Continuing Education

Chair(s):
Zhoumeng Lin, University of Florida; and Wei-Chun Chou, University of California.

Machine-Learning and artificial intelligence approaches have substantially advanced many scientific disciplines, including toxicology and risk assessment. Integration of machine-learning algorithms with traditional toxicological models, such as physiologically based pharmacokinetic (PBPK) and quantitative structure-activity relationship (QSAR) models can generate more robust models that can be used to predict absorption, distribution, metabolism, excretion, and toxicity (ADMET) of hundreds or thousands of chemicals or nanoparticles in a highly efficient manner. Many toxicologists from academia, industry, and governmental agencies have leveraged machine-learning approaches to different applications, including development of new computational toxicology models and drug discovery and development, as well as regulatory approvals of drugs and medical devices. The importance of machine-learning approaches is also highlighted by the fact that there have been multiple Symposium and Workshop Sessions on machine learning in toxicology during the last few SOT Annual Meetings. However, the challenge of properly using machine-learning approaches to support toxicological studies lies in the need of multidisciplinary expertise, including computer science, programming, machine-learning algorithms, toxicology, mechanisms of toxicokinetics and toxicity, PBPK modeling, QSAR modeling, and risk assessment. Collaboration between computational scientists and mechanistic toxicologists are common and important to complete machine learning–based projects. However, it is not easy for computational scientists to understand ADMET mechanisms of xenobiotics, and likewise, it is not a simple task for mechanistic toxicologists to understand and implement machine-learning models.

To address this barrier and to facilitate applications of existing machine-learning approaches to further support toxicological studies, the objective of this Continuing Education (CE) minicourse is to introduce the fundamental concepts of machine-learning approaches and how these approaches can be applied to support toxicological studies. The first talk will give an overview on the basics of commonly used machine-learning approaches in the field of toxicology, including k-nearest neighbors, support vector machine, decision tree, random forest, artificial neural network, and deep learning. The first talk will also give an example of applying machine learning to predict a key pharmacokinetic property (i.e., the plasma half-life) of drugs and another example of building a machine learning–assisted PBPK model for nanoparticles. The second talk will introduce how to build machine learning–based QSAR models to predict carcinogenic and noncarcinogenic toxicities of xenobiotics. The second talk will also do a software demonstration to show what a machine learning–based QSAR model looks like and how to use it to perform predictions of ADMET properties of chemicals. This demonstration will be through Google Colab, so no software installation is required.

This CE course will be focused on examples that are relevant to toxicology. It is expected that this course will be beneficial for a diverse population of SOT attendees, including both experimental toxicologists and computational toxicologists at different career stages (e.g., students, postdocs, and senior scientists) across various sectors, including academia, industry, and regulatory agencies.

Chair(s):
Shannon Bell, RTI International; and Oswaldo Lozoya, RTI International.

Data management and sharing are critical components that ensure transparency, reproducibility, and broader scientific impact in biomedical and toxicological research. How to collect, report, and share data are critical components to the research process but can be challenging when considering computational approaches that take in multiple data streams to generate a diversity of results.

This course is designed for individuals who are utilizing computational and predictive modeling in their research projects to review some best practices to ensure their results can have maximal impact within the community as well as cover how to meet the US National Institutes of Health (NIH) data management and sharing (DMS) requirements in the context of predictive modeling. The course will delve into strategies for handling metadata and data standards, navigating how to share your data, leveraging secondary data for analysis, and identifying reporting standards, with a specific focus on predictive modeling. By the end of this course, participants will be equipped with the knowledge and tools necessary to effectively manage and share data generated from computational and predictive modeling activities in alignment with NIH DMS requirements, fostering greater collaboration and impact within the scientific community. We will work through examples together, and participants are encouraged to bring laptops to access materials during the workshop. Resources will be made available as part of the course materials to reserve time for working through examples.

Learning Objectives:

• Understand the role of data sharing in advancing toxicology research Define research benefits to data sharing, the key components of the NIH DMS policies, and implications of data sharing for researchers using computational and predictive modeling approaches
• Implement metadata and data standards in modeling workflows Learn how to structure and document data using appropriate metadata and data standards to facilitate interoperability and reuse
• Navigate data repositories and utilize secondary data Explore relevant data repositories suitable for biomedical data and understand the ethical and practical considerations of using secondary data in computational modeling
• Adhere to reporting standards for predictive models Identify and apply best practices for reporting predictive models, including data quality assurance, model validation, and transparent documentation

Chair(s):
Jessica Murray, US EPA; and Monita Sharma, PETA Science Consortium International e.V.

