bioavailability and pharmacological properties are also discussed along with the mechanism of action through which fucoidan inhibits proliferation or induces apoptosis in different tumors, suggesting it is a potential therapeutic agent for combating various tumors.

Satoshi Yokota, Kiyoshi Hashimoto, Takuya Sato et al.

We previously developed the acrosin-green fluorescent protein (GFP) transgenic neonatal mouse organ culture system for rapid and accurate assessment of testicular toxicity. This system effectively evaluates drug-induced toxicity in male germ cells before meiotic entry but cannot assess post-meiotic germ cell toxicity. For many chemicals, the specific stage of germ cell differentiation that is susceptible to toxicity remains unclear, highlighting the need for new methods. In this study, we incubated neonatal mouse testis organ cultures for 35 days to allow post-meiotic cells to develop. The tissue was then exposed to cisplatin to determine the cells that are targeted and to assess the reversibility of the toxicity. We monitored changes in tissue volume and GFP fluorescence, which tracks the progression of spermatogenesis, and confirmed findings by histological analysis. Cisplatin inhibited tissue growth and reduced GFP fluorescence in a concentration-dependent manner. Higher concentrations targeted not only spermatogonia, but also spermatocytes and spermatids. Recovery from toxicity was observed at clinically relevant doses. This study demonstrates that long-term mouse testis organ culture can be used to assess testicular toxicity, enabling the identification of specific germ cell stages targeted by chemicals such as cisplatin.

Kannan Sridharan, Gowri Sivaramakrishnan

Background: Beta-lactamase inhibitors (BLIs) are widely used with beta-lactam antibiotics to combat resistant infections, yet their safety profiles, especially for newer agents, remain underexplored. This study aimed to identify potential adverse event (AE) signals associated with BLIs using the USFDA Adverse Event Reporting System (USFDA AERS). Methods: The USFDA AERS was queried for AE reports involving FDA-approved BLIs from March 2004 to March 2024. After removing duplicates, only reports with BLIs listed as primary suspects were included. Disproportionality analysis was conducted using frequentist and Bayesian approaches, with statistical significance assessed by chi-square testing. Results: A total of 12,456 unique reports were analyzed. Common AEs across BLIs included hematologic disorders, hypersensitivity reactions, emergent infections, organ dysfunction, and neurological complications. Signal detection revealed specific associations: septic shock and respiratory failure with avibactam; lymphadenopathy and congenital anomalies with clavulanic acid; antimicrobial resistance and epilepsy with relebactam; disseminated intravascular coagulation and cardiac arrest with sulbactam; and agranulocytosis and conduction abnormalities with tazobactam. For vaborbactam, no distinct AE signals were identified apart from off-label use. Mortality was significantly more frequent with avibactam and relebactam (<i>p</i> < 0.0001). Conclusions: This analysis highlights a spectrum of AE signals with BLIs, including unexpected associations warranting further investigation. While some events may reflect comorbidities or concomitant therapies, these findings underscore the importance of continued pharmacovigilance and targeted clinical studies to clarify causality and ensure the safe use of BLIs in practice.

Fei ZHAI, Naipeng LIU, Shenshen WU et al.

