AOP4EUpest: mapping of pesticides in adverse outcome pathways using a text mining tool

Editorial trend: adverse outcome pathway (AOP) and computational strategy - towards new perspectives in ecotoxicology

The adverse outcome pathway (AOP) has been conceptualized in 2010 as an analytical construct to describe a sequential chain of causal links between key events, from a molecular initiating event leading to an adverse outcome (AO), considering several levels of biological organization. An AOP aims to identify and organize available knowledge about toxic effects of chemicals and drugs, either in ecotoxicology or toxicology, and it can be helpful in both basic and applied research and serve as a decision-making tool in support of regulatory risk assessment. The AOP concept has evolved since its introduction, and recent research in toxicology, based on integrative systems biology and artificial intelligence, gave it a new dimension. This innovative in silico strategy can help to decipher mechanisms of action and AOP and offers new perspectives in AOP development. However, to date, this strategy has not yet been applied to ecotoxicology. In this context, the main objective of this short article is to discuss the relevance and feasibility of transferring this strategy to ecotoxicology. One of the challenges to be discussed is the level of organisation that is relevant to address for the AO (population/community). This strategy also offers many advantages that could be fruitful in ecotoxicology and overcome the lack of time, such as the rapid identification of data available at a time t, or the identification of "data gaps". Finally, this article proposes a step forward with suggested priority topics in ecotoxicology that could benefit from this strategy.

Advancing Computational Toxicology by Interpretable Machine Learning

Chemical toxicity evaluations for drugs, consumer products, and environmental chemicals have a critical impact on human health. Traditional animal models to evaluate chemical toxicity are expensive, time-consuming, and often fail to detect toxicants in humans. Computational toxicology is a promising alternative approach that utilizes machine learning (ML) and deep learning (DL) techniques to predict the toxicity potentials of chemicals. Although the applications of ML- and DL-based computational models in chemical toxicity predictions are attractive, many toxicity models are "black boxes" in nature and difficult to interpret by toxicologists, which hampers the chemical risk assessments using these models. The recent progress of interpretable ML (IML) in the computer science field meets this urgent need to unveil the underlying toxicity mechanisms and elucidate the domain knowledge of toxicity models. In this review, we focused on the applications of IML in computational toxicology, including toxicity feature data, model interpretation methods, use of knowledge base frameworks in IML development, and recent applications. The challenges and future directions of IML modeling in toxicology are also discussed. We hope this review can encourage efforts in developing interpretable models with new IML algorithms that can assist new chemical assessments by illustrating toxicity mechanisms in humans.

AOP-helpFinder 2.0: Integration of an event-event searches module

To support the use of alternative methods in regulatory assessment of chemical risks, the concept of adverse outcome pathway (AOP) constitutes an important toxicological tool. AOP represents a structured representation of existing knowledge, linking molecular initiating event (MIE) initiated by a prototypical stressor that leads to a cascade of biological key event (KE) to an adverse outcome (AO). Biological information to develop such AOP is very dispersed in various data sources. To increase the chance of capturing relevant existing data to develop a new AOP, the AOP-helpFinder tool was recently implemented to assist researchers to design new AOP. Here, an updated version of AOP-helpFinder proposes novel functionalities. The main one being the implementation of an automatic screening of the abstracts from the PubMed database to identify and extract event-event associations. In addition, a new scoring system was created to classify the identified co-occurred terms (stressor-event or event-event (which represent key event relationships) to help prioritization and support the weight of evidence approach, allowing a global assessment of the strength and reliability of the AOP. Moreover, to facilitate interpretation of the results, visualization options are also proposed. The AOP-helpFinder source code are fully accessible via GitHub, and searches can be performed via a web interface at http://aop-helpfinder-v2.u-paris-sciences.fr/.

Application of AOPs to assist regulatory assessment of chemical risks - Case studies, needs and recommendations

While human regulatory risk assessment (RA) still largely relies on animal studies, new approach methodologies (NAMs) based on in vitro, in silico or non-mammalian alternative models are increasingly used to evaluate chemical hazards. Moreover, human epidemiological studies with biomarkers of effect (BoE) also play an invaluable role in identifying health effects associated with chemical exposures. To move towards the next generation risk assessment (NGRA), it is therefore crucial to establish bridges between NAMs and standard approaches, and to establish processes for increasing mechanistically-based biological plausibility in human studies. The Adverse Outcome Pathway (AOP) framework constitutes an important tool to address these needs but, despite a significant increase in knowledge and awareness, the use of AOPs in chemical RA remains limited. The objective of this paper is to address issues related to using AOPs in a regulatory context from various perspectives as it was discussed in a workshop organized within the European Union partnerships HBM4EU and PARC in spring 2022. The paper presents examples where the AOP framework has been proven useful for the human RA process, particularly in hazard prioritization and characterization, in integrated approaches to testing and assessment (IATA), and in the identification and validation of BoE in epidemiological studies. Nevertheless, several limitations were identified that hinder the optimal usability and acceptance of AOPs by the regulatory community including the lack of quantitative information on response-response relationships and of efficient ways to map chemical data (exposure and toxicity) onto AOPs. The paper summarizes suggestions, ongoing initiatives and third-party tools that may help to overcome these obstacles and thus assure better implementation of AOPs in the NGRA.

