Events

Opportunities to Harness AI and Machine Learning in Drug Discovery Toxicology

Hosted by: The SOT Regulatory and Safety Evaluation and Computational Toxicology Specialty Sections

Registration required for this free webinar.

Register Now

Speaker(s):
Yodi Melnikov, PhD, Principal Data Scientist, Computational Toxicology, Genentech

Description:
Yodi Melnikov, PhD, is a principal data scientist in the safety assessment group at Genentech where he has focused on developing artificial intelligence and machine learning (AI/ML) models for early hazard identification in small molecule development. More recently, Yodi has been responsible for developing and optimizing bioinformatics pipelines for cross-species and cross-platform data inference in safety assessment, as well as data related data integrity and robust data inference work.

In this webinar, Dr. Melnikov will describe the various opportunities to leverage AI/ML in the drug discovery and development process to ultimately support decision making, improve efficiencies, and potentially reduce animal use. Case studies on modeling and practical application of toxicological endpoints ranging from simple single endpoint prioritization models like hERG, to more complex models combining data from multiple sources such as secondary pharmacology and drug induced liver injury, will be presented.

Registration required for this free webinar.

Register Now

Computational Toxicology Specialty Section 2024 Award Winner Presentations

Hosted by: The SOT Computational Toxicology Specialty Section

Registration required for this free webinar.

Register Now

The 2024 winners of the Computational Toxicology Specialty Section (CTSS) awards will present their work in this webinar, which will conclude with a question-and-answer segment.

Presentations:

• “Identifying Gene Predictors of Chemicals Linked with Breast Cancer: A Machine Learning Analysis of MCF7 Cellular Transcriptomic Screening Data”
  Lauren Koval, PhD Candidate, The University of North Carolina at Chapel Hill, Yves Alarie Diversity Award
• “A Cheminformatics Workflow for Higher-Throughput Modeling of Chemical Exposures from Biosolids”
  Paul Kruse, Postdoctoral Fellow, US EPA, Postdoctoral Award
• “Predicting Chemical Immunotoxicity Through Data-Driven QSAR Modeling of AhR Agonism and Related Toxicity Mechanisms”
  Nada Daood, PhD Candidate, Rowan University, Elsevier Student Award

Registration required for this free webinar.

Register Now

Advancing Toxicology in Drug Discovery using Generative Adversarial Networks

Hosted by: The SOT Drug Discovery Toxicology and Computational Toxicology Specialty Sections

Speaker(s):
Weida Tong, PhD, Director, Division of Bioinformatics and Biostatics, US FDA National Center for Toxicological Research (NCTR)
Zhichao Liu, PhD, Head of Computational Toxicology, Boehringer Ingelheim

Description:
Toxicological research is undergoing a transformative shift with the incorporation of advanced computational models. Among these, Generative Adversarial Networks (GANs) are emerging as a pivotal tool in addressing challenges associated with data scarcity, predictability, and the nuanced complexity of toxic responses. This webinar delves into the integration of GANs into toxicology, from their foundational principles to applications in drug discovery. Attendees will gain insights into the capabilities and potential of GANs, fostering a better understanding of their role in modern toxicological studies. This webinar will cater to both novices and experts in the intersection of toxicology and generative artificial intelligence.

Webinar Recording

Dr. Weida Tong Presentation | Dr. Zhichao Liu Presentation

CTSS—Computational Toxicology Top 3 Abstracts of 2023

Hosted by: The SOT Computational Toxicology Specialty Section

The 2023 Top 3 abstract awardees of the Computational Toxicology Specialty Section (CTSS) presented their work in this webinar, which will conclude with a question & answer segment.

