The Leadscope bacterial mutagenicity expert alerts journey to universal regulatory acceptance

It has taken over 10 years, but now Leadscope’s expert rule-based methodology is used every day to support submissions to regulatory agencies around the world, including pharmaceutical impurity assessments aligned with the ICH M7 guideline1, assessment of extractable and leachables, along with other applications. The following blog reviews this journey.

The first step in this process was to collect and consolidate alerts reported in the literature, including the seminal publications by John Ashby and Raymond Tennant.2,3 These alerts were then qualified, refined and new alerts identified using a large database of chemicals with bacterial mutagenicity study results assembled from quality public sources. Over the years, it has also been possible to further validate and refine the alerts using proprietary data. A new methodology, referred to as SAR fingerprinting, was developed to allow companies to share structure-activity (mutagenicity) knowledge without disclosing proprietary chemicals or study results.4,5 The latest version made use of around 50,000 chemicals. These exercises have identified mitigating factors which significantly minimize the number of false positives results as compared to using public domain alerts alone.

At the same time, individual alerts have been annotated with knowledge of their mechanistic basis. This is generally a highly time-consuming manual activity to summarize the literature. This has been supplemented with automated annotation of tester strain sensitivities associated with each alert, since knowing the tester strains likely to be positive (based on any alerts that fire) may support an expert review or streamlined in vitro testing. In the latest version of the alerts, the system can now even predict direct- versus indirect-acting mutagens.

Over the years we have also developed computational tools to support applying the alerts as part of an overall hazard assessment (including statistical-based methods as well as bacterial mutagenicity and carcinogenicity database searching). Since most regulatory guidelines emphasize the importance of being aligned with OECD validation principles6, a new methodology for applicability domain assessment was developed. Another important consideration was computational tools to ensure the results are totally transparent, including access to all chemicals matching the alerts from our database of around 20,000 chemicals alongside information on their mechanistic basis. Such methods are critical to enable an expert review of the prediction.

We have been engaged in a series of collaborative working groups to both assess the performance of the technology7,8, while also developing principles and procedures for performing such computational toxicology assessments9,10,11.

It has been rewarding to see how such methods have evolved where they are now relied upon for international regulatory submissions. Many lessons have been learned along this journey which are being applied in the development of new in silico approaches to support emerging regulatory and industrial applications.

If you would like to discuss in more detail or collaborate on emerging in silico methods, please get in touch (Glenn J. Myatt; glenn.myatt@instem.com).

References

  1. ICH, M7 (R1) Assessment and control of DNA reactive (mutagenic) impurities in pharmaceuticals to limit potential carcinogenic risk, European Medicines Agency, 2017. https://database.ich.org/sites/default/files/M7_R1_Guideline.pdf.
  2. Ashby, J., Tennant, R.W., 1988. Chemical structure, Salmonella mutagenicity and extent of carcinogenicity as indicators of genotoxic carcinogenesis among 222 chemicals tested in rodents by the U.S. NCI/NTP. Mutat. Res. 204, 17–115.
  3. Myatt, G.J., Beilke, L.D., Cross, K.P. (2016) In Silico Tools and their Application in: Reference Module in Chemistry, Molecular Sciences and Chemical Engineering (eds J. Reedijk). Elsevier. 10.1016/B978-0-12-409547-2.12379-0
  4. E. Ahlberg, et al., Extending (Q)SARs to incorporate proprietary knowledge for regulatory purposes: A case study using aromatic amine mutagenicity, Regul. Toxicol. Pharmacol. 77 (2016) 1–12. https://doi.org/10.1016/j.yrtph.2016.02.003.
  5. Amberg, A., et al. (2018) Extending (Q)SARs to incorporate proprietary knowledge for regulatory purposes: is aromatic N-oxide a structural alert for predicting DNA-reactive mutagenicity? Mutagenesis 34, 67–82. 10.1093/mutage/gey020
  6. OECD, Guidance Document on the Validation of (Quantitative) Structure-Activity Relationship [(Q)SAR] Models, OECD Environment, Health and Safety Publications, Paris, 2007. https://doi.org/10.1787/9789264085442-en.
  7. Honma, M., Kitazawa, A., Cayley, A., et al., (2019) Improvement of quantitative structure–activity relationship (QSAR) tools for predicting Ames mutagenicity: outcomes of the Ames/QSAR International Challenge Project, Mutagenesis, 34(1), 3–16, https://doi.org/10.1093/mutage/gey031
  8. Myatt, G.J., et al., 2022. Implementation of in silico toxicology protocols within a visual and interactive hazard assessment platform. Comput. Toxicol. 21, 100201. https://doi.org/10.1016/j.comtox.2021.100201
  9. Amberg, A., et al., 2016. Principles and procedures for implementation of ICH M7 recommended (Q)SAR analyses. Regul. Toxicol. Pharmacol. 77, 13–24. https://doi.org/10.1016/j.yrtph.2016.02.004
  10. A. Amberg, et al., Principles and procedures for handling out-of-domain and indeterminate results as part of ICH M7 recommended (Q)SAR analyses, Regul. Toxicol. Pharmacol. 102 (2019) 53–64. https://doi.org/10.1016/j.yrtph.2018.12.007.
  11. Hasselgren, C., et al., (2020) Management of Pharmaceutical ICH M7 (Q)SAR Predictions – The Impact of Model Updates, Regul. Toxicol. Pharmacol., 118, 104807. https://doi.org/10.1016/j.yrtph.2020.104807