In vitro methods are increasingly being deployed to assess the toxicity of inhaled substances, but it can be challenging to ensure their relevance to human exposure scenarios. The use of in vitro data to predict human health effects due to inhalation requires the consideration of several parameters including physicochemical properties of the inhaled substance, target exposure scenarios, anatomy of the respiratory system, and understanding of deposition and uptake mechanisms in both the respiratory tract and in vitro exposure systems. Inhaled substances are incredibly diverse and include aerosols and various gases (e.g., reactive, volatile organic compounds [VOCs], and “Category 2”), all of which possess distinct physicochemical properties that must be considered when selecting dosimetry models to predict human exposure and when optimizing in vitro assay conditions.

This Continuing Education course is intended for those who conduct or interpret data from in vitro inhalation studies and new approach methods (NAMs). This course will introduce how various dosimetry models are applied in risk assessment workflows and how in silico approaches can be utilized to predict human exposures and improve in vitro study design. Attendees will learn the fundamentals of the US EPA multiple-path particle dosimetry (MPPD) model, computational fluid dynamics (CFD), physiologically based pharmacokinetic (PBPK) models, hybridized modeling approaches (e.g., CFD-PBPK), and the in vitro sedimentation, diffusion, dissolution, and dosimetry model.

This course will begin with a risk assessor’s perspective that includes an overview of inhalation toxicology and dosimetry, highlighting how physicochemical properties interact with features of respiratory anatomy and physiology to determine deposition and uptake. We will cover how data availability coupled with these considerations dictate dosimetry model selection and requirements for their use. Following this critical background information, a practical demonstration of the US EPA MPPD v.2.0 model will be included. Subsequent presentations will share technical features and best applications for CFD, PBPK, and hybridized CFD-PBPK models, along with case studies for a broad array of inhalable compounds including soluble and insoluble aerosols, reactive gases, and vapors. Finally, the last two speakers will focus on dosimetry modeling approaches and quantitative workflows for in vitro test systems and will share best methods to translate in vitro doses to human exposure scenarios. Attendees will learn how to incorporate these methods when designing in vitro studies and how to improve in vitro to in vivo extrapolation (IVIVE) for inhalation.

The course will be of interest to those engaged in the development of in vitro and alternative test methods, inhalation toxicology, risk assessment, regulatory and safety evaluation, and dosimetry modeling.

Chair(s):
Emily Reinke, Inotiv; and Donna Macmillan, International Collaboration on Cosmetics Safety.

The global push toward the 3Rs and increased restrictions on animal methods requires the development of nonanimal methods that can fulfill regulatory requirements for specific toxicity endpoints. As methods have been developed and processed through global harmonization pipelines, like the Organisation for Economic Co-operation and Development (OECD) Test Guidelines Programme, multiple lessons have been learned, and the processes by which methods are developed, validated, and implemented have been refined.

In this course, attendees will learn the key factors that must be considered in developing methods, concepts of method readiness, validation and peer review, and the processes that are available within OECD to bring methods to a stage where they can be implemented by industry and the regulatory community. After a short introduction, the first talk will highlight the major aspects to determine if a developed method is ready for the next step, prevalidation and peer review with case examples of developed tests for endocrine disruptors identification. The next speaker will provide insight into the process of validation, the key concepts, and how validation is changing to become more efficient and robust, with reference to the current efforts to update OECD Guidance Document 34. The scientific confidence, context of use, and efforts to standardize new approach methodologies (NAMs) to address complex endpoints can be evaluated within the OECD project on integrated approaches for testing and assessment (IATA), a way to evaluate methods in a chemical-agnostic approach, as will be covered by the third speaker. The fourth speaker will address mechanisms by which methods, including IATAs, can be evaluated under the OECD test guideline program and developed into a defined approach, which relies on a strict data interpretation procedure for determination of the endpoint outcome. Finally, the last speaker will give the perspective of industry on implementation of accepted NAMs and the impact on testing requirements.

Upon completion of the course, attendees will have a better understanding of how to bring a method from development to validation, build confidence, and standardize aspects of the approach on the path to harmonization.

Chair(s):
Madhavi Billam-Mukhi, Meta Platforms, Inc.; and Janet Zang, US FDA.

Building on the success of an Informational Session titled “From My Cosmetics to Smart Watch: Toxicology Touches It All” held during the 2024 SOT Annual Meeting and ToxExpo, this Continuing Education (CE) course expands from the basic concepts offered at the Informational Session to an in-depth and hands-on course to evaluate the safety of diverse categories of products, ranging from medical devices to cosmetics and consumer products, such as virtual reality headsets.