Presentations:

• “Data Fusion by Matrix Completion for Exposome Target Interaction Prediction”
  (Abstract No. 3077) Kai Wang, PhD, Postdoctoral Fellow, University of Michigan, Ann Arbor
• “Framework for In Sillico Toxicity Screening of Novel Odorants”
  (Abstract No. 3698)
  Isaac Mohar, PhD, Principal Toxicologist, Gradient
• “Computational Analysis of Discontinued Neurological Drugs without Defined Primary Target Pharmacology”
  (Abstract No. 3059) Mohan Rao, PhD, Toxicologist, Neurocrine Biosciences

Webinar Recording
Dr. Kai Wang Presentation | Dr. Isaac Mohar Presentation | Dr. Mohan Rao Presentation

ToxAIcology—The Future of Toxicology is AI

Speaker(s):
Thomas Hartung, MD, PhD, Professor & Chair, Johns Hopkins Bloomberg School of Public Health

Description:
The 2007 NAS report on Toxicity Testing for the 21st Century was a watershed moment for toxicology. Since then, the discussion is no longer whether to change but how and how fast? With knowledge in the life sciences doubling every 3-4 years, we now have sixteen times more understanding and a number of disruptive technologies have evolved which were not anticipated in the report, such as Microphysiological Systems (MPS) and Machine Learning, also known as Artificial Intelligence (AI). To embrace these developments and move toxicology to a more wholistic and integrated paradigm, the Basic Research Office of the Office of the Under Secretary of Defense for Research and Engineering (OUSD (R&E)) hosted a Future Directions workshop Advancing the Next Scientific Revolution in Toxicology in April 2022 at the Basic Research Innovation Collaboration Center (BRICC). Scientific, technical experts, and agency observers developed a report, laying out how recent developments can be embraced and set the direction of “Toxicology for the 21st Century 2.0” in the next decades.

Computational approaches, especially AI, play a key role here:

1. A central role of Exposomics to change to more exposure-driven toxicology, with AI enabling us to make sense of “omics” (big) data
2. Predictive toxicology through automated read-across such as read-across-based structure-activity relationships (RASAR)
3. The computational modeling of in vitro tests and MPS
4. Digital pathology through image analysis
5. Information extraction by Natural Language Processing of scientific literature and the grey information of the internet as well as curated databases of legacy data
6. Evidence integration of different evidence streams allows probabilistic risk assessment

The EU project ONTOX is working toward the implementation of some of these goals.

Webinar Recording

Dr. Thomas Hartung Presentation

Hosted by: SOT Arab Toxicologists Association

Speaker(s):
Dr. Ahmed Abdelmoneim, BVMS, MSc, PhD, Louisiana State University
Dr. Tamer Mansour, MBChB, MS, PhD, University of California Davis

Description:
Next-generation sequencing (NGS) and bioinformatic technologies enabled new directions to address research questions that could not be previously considered due to cost/time prohibiting factors and the lack of molecular information. In the field of toxicology, NGS and bioinformatics helped us gain a great deal of insight on chemical-induced proteome-, genome- and transcriptome-wide effects and discover novel biomarkers of toxicant exposure and effects. However, to exploit the full potential of these technologies in toxicological studies, researchers need to learn more about their applications and limitations.

In this webinar, Dr. Mansour and Dr. Abdelmoneim will provide an overview on the different branches of bioinformatics and their applications in toxicological research. They will discuss how to design your first bioinformatic study and points to consider. They will also present a detailed description of how bioinformatics tools were used to analyze datasets associated with a toxicological investigation.

Webinar Recording

Presentation

Artificial Intelligence Enables Structural Toxicity Testing for Endpoint and Multiple-Timepoint Assays

Speaker(s):
Alexandre Ribeiro, PhD, Senior Scientist, Hovione

Description:
Drug toxicity leads to the attrition of more than one-third of drug candidates and presents a high financial burden in drug development. To reduce this cost, there is a need for more predictive approaches involving technologies that better assess drug toxicity in the early preclinical stages of drug development. Cells that represent the function of specific tissues can be used for predicting toxic drugs effects. To evaluate cellular drug-induced structural toxicity, intracellular structures are labeled and imaged using high-content microscopy to detect intracellular damage with image analysis techniques. Such approaches are more accurate than human inspection of images but may miss subtle structural changes that are not easily visualized or are too complex to measure with traditional image analysis. Moreover, they are rarely amenable to live-cell bright-field imaging.

In this event sponsored by two Society of Toxicology (SOT) component groups, the American Association of Chinese in Toxicology (AACT) and the Computational Specialty Section (CTSS), we present a novel image-based artificial intelligence (AI) tool for quantifying subtle structural changes in cell-based models. The inputs are a collection of cellular images captured at multiple doses for the drugs of interest and a control set of images with only the vehicle applied. The output is a metric of structural change for each drug dose relative to the control. The technology introduces a new approach for structural toxicity testing that is unbiased and provides a high level of sensitivity that has not previously been possible. Through the process of training deep neural networks, the system learns on its own what features within the images, if any, are contributing to the structural changes. The proposed method is agnostic to imaging modality and works for fixed and stained cells (fluorescence images) as well as live cells (bright-field images), enabling drug testing at multiple timepoints.