New in silico paper on acute toxicity

We are pleased to announce the publication of a new paper “Principles and Procedures for Assessment of Acute Toxicity Incorporating In Silico Methods”.1 This paper presents the results from a significant cross-industry collaboration to support the application of in silico methods for (1) product research and development, (2) product approval and registration as well as (3) the transport, storage and handling of chemicals.

The paper also complements the recent paper assessing whether (Q)SAR models are fit-for-purpose for classification and labelling.2

The new paper describes:

  • a framework for hazard assessment of acute toxicity using different sources of information including in silico methods alongside in vitro or in vivo experiments
  • the endpoints from in vitro studies commonly used for predicting acute toxicity
  • key pathways and key triggering mechanisms underlying acute toxicity
  • the state-of-the-art in prediction of acute toxicity
  • an expert review using weight-of-evidence considerations
  • diverse and practical use cases using in silico approaches

We wish to thank and congratulate all collaborators on this important project supporting the 3Rs.

Please get in touch if you would like to discuss this paper or get involved with other collaborative working groups (Glenn Myatt; glenn.myatt@instem.com).

Reference

  1. Zwickl, C. et al., Principles and Procedures for Assessment of Acute Toxicity Incorporating In Silico Methods, Computational Toxicology, 100237. https://doi.org/10.1016/j.comtox.2022.100237
  2. Bercu, J. et al., A cross-industry collaboration to assess if acute oral toxicity (Q)SAR models are fit-for-purpose for GHS classification and labelling, Regulatory Toxicology and Pharmacology, 120, 2021, 104843. https://doi.org/10.1016/j.yrtph.2020.104843

Neurotoxicity hazard assessment framework that integrates in silico approaches

Earlier this year, we described the results from a collaborative working group that developed a neurotoxicity hazard assessment framework that integrates in silico approaches1. This cross-industry collaboration resulted in a publication within a special issue of the Journal of Computational Toxicology2.

Dr. Arianna Bassan will be presenting the project on July 22nd, 2022 at 9:00 am EST in an upcoming webinar organized by International Neurotoxicology Association titled “Seizing the moment: Issues and technological advancements in mitigation strategies for drugs with seizure liability” 3. In this talk, in silico methods available today that support the assessment of neurotoxicity based on knowledge of chemical structure will be reviewed, followed by the presentation of a conceptual framework for the integration of in silico methods with experimental information.

Please get in touch with me (Glenn Myatt; glenn.myatt@instem.com) if you would like more information on assessment of neurotoxicity.

References

  1. https://insilicoinsider.blog/2022/04/07/neurotoxicity-hazard-assessment-framework-that-integrates-in-silico-approaches/
  2. Crofton et al (2022) Current status and future directions for a neurotoxicity hazard assessment framework that integrates in silico approaches, Computational Toxicology, 22, 100223. https://doi.org/10.1016/j.comtox.2022.100223
  3. https://www.neurotoxicology.org/ina-webinars/

Novel application of in silico methods in the assessment of Extractables and Leachables

New therapeutic modalities play a critical role in our health and safety. Novel therapeutics may be comprised of biologically based molecules including peptides, monoclonal antibodies, and genetic materials. The quality and safety of these products can be assessed using experimental systems; however, it is important to ask whether in silico methods can add value to the analysis workflow. In silico methods have an advantage of providing a rapid screening of effects that could be used in early stages of development to triage and prioritize chemicals based on the likelihood that they may lead to the assessed effect.