In an era with novel products constantly being developed, toxicologists are often challenged with the absence of standardized assessment approaches when assessing the safety of emerging day-to-day consumer products with skin contact, such as virtual reality headsets and other wearable digital devices. This broadreaching course aims to provide extensive knowledge along with relevant case studies, risk assessment approaches, and best practices drawn from established sectors like medical devices and cosmetics that can be adapted to evaluate new product categories.

The session will start with an overview of the diverse spectrum of consumer products and considerations for their safety evaluation. The second presentation will focus on skin sensitization, a common adverse effect following dermal contact. The speaker will elucidate the toxicological mechanisms of chemical-induced sensitization, along with examples of diverse types of dermal sensitizers across various product categories. Next, the session will delve into three consecutive talks that provide in-depth knowledge and practical safety assessment approaches in medical devices, cosmetic products, and wearables. These talks will include examples, case studies, or exercise problems to guide the attendees through highlighted principles and methods. The session will end with a final presentation on the US FDA adverse event reporting system for cosmetic products, which underlines the importance of postmarket surveillance as a critical element that complements premarket safety evaluation and risk assessment to protect consumer health.

This CE course’s main objective is to provide a comprehensive overview of toxicological review considerations across diverse product categories. From material selection to adverse event reporting, the session offers a holistic understanding of consumer product safety assessment. Importantly, this session represents a strong collaboration across academia, industry, expert consultants, trade associations, and government, marking the first time such diverse expertise has been brought together for training on consumer product safety evaluation at the SOT Annual Meeting and ToxExpo. It underscores the practical application of nonanimal approaches and highlights the broader role of toxicology beyond traditional sectors like pharmaceuticals and agrochemicals. Notably, this CE course explores uncharted territory, addressing areas where traditional textbooks lack coverage. It serves as a valuable resource for both budding toxicologists and experienced risk assessors who plan to expand their skills to diverse sectors, including less mainstream product categories where toxicology is pivotal.

Chair(s):
Esther Haugabrooks, Coca-Cola Company; and Matthew Linman, Trinity Consultants.

If we are what we eat, then ensuring we are safe is essential. Food toxicologists in academia, industry, and government face unique challenges and drivers that influence the evaluation or reevaluation of raw materials, ingredients, or articles associated with food or beverages. The category of food is not only broad but also dynamic, with thousands of food matrices of varying compositions generating critical interactions between consumer demands, idiosyncrasies across susceptible populations, and sometimes varying regional regulatory requirements. How do safety professionals safeguard the food supply and address the diverse risk considerations across different food types?

This course, geared toward toxicologists in the food industry, aims to enhance expertise in evaluating toxicology studies while honing critical-thinking skills in assessing the weight of evidence and study design. It is also for those toxicologists curious about how toxicology is applied to food safety or who want to learn from another subdiscipline in the field. Presentations will include establishing food safety for ingredients, developing safety narratives that align with regulatory requirements, and optimizing regulatory approvals within food additives, color additives, and GRAS substances. The course will culminate in an interactive case study where participants apply the concepts covered to test if we can find consensus in food toxicology evaluations.

Chair(s):
Anna Lowit, US EPA; and Gina Hilton, PETA Science Consortium International e.V.

Risk assessment is a critical tool used worldwide in the safety evaluation of agrochemicals, environmental chemicals, pharmaceuticals, cosmetics, and metabolites/degradates/leachates along with other materials and sectors.

This course aims to delve into advancements in risk assessment and risk characterization to highlight the current state-of-the-science regarding the paradigms, methodologies, and technologies available for enhancing the safety evaluation of chemicals. This course departs from conventional risk assessment training, which typically offers cookie-cutter descriptions on topics such as hazard characterization and identification, dose-response assessment, and default uncertainty factors. Rather, this course will introduce the use of problem formulation to leverage innovative, fit-for-purpose approaches for chemical risk assessment in the 21st century. Specifically, it will introduce advances in applying new approach methods (NAMs) to derive regulatory guidance values; developing computational tools to integrate mechanistic data in chemistry and biology to better predict exposures, hazards, and dose-response relationships; using novel modeling and monitoring approaches for exposure assessment; drawing weight of evidence–based conclusions about potential health risks; and evaluating context-specific uncertainty in lieu of defaults. This training will also provide examples of narrative-based descriptions of risk characterizations essential for effective engagement with stakeholders and decision-makers.