The data presented are from two proof-of-concept experiments using iPSC-derived hepatocytes: 1) Endpoint Assay: Cells were cultured on 96-well plates. Multiple doses of tamoxifen and aspirin were applied for 48 hours, after which cells were fixed and stained with phalloidin, and fluorescence microscopy images were captured. For tamoxifen-treated hepatocytes, the method detected dose-dependent structural changes, and showed higher sensitivity than a cytochrome P450 3A4 assay. No changes were detected for aspirin. 2) Multiple Timepoint Assay: Cells were cultured on 96-well plates. Multiple doses of tamoxifen were applied, after which live-cell bright-field images were captured every 3 hours for 2 days. The method detected dose-dependent morphological changes caused by tamoxifen over time and showed better sensitivity than a caspase-3/7 apoptosis assay designed for screening toxicity in live cells. Results from the endpoint assay based on phalloidin imaging correlated highly with results from our multiple time point assay based on brightfield imaging. The approach can enable chronic studies in the same culture plate.

This one-hour webinar will include a presentation by Dr. Ribeiro and a time set aside for questions.

Webinar Recording

Dr. Alexandre Ribeiro Presentation

2022 Computational Toxicology Specialty Section Award Winner Presentations

The 2022 awardees of the Computational Toxicology Specialty Section (CTSS) will present their work in this webinar, which will conclude with a question and answer segment.

Presenters:

CTSS Paper of the Year Award

Heather Ciallella (Rutgers University)

"Revealing Adverse Outcome Pathways from Public High-Throughput Screening Data to Evaluate New Toxicants by a Knowledge-Based Deep Neural Network Approach.” Environ. Sci. Technol.2021, 55, 10875−10887

CTSS Student Travel Award

Xuelian Jia (Rutgers Center for Computational and Integrative Biology)

“Mechanism-Driven Modeling of Drug-Induced Liver Injury Using Structural Alerts and an Oxidative Stress Screening Assay”

Elsevier/CTSS Postdoctoral Award

Dr. Adrian Green (North Carolina State University)

“Leveraging High-Throughput Screening Data, Deep Neural Networks, and Conditional Generative Adversarial Networks to Advance Predictive Toxicology”

2022 Winner of the Yves Alarie Diversity Award

Linlin Zhao, PhD (work done at Genentech, Inc., currently at Amazon)

“Application of a Matrix Factorization Method for Kinase Data to Support Safety Profiling”

Moderators:

Catrin Hasselgren, Genentech, CTSS Past President

Patricia Ruiz, ATSDR/CDC

Webinar Recording
Award Presentation

Integrating Mass Spectrometry Non-Targeted Analysis and Computational Toxicology to Characterize Chemical

Speakers:
Antony Williams, US EPA, Center for Computational Toxicology and Exposure
Ron Brown, Risk Science Consortium

Materials:
Webinar Recording

Williams Presentation

2022 Computational Toxicology Specialty Section Webinar: The Predictive Toxicogenomics Space (PTGS) Modeling Tool Captures Diverse Cellular and Organ Toxicity Mechanisms and Serves for In Vitro Model-Driven Prediction of Drug-Induced Liver Injury

Abstract:
“Toxicogenomics” represents a steadily developing “Big Data” informatics analysis field. Increasing amounts of safety testing-derived gene expression data require interpretation related to existing knowledge for characterizing hazard and risks coupled to agents such as drugs, chemicals and nanomaterials. We have elucidated toxicity mechanisms from embracing the network character of systems biology as well as the complementary linear analysis scheme characteristic of the adverse outcome pathway (AOP) concept. An “artificial intelligence”-derived 14 gene component-based “predictive toxicogenomics space (PTGS)” tool generates toxicity estimates intrinsic to omics-data via broad coverage of toxicity reactions and mechanisms. The tool enables application of in vitro data to assess tissue injury in multiple organs of experimental animals subjected to repeated-dose toxicity bioassays, including the accurate prediction of human drug-induced liver injury.