BackgroundThe population with diabetes in China is increasing year by year. Current research has found that either air pollution or temperature has an impact on the occurrence and development of diabetes, but the interaction between the two is unclear yet. ObjectiveTo investigate the effects and the lag effects of air pollutants and temperature on the risk of hospital admission for type 2 diabetes in Hefei, Anhui Province from 2016 to 2019, as well as to verify potential interaction between air pollutants and temperature. MethodsThis study collected hospital admission data for patients with type 2 diabetes from a tertiary hospital in Hefei, Anhui Province, and the corresponding monitoring data on air pollutants and meteorological factors from 2016 to 2019. Firstly, a distributed lag non-linear model (DLNM) was used to explore the effects of air pollutants and temperature on the risk of hospital admission for type 2 diabetes. Subsequently, a bivariate response surface model was used to explore potential interaction between temperature and various pollutants on frequency of hospital admission due to diabetes. Temperature was further divided into lower, medium, and higher levels by percentiles during the study period, and the potential interaction between air pollutants and temperature strata were verified . ResultsAfter controlling long-term trend, seasonal trend, holiday effect, and day of the week effect, the results of single pollutant models showed that for every 10 μg·m−3 increase in fine particulate matter (PM2.5), inhalable particulate matter (PM10), and nitrogen dioxide (NO2), the relative risk (RR) values for hospital admission due to type 2 diabetes were 1.032 (95%CI: 1.021, 1.043), 1.018 (95%CI: 1.008, 1.026), and 1.037 (95%CI: 1.016, 1.058), respectively; for every 1 mg·m−3 increase in carbon monoxide (CO), the RR value for hospital admission due to type 2 diabetes was 1.319 (95%CI: 1.163, 1.495); the increases in sulfur dioxide (SO2), ozone (O3), and daily average temperature showed no statistically significant impact on hospital admission due to type 2 diabetes. The results of bivariate response surface models suggested that daily average temperature and various pollutant levels spontaneously affected the risk of hospital admission for type 2 diabetes, but the stratified analysis did not find significant differences in the effect of PM2.5 on the risk of hospital admission due to type 2 diabetes across different temperature strata. ConclusionIncreases in the concentrations of PM2.5, PM10, NO2, and CO elevate the risk of hospital admission for type 2 diabetes. This study could not confirm the interactions between daily average temperature and various pollutants.

Alberto Boretti

The integrity of environmental toxicology is undermined by selective risk assessments that focus intently on certain chemicals while overlooking others. Glyphosate, one of the most widely used herbicides, serves as a case study of how regulatory decisions can be shaped by incomplete or biased evidence. This paper argues for a holistic approach to toxicology, calling for balanced assessments that consider both health risks and societal benefits. It critically examines current regulatory practices concerning glyphosate, investigating its association with non-Hodgkin’s lymphoma and its positive effects on agricultural productivity and food security. While definitive evidence linking glyphosate to cancer remains inconclusive, its role in enhancing crop yields, by as much as 20 % in some regions, has had measurable benefits for food security and public health. The paper advocates for regulatory frameworks that transparently weigh these societal benefits against potential health risks, particularly in settings of occupational exposure, where the need for balanced assessment is especially pressing. Through a narrative review of major studies, this paper underscores the need for transparency, accountability, and evidence-based approaches in environmental regulation. Such practices are essential for crafting policies that not only mitigate risk but also promote global food security and well-being. By integrating both risks and benefits into the regulatory process, the study proposes an inclusive and data-driven approach to chemical policy that aligns with the broader goals of sustainability and public health.

Bhavya Mehta, Kush Kothari, Reshmika Nambiar et al.

Traditional methods like Graph Convolutional Networks (GCNs) face challenges with limited data and class imbalance, leading to suboptimal performance in graph classification tasks during toxicity prediction of molecules as a whole. To address these issues, we harness the power of Graph Isomorphic Networks, Multi Headed Attention and Free Large-scale Adversarial Augmentation separately on Graphs for precisely capturing the structural data of molecules and their toxicological properties. Additionally, we incorporate Few-Shot Learning to improve the model's generalization with limited annotated samples. Extensive experiments on a diverse toxicology dataset demonstrate that our method achieves an impressive state-of-art AUC-ROC value of 0.816, surpassing the baseline GCN model by 11.4%. This highlights the significance of our proposed methodology and Few Shot Learning in advancing Toxic Molecular Classification, with the potential to enhance drug discovery and environmental risk assessment processes.

Enkai Li, Chuang Li, Nathan Horn et al.