In considering how in silico models could be used to support the development and quality of therapies consisting of biomolecules, we can leverage the ability of models to predict chemical reactivity. Active pharmaceutical ingredients which consist of biomolecules (biomolecule-APIs) may be exposed to leachables from container/closure systems, delivery devices or manufacturing systems. The leachables may react with the biomolecule and potentially impact the safety and/or quality of the drug product. In this context we investigated the use of in silico methods which either predict peptide reactivity, direct acting mutagenicity (as a surrogate for reactivity with a DNA or RNA containing API) or assigns a reaction domain (description of the nature by which a covalent interaction may or may not occur) to predict the potential chemical reactivity of leachables in an extractables and leachables (E&L) database. The prevalence of potentially reactive chemicals was low with 16.7% of the set of 801 leachables predicted as reactive based on assignment to a domain; whereas, 16.1% and 5.4% were considered reactive based on the peptide reactivity model and direct acting mutagenicity alerts respectively. To validate the use of such models in early screening, we selected 22 representative E&L chemicals and experimentally determined the reactivity of these chemicals with insulin glargine. The peptide reactivity and reaction domain models were used to predict the observed reactivity. A consensus approach based on the model results, was found to be conservative; that is, most of the reactive chemicals were flagged for reactivity. Factors such as steric hindrance was identified as an important consideration for expert review.

Leachables often exist at very low concentration, and a flag by the in silico models may not necessarily translate into a safety or quality issue. As such, a workflow, which entails a review of model predictions to ensure reliability of the prediction, as well as product quality and safety attributes could be referenced to determine whether follow-up studies are necessary.

This work was conducted by members of the Biomolecule Reactivity Consortium: a group of experts with diverse experience and expertise in the assessment of extractables and leachables. A manuscript is being prepared for publication. Click here to access a recorded presentation on this work which was presented at the E&L USA 2022 conference.

If you would like to learn more about this initiative, or the group’s activities, please feel free to contact me at candice.johnson@instem.com.

Special issue of the Journal of Computational Toxicology

A special issue of the Journal of Computational Toxicology on the in silico toxicology protocol initiative1 is currently being finalized. Myself (Glenn Myatt), Kevin Cross and Candice Johnson from Instem were happy to support this effort as guest editors and many of the articles are already available on-line.

The in silico toxicology protocol initiative was established to create a clear framework and process for performing a computational toxicology assessment to ensure such assessments are performed in a consistent and repeatable manner based on the best science and best practices in the field. These protocols also help to defend the use of in silico models to colleagues, peers, and regulators. To date, publications detailing the protocol framework2 as well as two protocols to support genetic toxicology3 and skin sensitization4 have been published.

This special edition covers a series of new publications describing the projects goals and overriding framework, how confidence in any assessments can be systemically and transparently determined and reviews the current state of the science in the prediction of liver, heart, lung, and kidney toxicity as well as carcinogenicity and neurotoxicity. A paper also introduces a new visual and interactive platform to support a rapid execution and documentation aligned with the protocols.

There are currently seven publications in this edition covering:

  1. Increasing the acceptance of in silico toxicology through development of protocols and position papers
  2. Evaluating confidence in toxicity assessments based on experimental data and in silico predictions
  3. In silico approaches in organ toxicity hazard assessment: current status and future needs in predicting liver toxicity
  4. In silico approaches in organ toxicity hazard assessment: current status and future needs for predicting heart, kidney and lung toxicities
  5. In silico approaches in carcinogenicity hazard assessment: current status and future needs
  6. Current status and future needs for a neurotoxicity hazard assessment framework that integrates in silico approaches
  7. Implementation of in silico toxicology protocols within a visual and interactive hazard assessment platform

If you would like to discuss any of these publications or any upcoming collaborative working groups, please get in touch (Glenn Myatt; glenn.myatt@instem.com).