Speaker(s):
Antony Williams, US EPA, Center for Computational Toxicology and Exposure
Ron Brown, Risk Science Consortium

Webinar Recording

Roland Grafström and Pekka Kohonen Presentation

Integrating Computational Tools into Carcinogenicity Assessments—Part II

Description:
This is the second session of a two-part webinar series exploring computational tools and predictive models for assessing carcinogenicity. The goal of the series is to present and discuss opportunities to improve on current methods and increasing their implementation in safety evaluations of drugs and potentially hazardous chemicals. This session will discuss the use of in silico methods, including artificial intelligence and machine learning as part of a weight of evidence in this process. It will also cover fundamental concepts of carcinogenesis and how in silico tools can be developed and used to better understand mechanistic aspects. The webinar will be of broad interest to SOT members, especially graduate students, and postdoctoral fellows who may not directly work in the field of carcinogenesis or computational toxicology but wish to expand their knowledgebase.

The webinar will consist of two 20-minute presentations followed by a 20-minute moderated question/answer session with the speakers.

The Use of VEGA and JANUS Software to Evaluate Carcinogenicity of Botanicals

Speaker(s):
Dr. Emilio Benfenati, PhD, Istituto Mario Negri, Milan, Italy

Abstract:
We are exposed to thousands of botanicals through the diet and other products, such as cosmetics. Existing in silico models can help, but are not sufficient to explore these substances. The combination of read-across and in silico models improves the possibility to get a reasonable evaluation for a number of botanicals. For this purpose, we used VEGA, JANUS, and ad hoc software to cluster similar compounds, and screen analogs.

In Silico Approaches in Carcinogenicity Hazard Assessment: Current Status and Future Needs

Speaker(s):
Dr. Raymond Tice, PhD, RTice Consulting, Hillsborough, NC, USA

Abstract:
As part of an international effort to develop in silico toxicological protocols, a consortium of scientists across several industries and governmental agencies evaluated the extent to which in silico models exist for each of the 10 key characteristics (KCs) of carcinogens and indicated where experimental methods need to be implemented and robust databases generated to enable the development of reliable in silico models. This effort also highlighted the likely interactions among the KCs with the different stages of carcinogenesis.

Webinar Recording

Dr. Emilio Benfenati Presentation | Dr. Raymond Tice Presentation

Integrating Computational Tools into Carcinogenicity Assessments

Description:
The webinar will explore computational tools and predictive models for assessing carcinogenicity as well as address opportunities to improve on current methods and increase their implementation in safety evaluations for drugs and potentially hazardous chemicals.

This session will introduce foundational aspects of computational toxicology and detail how computational methods can be used to improve cancer safety assessments. The primary concepts covered in this webinar will include 1) an overview of in silico methods, including artificial intelligence and machine learning, and 2) potential application of in silico methods in regulatory science. The webinar will be of broad interest to SOT members, especially graduate students, and postdoctoral fellows who may not directly work in the field of carcinogenesis or computational toxicology but wish to expand their knowledgebase.

The webinar will consist of two 20-minute presentations followed by a 20-minute moderated question/answer session with the speakers.

In Silico Methods for the Prediction of Carcinogenicity

Speaker(s):
Dr. Mark Cronin, PhD, School of Pharmacy and Biomolecular Science, Liverpool John Moores University, Liverpool, UK

Abstract:
An overview of in silico methods to predict carcinogenicity will be provided. The complexity of this endpoint means it is one of the most challenging adverse outcomes to predict, although much progress has been made, especially for DNA reactive carcinogens. The methods currently available and routinely used include read-across, structural alerts and quantitative structure-activity relationships (QSARs). Potential future applications of machine learning and artificial intelligence will be considered.

The EURL ECVAM Curated Database of Genotoxicity & Carcinogenicity Results: Opportunities for Application in Regulatory Science.

Speaker(s):
Dr. Federica Madia, PhD, European Commission, Joint Research Centre (JRC) Ispra, Italy

Abstract:
Collection of data, consolidation of databases, criteria for the selection and expert review of data results are all prerequisites for the design and development of new testing tools. The EURL ECVAM Consolidated Genotoxicity and Carcinogenicity Database as a reference for a number of scientific endeavors in the area of genotoxicity and carcinogenicity testing across different product-type sectors is presented as well the potential application within regulatory science.