The presence of deoxynivalenol (DON), one of the most frequently occurring mycotoxin, in food and feed has been considered a risk factor to both human and animal health. Molecular mechanisms that regulate DON effects in tissues are still poorly understood. However, recent evidence suggests that nuclear factor erythroid 2-like 2 (Nrf2) may be a major target during mycotoxin-induced intestinal barrier dysfunction. Although quercetin, a plant-derived flavonoid, is known to induce the activation of Nrf2 signaling pathway, its potential to mitigate effects of DON and the implication of Nrf2 in its physiological effects is poorly understood. Therefore, this study was conducted to investigate the protective effects of quercetin in alleviating the DON-induced barrier loss and intestinal injuries in IPEC-J2 cells and weaned piglets and determine the potential role of Nrf2. Quercetin treatment dose-dependently increased mRNA expression of Nrf2 target gene, NQO-1, and concomitantly increased the expression of claudin-4 at both mRNA and protein levels. Quercetin supplementation also reversed the reduction of claudin-4 caused by DON exposure in vivo and in vitro. The decreased membrane presence of claudin-4 and ZO-1 induced by DON was also blocked by quercetin. Furthermore, quercetin attenuated the endocytosis and degradation of claudin-4 caused by DON exposure. The effects of quercetin also included the restoration of transepithelial electrical resistance (TEER) and reduction of FITC-dextran permeability that have been perturbed by DON. However, the protective effects of quercetin against DON exposure were abolished by a specific Nrf2 inhibitor (brusatol), confirming the importance of Nrf2 in the regulation of TJP expression and barrier function by quercetin. In vivo study in weaned pigs showed that DON exposure impaired villus-crypt morphology as indicated by diffuse apical villus necrosis, villus atrophy and fusion. Notably, intestinal injuries caused by DON administration were partly mitigated by quercetin supplementation. Collectively, this study shows that quercetin could be used to prevent the DON-induced gut barrier dysfunction in humans and animals and the protective effects of quercetin against DON-induced intestinal barrier disruption is partly through Nrf2-dependent signaling pathway.

Louise Ramhøj, Marta Axelstad, Yoni Baert et al.

Current test strategies to identify thyroid hormone (TH) system disruptors are inadequate for conducting robust chemical risk assessment required for regulation. The tests rely heavily on histopathological changes in rodent thyroid glands or measuring changes in systemic TH levels, but they lack specific new approach methodologies (NAMs) that can adequately detect TH-mediated effects. Such alternative test methods are needed to infer a causal relationship between molecular initiating events and adverse outcomes such as perturbed brain development. Although some NAMs that are relevant for TH system disruption are available–and are currently in the process of regulatory validation–there is still a need to develop more extensive alternative test batteries to cover the range of potential key events along the causal pathway between initial chemical disruption and adverse outcomes in humans. This project, funded under the Partnership for the Assessment of Risk from Chemicals (PARC) initiative, aims to facilitate the development of NAMs that are specific for TH system disruption by characterizing in vivo mechanisms of action that can be targeted by in embryo/in vitro/in silico/in chemico testing strategies. We will develop and improve human-relevant in vitro test systems to capture effects on important areas of the TH system. Furthermore, we will elaborate on important species differences in TH system disruption by incorporating non-mammalian vertebrate test species alongside classical laboratory rat species and human-derived in vitro assays.

Max Menssen

A prediction interval is a statistical interval that should encompass one (or more) future observation(s) with a given coverage probability and is usually computed based on historical control data. The application of prediction intervals is discussed in many fields of research, such as toxicology, pre-clinical statistics, engineering, assay validation or for the assessment of replication studies. Anyhow, the prediction intervals implemented in predint descent from previous work that was done in the context of toxicology and pre-clinical applications. Hence the implemented methodology reflects the data structures that are common in these fields of research. In toxicology the historical data is often comprised of dichotomous or counted endpoints. Hence it seems natural to model these kind of data based on the binomial or the Poisson distribution. Anyhow, the historical control data is usually comprised of several studies. These clustering gives rise to possible overdispersion which has to be reflected for interval calculation. In pre-clinical statistics, the endpoints are often assumed to be normal distributed, but usually are not independent from each other due to the experimental design (cross-classified and/or hierarchical structures). These dependencies can be modeled based on linear random effects models. Hence, predint provides functions for the calculation of prediction intervals and one-sided bounds for overdispersed binomial data, for overdispersed Poisson data and for data that is modeled by linear random effects models.