The first presentation will cover the fundamental principles of risk assessment with a contemporary twist, including discussions on the continuum of various types of risk assessments that support different risk management purposes, tiered approaches for addressing risk assessment questions, the importance of problem formulation, and types of toxicology used in hazard assessment (including traditional animal studies, human studies, and NAMs), along with hazard assessment and hazard characterization. The second presentation will provide an in-depth background for exposure assessment and risk characterization, emphasizing advancements in these areas such as computational modeling, cell-based assays, and nonmammalian systems. Following the two introductory presentations, this training will take a deep dive into case studies representing the risk assessment spectrum with respect to data availability, allotted time and resources, decision context, and associated uncertainty. The third presentation will feature a case study based on new chemicals exposure and risk assessment (e.g., like those performed under the US Toxic Substances Control Act) where there are no data requirements and the statute requires risk assessments and decision-making within 90 days. This presentation will also highlight the use of a variety of NAMs and rapid exposure assessment. The final presentation will showcase case studies demonstrating weight of evidence–based risk assessment for agrochemicals (e.g., like those conducted for registration review under the US Federal Insecticide, Fungicide, and Rodenticide Act), including the use of NAMs. Presentations three and four will include example narrative descriptions appropriate for use in risk characterization to support risk communication. The weight of evidence approach involves highly refined exposure assessment and extensive hazard data obtained from both traditional in vivo animal studies and NAMs. Where applicable, computational tools are used to integrate mechanistic data from multiple sources for the retrospective interpretation or extrapolation of dose-response data, as well as the prospective design of studies to address data and knowledge gaps. At the end of presentations three and four, a hands-on exercise will provide attendees with the opportunity to work as a team, using mock data provided to them to make risk assessment decisions and offer rationales to justify their decisions. The session will close with time for general questions from attendees.

Chair(s):
Stephen Edwards, US EPA; and Clemens Wittwehr, European Commission Joint Research Centre.

Environmental science research is becoming increasingly more integrated to address complex, 21st-century problems. This, coupled with an exponential increase in research outputs, makes it impossible for an individual researcher to comprehensively review the existing knowledge relevant for their research. Computational models can support researchers by integrating data, but they require structured inputs based on our existing knowledge of toxicological mechanisms and exposure pathways. Knowledge bases address this challenge by aggregating existing knowledge from a given research area and providing it in a more generally usable form. This increases the user base for that existing knowledge and prevents a duplication of effort by many research teams assembling the same information independently. The ultimate goal for environmental knowledge bases is to provide knowledge in a computable form that can support models used to improve human health and the environment. Current trends in funding and emphasis on scientific impact have shifted existing and new knowledge-base efforts toward more applied uses and should accelerate the pace toward reaching this goal.

This course will include a series of presentations that demonstrate the rich history of environmentally focused knowledge bases as well as new and emerging resources. The first knowledge base for discussion will be the ChemExpo knowledge base, which was first released in 2023. This resource integrates exposure data from a variety of sources and makes it readily available to support risk assessment applications. This represents an example of an early stage knowledge base and highlights the value of structured exposure information to support regulatory decisions. The next talk will connect exposure information with biological effects via the Comparative Toxicogenomics Database (CTD). Over the last two decades, CTD has expanded to include not only toxicogenomics data but also associations among chemicals, exposure, phenotypes, and adverse outcomes. More recent tools have focused on integrating data across these categories to enable scientists to design testable hypotheses regarding mechanisms of toxicity, including adverse outcome pathways. This will be followed by a similar survey of the Adverse Outcome Pathway Wiki that was first released in 2013 with a goal to support the use of new alternative methods for environmental decision-making. This will include examples showing direct impact on regulatory decisions and the development of computational models and upcoming changes to increase the utility of this resource in the future. The next talk will highlight the use of knowledge bases to enhance decision-making within the agrochemical sector. Finally, a knowledge base that is emerging from the European Partnership for the Assessment of Risks from Chemicals funded through the Horizon Europe Programme will highlight a new theme where knowledge bases are developed as part of these large transdisciplinary projects to increase the impact of their collective research. This will provide a glimpse into the earliest stages of knowledge-base development with an emphasis on community engagement and defining the purpose and features of the knowledge base. The learning objectives for this course are to (1) enable participants to navigate three public knowledge bases covering the full source to outcome continuum, (2) show participants how they can develop custom knowledge bases to support their research, (3) help participants understand the state-of-the-science regarding environmental knowledge bases and what to expect in the coming years, and (4) demonstrate the value of environmental knowledge bases and the importance of community participation in expanding these resources.