Webinar Recording

Dr. Mark Cronin Presentation | Dr. Federica Madia Presentation

Moving from One-Size-Fits-All to Fit-for Purpose TTC Values

Speaker(s):
Ron Brown, PhD, Risk Sciences Consortium
Grace Patlewicz, PhD, US EPA

Description:
Threshold of Toxicological Concern (TTC) values have historically been derived from comprehensive datasets that include structurally diverse compounds that span a wide range of toxicological or carcinogenic potency. The goal of this "one-size-fits-all" approach is to derive a TTC value that is adequately protective for any data-poor compound that a person may be exposed to, with the exception of very highly potent Cohort of Concern compounds. However, there is increasing interest in developing fit-for-purpose TTC values for compounds in specific use categories (e.g., cosmetics, medical devices, food, consumer) or structural classes. This webinar will explore recent efforts to develop fit-for-purpose TTC values and describe the factors that should be considered (e.g., chemical space) when deriving TTC values for compounds in specific use or structural categories.

Webinar Recording

Webinar Intro Slides | Webinar Presentation

Application of In Vitro and In Silico Data in Predictive Modeling of Human Organ Toxicity

Speaker(s):
Ruili Huang, PhD, National Center for Advancing Translational Sciences (NCATS), National Institutes of Health (NIH)

Description:
Traditional toxicity testing reliant on animal models is costly and low throughput, posing a significant challenge with increasing numbers of environmental chemicals that need toxicological evaluation. This presentation will describe machine learning models for the prediction of various human organ-level toxicity endpoints, e.g., carcinogenicity, cardio, developmental, hepato, nephro, neuro, reproductive, and skin toxicities, using chemical structure and Tox21 in vitro quantitative high-throughput screening (qHTS) assay data. The assay targets contributing the most to model performance could be used to provide insight into the mechanisms and pathways related to different toxicity endpoints. A systematic approach for toxicity pathway identification will also be discussed.

Webinar Recording

Webinar Slides

MDCPSS and CTSS Webinar: Practical Application of Computational Models to Predict Release Kinetics and Toxicity of Compounds Released from Medical Devices

Speaker(s):
Ron Brown, Risk Science Consortium, LLC
David Saylor, US Food and Drug Administration

Use of Computational Models to Predict the Toxicity of Compounds Released from Medical Device Materials, Ron Brown, Risk Science Consortium, LLC

Toxicity data are not available for many compounds released from medical device materials.  However, computational models can be used to help predict the toxicity of data-poor compounds.  This webinar will show how computational toxicology models can be used for the biological safety assessment of medical devices and will provide some tips on the practical application of the models and interpretation of model-derived predictions.  Some of the more recently developed tools to help with compound grouping and Read Across will be featured and promising models to provide quantitative toxicity threshold values (e.g. NOAEL, LOAEL) will be explored, along with models to predict biocompatibility endpoints such as skin sensitization.

Physics-based models to predict patient exposure to medical device leachables, David Saylor, US Food and Drug Administration

Medical device materials contain chemicals that may pose toxicological concern(s) if released in sufficient quantities.  Toxicological risk assessment approaches are increasingly being used in lieu of animal testing to address these concerns.  Currently, these approaches rely primarily on in vitro extraction testing to estimate the potential for patients to be exposed to chemicals that may possibly leach out of device materials, but the clinical relevance of the test results are often ambiguous.  Recent developments suggest physics-based models can be used to provide more clinically relevant exposure estimates.  However, the lack of data available to parameterize and validate these models presents a barrier to routine use. This presentation will provide an overview of these approaches, including considerations in developing and parameterizing exposure models, strategies that can be used to address the challenges associated with limited data, and potential future directions and improvements.