References

  1. https://www.sciencedirect.com/journal/computational-toxicology/special-issue/10ZD1MJ9982
  2. G.J. Myatt et al., In silico toxicology protocols, Regul. Toxicol. Pharmacol. 96 (2018) 1–17. https://doi.org/10.1016/j.yrtph.2018.04.014.
  3. C. Hasselgren et al., Genetic toxicology in silico protocol, Regul. Toxicol. Pharmacol. 107 (2019) 104403. https://doi.org/10.1016/j.yrtph.2019.104403.
  4. C. Johnson et al., Skin sensitization in silico protocol, Regul. Toxicol. Pharmacol. 116 (2020) 104688. https://doi.org/10.1016/j.yrtph.2020.104688.

6-year anniversary of the ICH M7 principles and procedures publication

Over 6 years ago, the ICH M7 pharmaceutical impurities guideline1 was in its implementation phase, and we were approached to consider writing a cross-industry publication to outline a protocol for performing a (Q)SAR assessment aligned with the guideline. A collaborative working group was established and work began to create this publication. The paper was published in 2016 and outlined considerations for incorporating experimental data on the impurities, rules and principles for performing a (Q)SAR analysis including expert review considerations, recommendations on formats to document the results along with case studies.2

In 2019, it was followed up with a paper outlining how to support indeterminate and out-of-domain (Q)SAR results that included case studies and expert review approaches.3 The paper also outlined an analysis that documented the risk of missing a mutagenic impurity based on the differing results from the two (Q)SAR methodologies.

Both papers are open access and have been widely used and cited, with 77 combined citations including the ICH M7 Q&A4. The principles and procedures have also been incorporated into software applications. For example, the Leadscope model applier includes an ICH M7 protocol implementation that streamlines the application of ICH M7 aligned analyses based on these publications, including tools to guide an expert review.5,6 Instem also uses these publications to support its Predict™ computational toxicology services.7

The development of these papers also helped a series of follow-on activities, including the development of an in silico toxicology protocol framework8 as well as other applications that incorporate similar principles, such as the hazard assessment of extractables and leachables.

If you would like to discuss any of these papers or projects, please get in touch (Glenn Myatt; glenn.myatt@instem.com).

References

  1. ICH, 2017. ICH guideline M7 (R1). Assessment and control of DNA reactive (mutagenic) impurities in pharmaceuticals to limit potential carcinogenic risk (No. EMA/CHMP/ICH/83812/2013), ICH Harmonised Guideline. European Medicines Agency. https://database.ich.org/sites/default/files/M7_R1_Guideline.pdf
  2. Amberg, A., et al., 2016. Principles and procedures for implementation of ICH M7 recommended (Q)SAR analyses. Regul. Toxicol. Pharmacol. 77, 13–24. https://doi.org/10.1016/j.yrtph.2016.02.004
  3. A. Amberg et al., 2019 Principles and procedures for handling out-of-domain and indeterminate results as part of ICH M7 recommended (Q)SAR analyses, Regul. Toxicol. Pharmacol. 102, 53–64. https://doi.org/10.1016/j.yrtph.2018.12.007.
  4. ICH M7 Guideline: Assessment and control of DNA reactive (mutagenic) impurities in pharmaceticals to limit potential carcinogenic risk. Questions and Answers. https://www.ich.org/page/multidisciplinary-guidelines#7-3
  5. https://www.instem.com/solutions/insilico/computational-toxicology.php
  6. Myatt, G.J., Bassan, A., Bower, D., Johnson, C., Miller, S., Pavan, M., Cross, K.P., 2022. Implementation of in silico toxicology protocols within a visual and interactive hazard assessment platform. Comput. Toxicol. 21, 100201. https://doi.org/10.1016/j.comtox.2021.100201
  7. https://www.instem.com/solutions/insilico/predict.php
  8. Myatt, G.J., et al., 2018. In silico toxicology protocols. Regul. Toxicol. Pharmacol. 96, 1–17. https://doi.org/10.1016/j.yrtph.2018.04.014

Advancing Prediction of Nitrosamine Carcinogenicity Potency

This week, Dr. Kevin Cross from Instem is presenting at the 2022 Genetic Toxicology Association meeting on recent progress in predicting N-nitrosamine carcinogenicity potency. The presentation outlines progress over the last year and provides an update on different collaborative working groups, including a recently announced EMA-MutAmind project funded by the European Medicine’s Agency and led by Fraunhofer ITEM that includes Instem as a partner.1

The presentation provides an overview of the different reaction mechanisms involved in N-nitrosamines mutagenicity and compares Nitrosamine Drug Substance-Related Impurities (NDSRIs) against historical information on tested nitrosamines. This includes an assessment of historical Ames negative studies that have historically tested positive for rodent carcinogenicity. Recent testing of these ‘false negatives’ are presented and discussed relative to experimental protocols used and structural classes where issues occur, allowing for more directed future assessments of nitrosamines.