Webinar Recording

Ron Brown | David Saylor

State of the Science: QSAR Modeling of Skin Sensitization

Pred-Skin: A Web Portal for Accurate Prediction of Human Skin Sensitizers

Speaker(s):
Vinicius Alves, PhD, Researcher Assistant Professor, Division of Chemical Biology and Medicinal Chemistry, University of North Carolina at Chapel Hill

Description:
Safety assessment is an essential component of the regulatory acceptance of industrial chemicals. Previously, we have developed a model to predict the skin sensitization potential of chemicals for two assays, the human patch test and murine local lymph node assay and implemented this model in a web portal. Here, we report on the substantially revised and expanded freely available web tool, Pred-Skin version 3.0. This up-to-date version of Pred-Skin incorporates multiple quantitative structure–activity relationship (QSAR) models developed with in vitro, in chemico, and mice and human in vivo data, integrated into a consensus naïve Bayes model that predicts human effects. Individual QSAR models were generated using skin sensitization data derived from human repeat insult patch tests, human maximization tests, and mouse local lymph node assays. In addition, data for three validated alternative methods, the direct peptide reactivity assay, KeratinoSens, and the human cell line activation test, were employed as well. Models were developed using open-source tools and rigorously validated according to the best practices of QSAR modeling. Predictions obtained from these models were then used to build a naïve Bayes model for predicting human skin sensitization with the following external prediction accuracy: correct classification rate (89%), sensitivity (94%), positive predicted value (91%), specificity (84%), and negative predicted value (89%). As an additional assessment of model performance, we identified 11 cosmetic ingredients known to cause skin sensitization but were not included in our training set, and nine of them were accurately predicted as sensitizers by our models. Pred-Skin can be used as a reliable alternative to animal tests for predicting human skin sensitization.

Evaluation of the Global Performance of Eight In Silico Skin Sensitization Models Using Human Data

Speaker(s):
Emily Golden, MS Graduate Student, Center for Alternatives to Animal Testing, Johns Hopkins University

Description:
Allergic contact dermatitis, or the clinical manifestation of skin sensitization, is a leading occupational hazard. Several testing approaches exist to assess skin sensitization, but in silico models are perhaps the most advantageous due to their high speed and low-cost results. Many in silico skin sensitization models exist, though many have only been tested against results from animal studies (e.g., LLNA); this creates uncertainty in human skin sensitization assessments in both a screening and regulatory context. This project’s aim was to evaluate the accuracy of eight in silico skin sensitization models against two human data sets: one highly curated (Basketter et al., 2014) and one screening level (HSDB). The binary skin sensitization status of each chemical in each of’s QSAR Toolbox, UL’s REACHAcross™, Danish QSAR Database, TIMES-SS, and Lhasa Limited’s Derek Nexus). Models were assessed for coverage, accuracy, sensitivity, and specificity, as well as optimization features (e.g., probability of accuracy, applicability domain, etc.), if available. While there was a wide range of sensitivity and specificity, the models generally performed comparably to the LLNA in predicting human skin sensitization status (i.e., approximately 70–80% accuracy). Additionally, the models did not mispredict the same compounds, suggesting there might be an advantage in combining models. In silico skin sensitization models offer accurate and useful insights in a screening context; however, further improvements are necessary so these models may be considered fully reliable for regulatory applications.

Skin Sensitization In Silico Protocol

Speaker(s):
Glenn Myatt, PhD, CEO, Leadscope, Inc.

Description:
In silico toxicology protocol for skin sensitization computational frameworks that incorporate in silico best practices will have a critical role in expanding the application of computational toxicology. The in silico toxicology protocol consortium is one example which is developing a series of protocols (equivalent to experimental test guidelines) to support the acceptance and adoption of such methods. A general framework for organizing such protocols has been published with endpoint-specific protocols completed for genetic toxicology and skin sensitization. A series of additional protocols are in development. The skin sensitization protocol incorporates recent advances in the understanding of key events along the adverse outcome pathway. These events, along with other supporting information, are incorporated into the in silico toxicology hazard assessment framework. Information from both experimental data and/or in silico model results for individual effects or mechanisms are incorporated within this framework. Based on the relevance of the mechanisms and effects as well as the strengths and limitations of the experimental systems used to identify them, rules and principles are defined for deriving skin sensitization assessments based on the weight of the information. Further, the assignments of reliability and confidence scores that reflect the overall strength of the assessment are discussed. This presentation will provide an overview of the in silico toxicology protocol framework, describe the skin sensitization hazard assessment framework and present a worked example.

Webinar Recording

Vinicius Alves | Emily Golden | Glenn Myatt

Speaker(s):
Richard Goodwin, PhD, Head of Imaging & AI Imaging & Data Analytics, Clinical Pharmacology & Safety Sciences, AstraZeneca R&D

Matthew Jacobsen, PhD, DipACVP, Director of Pathology Regulatory Safety Centre of Excellence, Clinical Pharmacology & Safety Sciences, AstraZeneca R&D

Description:
Novel and integrated molecular imaging technologies are offering a new view on the molecular and cellular landscape across the tissue microenvironment. They are able to map in unprecedented detail the impact of new therapies and so provide new ways to study disease, the patient population and the efficacy and safety of drugs. This is empowering its scientists to see the complexity of a disease in unprecedented detail to enable effective development and selection of new medicines. Today cutting-edge imaging technologies are increasingly used to support studies in to the efficacy and safety of drugs through the R&D pipeline. This presentation will introduce the range of novel in vivo and ex vivo imaging technologies employed, describe the data challenges associated with scaling up the use of molecular imaging technologies and address the new data integration and mining challenges. Novel computational methods are required for large cohort imaging studies that involve tissue based multi-omics analysis that integrate spatial relationships in unprecedented detail.

In parallel to the development and deployment of novel molecular imaging technologies, a revolution in toxicological digital pathology is also occurring. We are moving to a digital future where our pathologists are assessing and reviewing safety studies digitally. This requires a significant change and acceptance by regulators. During this presentation we will also hear how we are transforming toxicologic pathology peer review with the first GLP validation of the digital pathology and how this is a critical step and delivering a digital future for drug discovery and development.

Webinar Recording

Richard Goodwin Presentation | Matthew Jacobsen Presentation | Intro slides Richard Matt Presentation

Description:
This virtual annual business meeting included a brief welcome from the CTSS President, highlights announcing 2020–2021 events and accomplishments, recognition to sponsors and supporters, and an award ceremony followed by awardee presentations.

Presentation

Predictive Approaches for Assessing Cross-Species Conservation of Endocrine Targets

Speaker(s):
Sara Vliet, PhD, ORISE Postdoctoral Fellow, United States Environmental Protection Agency (EPA), Great Lakes Toxicology

Description:
Developed computational predictions for cross-species susceptibility to chemical exposures of agency concern. This session will highlight current advances in computational tools and, through an Androgen Receptor case study, demonstrate current applications. This presentation will cover efforts to corroborate computational predictions through systematic review of existing data.

Dr. Vliet Presentation

Deep Neural Network Modeling Identifies NSAIDs as Highly Bioactivated Hepatotoxic Metabolites

Speaker(s):
Mary Schleiff (formerly Davis), Doctor of Philosophy Candidate, University of Arkansas for Medical Sciences, Recipient of the 2019–2020 CTSS Pfizer Student Travel Award

Description:
Computational approaches and tools are used to identify a subset of structurally similar drugs with variable hepatotoxic risk on the US market: diphenylamine non-steroidal anti-inflammatory drugs (NSAID). A Xenosite metabolic model was used to identify potential mechanisms and sites of metabolic bioactivation. All seven diphenylamine NSAIDs are predicted to form quinone-species reactive metabolites but the sites and likelihood of bioactivation varied depending upon their minor structural differences (methylation, halogenation, etc.). This work can guide pharmaceutical development and identify at-risk patients to promote more personalized patient care.

Dr. Schleiff Presentation

Adverse Outcome Pathway (AOP) Network Guided High-Dimensional Modeling for Drug Induced Liver Injury

Speaker(s):
Dong Wang, Senior Statistician, US Food and Drug Administration (FDA), National Center for Toxicological Research, Recipient of the 2019–2020 CTSS Pfizer Paper of the Year

Description:
Adverse outcome pathway (AOP) networks were used to filter high dimensional in vitro assay data for constructing predictive models of drug induced liver injury (DILI). The findings were also corroborated with data from post market surveillance.

Wang Presentation

Webinar Recording

Intro and Concluding Sildes

Artificial Intelligence in the Design of Safer Medicines—Science or Science Fiction?