The presentation is available to view:

<View presentation>

Please get in touch if we can help answer any questions on this important topic (Glenn Myatt, glenn.myatt@instem.com).

References

  1. Instem Awarded EMA Research Grant. https://www.instem.com/news/articles/0905-instem-awarded-EMA-research-grant.php

Predicting toxicity using read-across

Read-across is an example of a flexible in silico toxicology methodology whereby a toxicity prediction for a chemical (often referred to as the target compound) is made using experimental data on one or more similar analogs (often referred to as source compound(s)). This is a lengthy process highly reliant on an expert review of the information to identify and justify the analogs selected and the prediction made.

Read-across is used in many current and emerging applications, including supporting the assessment of repeated dose toxicity of chemicals or N-nitrosamines carcinogenicity potency categories.

At the heart of any read-across is the need to identify suitable analogs. This is where comprehensive high-quality structure-searchable databases are critical, containing both public and proprietary experimental data. Global similarity is often used to search such databases by identifying and prioritizing chemical analogs based on their whole molecule similarity. However, in many instances it is important to retrieve analogs that have a local similarity around the structural class that is most likely associated with the toxicological endpoint being assessed. This may also consider the structural environment responsible for any activation or deactivation of the specific toxicity of interest. The use of structural alerts or QSAR models may support this assessment and justification, when run against both the target and source compounds(s). Actually, mechanistic similarity can be in general explored by running alert profilers, such as reaction domain or bioactivation profilers, across both the target and source chemicals to identify similarities and differences. In this way, reactivity of the source and target chemicals is compared in terms of structural alerts or reactive features. Indeed, justifying the selected analogs based on common biological properties or mechanisms, and toxicity profiles is key in a read-across framework. Further analysis of the physico-chemical properties is also prudent.

The read-across process can be assisted and simplified by the availability of both high-quality structure-searchable databases coupled with user-friendly read-across tools that are aligned with regulatory expectations, including appropriate documentation of the results. We are actively developing such methodologies and technologies to support further acceptance of read-across. If you are interested in collaborating in this area, please get in touch (Glenn Myatt; glenn.myatt@instem.com).

In silico applications in the assessment of Extractables and Leachables

The Product Quality Research Institute (PQRI) recently published on their recommendations for safety thresholds and best practices for extractables and leachables (E&L) in Parenteral Drug products1. The publication describes, and addresses, issues related to the definition and procedures for defining safety thresholds. Within the document, areas of application for in silico methods are also defined. We see three main categories for the in silico approaches.

  1. Established procedures

Such methods have well defined and documented procedures and could be accepted for regulatory submissions. The in silico assessment of mutagenicity falls within this category and it is important to examine the factors/characteristics of this assessment type as we look forward to developing other areas. There is scientific consensus on the use and application of in silico approaches to assess mutagenicity based on the publication of a detailed guideline, ICH M72. Further, the availability of data3, and growth of SAR principles4 and application of the methods5 have allowed for advanced interpretation of results.

  1. Evolving procedures

The areas which are considered evolving have foundational elements laid out and the application structure is being defined. In these areas, models are available, SAR understanding is acceptable and various application methods are being considered6. Skin sensitization and irritation/corrosion models appear to fall into this category. In these areas there is a need for documented procedures and questions remain as how to interpret and utilize potency predictions in the context of safety thresholds. The ELSIE consortium is working on advancing this area and others.

  1. Emerging applications 

Within the publication by PQRI, special consideration is given to biological products where the potential for a reactive E&L compound to react with therapeutic proteins is mentioned. Here the proof of concept for the application of theoretically relevant models is needed. As such, efforts are ongoing with the in silico protocol initiative to examine the performance of relevant models against empirical data and examine the scope of potentially reactive chemicals within the E&L chemical space.