Speaker(s):
Nigel Greene, AstraZeneca

Description:
There are strong economic drivers to reduce the costs associated with the discovery of new medicines. However, drug discovery and development are multiparameter optimization problems that require a novel medicine to have a fine balance between its efficacy, ADME properties and safety. Although the number of clinical failures from safety has been reduced in recent years, there are still improvements that could be made. Data science and artificial intelligence is a potential method to both improve the safety profile of new drugs as well as reduce the costs and time to bring these to the clinic. This talk will highlight some of the current investments in computational methods and highlight some of the key gaps in realizing these benefits.

Webinar Recording

Intro Slides

Computational Methods in Next-Generation Risk Assessment of Consumer Products

Speaker(s):
Steve Gutsell, PhD, Unilever Safety and Environmental Assurance Centre
Colworth Science Park, Sharnbrook, London, United Kingdom

Abstract:
The use of computational models is common-place across the consumer product development pipeline. The mantra of “in silico first” applies to almost all areas of efficacy and safety. Safety assessments of novel chemicals (or novel uses of existing chemicals) is required to cover human exposure and release to the environment. A key principle is that modern risk assessments should be exposure led and quantitative, so that predictions relating to dose may be interpretable in the context of exposure. Interpreting outputs from in vitro assays requires an understanding of both cellular exposure and how these relate to in vivo internal concentrations. The concept of exposure-led assessment allows methods to identify and characterize hazard and risk to be applied in a tiered manner starting with exposure-based waiving e.g.TTC, DST, ecoTTC) and building through in silico, in chemico and in vitro assays combined in weight of evidence assessments. As with any safety assessment the need for robustness, reliability and traceability of decisions made using computational approaches is paramount to their acceptance and success. The concept and benefits of pathways-based approaches to risk assessment will also be introduced in the context of both environmental and human health risk assessment.

Webinar Recording

Webinar Presentation

US FDA Experience in the Regulatory Application of (Q)SAR Modeling

Speaker(s):
Naomi Kruhlak, BS, PhD
US Food & Drug Administration, Center for Drug Evaluation & Research

Description:
The US Food and Drug Administration (FDA) actively promotes and supports research into state-of-the-art computational toxicology approaches including (quantitative) structure-activity relationship, also known as(Q)SAR, modeling for regulatory decision-making. Within FDA’s Center for Drug Evaluation and Research, a significant focus of this research is on the safety assessment of drug impurities under the ICH M7 guideline, where a (Q)SAR prediction can be used as an alternative to empirical testing to assess mutagenic potential. The successful implementation of the ICH M7 guideline has led to the development of best practices for the conduct of (Q)SAR analyses, evaluation of model outputs, and submission of (Q)SAR data to regulators. Building on this success, new opportunities for the regulatory application of (Q)SARs are being explored for other endpoints including carcinogenicity and endocrine disruption, as well as for other components of FDA-regulated products beyond impurities.

This presentation will provide an in-depth look at current regulatory practice and expectations for the use of (Q)SAR models under ICH M7, as well as highlight emerging applications of the methodology to chemical entities and endpoints of regulatory importance to FDA.

Webinar Recording

Webinar Slides

Description:
This annual luncheon/business meeting included a brief welcome from the CTSS President, highlights announcing 2019–2020 events and accomplishments, recognition to sponsors and supporters, and an award ceremony followed by awardee presentations.

Presentation

An Introduction to In Silico Toxicology

Description:
In silico toxicology has a number of unique benefits compared to in vivo or in vitro methods: it is fast to run, it does not require any test material, and often provides an understanding of the structural (and mechanistic) basis for any toxicity prediction. In addition, it has been validated as fit-for-purpose for specific toxicology endpoints and there exist protocols and other documentation to support its adoption. As such, it is being successfully used in a variety of applications across toxicology.

This webinar will outline the process of developing and using in silico methods to predict toxicity. The two commonly used in silico methodologies, expert rule-based (or structural alerts) and statistical-based (or QSAR models), will be described. In silico models are built from existing knowledge or automatically derived from training sets of historical toxicity data. The construction of these models will be described as well as how these models could be used to make a prediction and support an expert review. A series of case studies will be used to illustrate this process using first principles and commercial software.

• Introduction to different methodologies
• Expert rule-based methodologies
• Statistical-based methodologies
• Predicting toxicity
• Expert review

Webinar Recording

Webinar Presentation