At the Extractables and Leachables 2022 conference, I presented on these topics with focus on emerging applications. Here we demonstrated the conservative nature of a consensus result based on a prediction of peptide reactivity and reaction domain profiling, while detailing a decision workflow for the application of in silico models to assess the reactivity of E&L molecules with biomolecules. Noted from the conference is the ongoing work in the application of in silico methods in the context of E&L evaluations; and additional details to be provided in the ISO-10993 guideline and various upcoming publications.

If you would like to get in touch, please do not hesitate to contact candice.johnson@instem.com.

References

  1. Product Quality Research Institute (PQRI). 2021. Safety Thresholds and Best Demonstrated Practices for Extractables and Leachables in Parenteral Drug Products (Intravenous, Subcutaneous, and Intramuscular). https://www.pda.org/bookstore/product-detail/6576-pqri-pdp.
  2. ICH, 2017. M7 (R1) Assessment and control of DNA reactive (mutagenic) impurities in pharmaceuticals to limit potential carcinogenic risk (No. EMA/CHMP/ICH/83812/2013), ICH Harmonised Guideline. European Medicines Agency. https://database.ich.org/sites/default/files/M7_R1_Guideline.pdf
  3. Landry, Curran et al. 2019. “Transitioning to Composite Bacterial Mutagenicity Models in ICH M7 (Q)SAR Analyses.” Regulatory Toxicology and Pharmacology 109: 104488. https://www.sciencedirect.com/science/article/pii/S0273230019302521.
  4. Ahlberg, Ernst et al. 2016. “Extending (Q)SARs to Incorporate Proprietary Knowledge for Regulatory Purposes: A  Case Study Using Aromatic Amine Mutagenicity.” Regulatory toxicology and pharmacology : RTP 77: 1–12.
  5. Amberg, Alexander et al. 2019. “Principles and Procedures for Handling Out-of-Domain and Indeterminate Results as Part of ICH M7 Recommended (Q)SAR Analyses.” Regulatory Toxicology and Pharmacology.
  6. Johnson, Candice et al. 2022. “Evaluating Confidence in Toxicity Assessments Based on Experimental Data and in Silico Predictions.” Computational Toxicology 21: 100204. https://www.sciencedirect.com/science/article/pii/S2468111321000505.

Neurotoxicity hazard assessment framework that integrates in silico approaches

This week we are pleased to welcome Dr. Kevin Crofton and Dr. Arianna Bassan as guest contributors to the blog.

Within the in silico toxicology project, a position paper on neurotoxicity is appearing in a special issue of the journal of Computational Toxicology1.

This paper discusses the need for the development of more informative new approach methodologies (NAM) to assess developmental (DNT) and adult (NT) neurotoxicity induced by xenobiotics. The use of NAMs (including in silico methodologies that predict toxicity from chemical structure such as QSARs and structural alerts) is ideally based on the understanding of the biological mechanism underpinning neurotoxicity (e.g., blocking of GABAA receptors by organochlorines leading to seizures and pyrethroid effects on voltage-gated sodium channels).

In our paper, we discuss the development of a decision assessment framework that builds on the emerging vision of mechanistically informed approaches. Such assessment framework requires the identification of relevant effects/mechanisms and endpoints together with their mutual relationships as it is meant to guide the integration of data from different sources (in vitro data, animal in vivo data, human data) for the assessment of neurotoxicity. The expedited advances in toxicology and the ever-increasing understanding of the biological pathways underlying NT and DNT will impose a continuous adaptation of this framework. However, in our work, we have tried to capture the current knowledge to provide a construct that integrates the use of in silico methods for DNT and NT assessments and that may inform future development of in silico approaches.

The paper also includes overviews on:

  • Known mechanisms underlying DNT/NT
  • Current state of the art in DNT/NT testing
  • In silico approaches for the assessment of neurotoxicity based on knowledge of chemical structure

This work should support the future development of protocols, namely standardized approaches, for the use of in silico methods for the assessments of NT and DNT based on chemical structures ensuring transparent, consistent, and defendable results. Because of the large numbers of chemicals used in all aspects of global commerce, such standardized procedures would be beneficial for compliance across related regulatory programs (e.g., FDA and EFSA, TSCA and REACH).

For more information on this collaboration or related projects, please contact Glenn Myatt (glenn.myatt@instem.com).

References

  1. Crofton et al (2022) , Current status and future directions for a neurotoxicity hazard assessment framework that integrates in silico approaches, 22, 100223, https://doi.org/10.1016/j.comtox.2022.100223