Characterization of Potential Adverse Outcome Pathways Related to Metabolic Outcomes and Exposure to Per- and Polyfluoroalkyl Substances Using Artificial Intelligence

Human exposure to per- and polyfluoroalkyl substances (PFAS) has been associated with numerous adverse health effects, depending on various factors such as the conditions of exposure (dose/concentration, duration, route of exposure, etc.) and characteristics associated with the exposed target (e.g., age, sex, ethnicity, health status, and genetic predisposition). The biological mechanisms by which PFAS might affect systems are largely unknown. To support the risk assessment process, AOP-helpFinder, a new artificial intelligence tool, was used to rapidly and systematically explore all available published information in the PubMed database. The aim was to identify existing associations between PFAS and metabolic health outcomes that may be relevant to support building adverse outcome pathways (AOPs). The collected information was manually organized to investigate linkages between PFAS exposures and metabolic health outcomes, including dyslipidemia, hypertension, insulin resistance, and obesity. Links between PFAS exposure and events from the existing metabolic-related AOPs were also retrieved. In conclusion, by analyzing dispersed information from the literature, we could identify some associations between PFAS exposure and components of existing AOPs. Additionally, we identified some linkages between PFAS exposure and metabolic outcomes for which only sparse information is available or which are not yet present in the AOP-wiki database that could be addressed in future research.

Natural language processing in toxicology: Delineating adverse outcome pathways and guiding the application of new approach methodologies

Adverse Outcome Pathways (AOPs) are conceptual frameworks that tie an initial perturbation (molecular initiating event) to a phenotypic toxicological manifestation (adverse outcome), through a series of steps (key events). They provide therefore a standardized way to map and organize toxicological mechanistic information. As such, AOPs inform on key events underlying toxicity, thus supporting the development of New Approach Methodologies (NAMs), which aim to reduce the use of animal testing for toxicology purposes. However, the establishment of a novel AOP relies on the gathering of multiple streams of evidence and information, from available literature to knowledge databases. Often, this information is in the form of free text, also called unstructured text, which is not immediately digestible by a computer. This information is thus both tedious and increasingly time-consuming to process manually with the growing volume of data available. The advancement of machine learning provides alternative solutions to this challenge. To extract and organize information from relevant sources, it seems valuable to employ deep learning Natural Language Processing techniques. We review here some of the recent progress in the NLP field, and show how these techniques have already demonstrated value in the biomedical and toxicology areas. We also propose an approach to efficiently and reliably extract and combine relevant toxicological information from text. This data can be used to map underlying mechanisms that lead to toxicological effects and start building quantitative models, in particular AOPs, ultimately allowing animal-free human-based hazard and risk assessment.

A Narrative Literature Review of Natural Language Processing Applied to the Occupational Exposome

The evolution of the Exposome concept revolutionised the research in exposure assessment and epidemiology by introducing the need for a more holistic approach on the exploration of the relationship between the environment and disease. At the same time, further and more dramatic changes have also occurred on the working environment, adding to the already existing dynamic nature of it. Natural Language Processing (NLP) refers to a collection of methods for identifying, reading, extracting and untimely transforming large collections of language. In this work, we aim to give an overview of how NLP has successfully been applied thus far in Exposome research. Methods: We conduct a literature search on PubMed, Scopus and Web of Science for scientific articles published between 2011 and 2021. We use both quantitative and qualitative methods to screen papers and provide insights into the inclusion and exclusion criteria. We outline our approach for article selection and provide an overview of our findings. This is followed by a more detailed insight into selected articles. Results: Overall, 6420 articles were screened for the suitability of this review, where we review 37 articles in depth. Finally, we discuss future avenues of research and outline challenges in existing work. Conclusions: Our results show that (i) there has been an increase in articles published that focus on applying NLP to exposure and epidemiology research, (ii) most work uses existing NLP tools and (iii) traditional machine learning is the most popular approach.

Adverse outcome pathway from activation of the AhR to breast cancer-related death

Adverse outcome pathways (AOPs) are formalized and structured linear concepts that connect one molecular initiating event (MIE) to an adverse outcome (AO) via different key events (KE) through key event relationships (KER). They are mainly used in eco-toxicology toxicology, and regulatory health issues. AOPs must respond to specific guidelines from the Organization for Economic Co-operation and Development (OECD) to weight the evidence between each KE. Breast cancer is the deadliest cancer in women with a poor prognosis in case of metastatic breast cancer. The role of the environments in the formation of metastasis has been suggested. We hypothesized that activation of the AhR (MIE), a xenobiotic receptor, could lead to breast cancer related death (AO), through different KEs, constituting a new AOP. An artificial intelligence tool (AOP-helpfinder), which screens the available literature, was used to collect all existing scientific abstracts to build a novel AOP, using a list of key words. Four hundred and seven abstracts were found containing at least a word from our MIE list and either one word from our AO or KE list. A manual curation retained 113 pertinent articles, which were also screened using PubTator. From these analyses, an AOP was created linking the activation of the AhR to breast cancer related death through decreased apoptosis, inflammation, endothelial cell migration, angiogenesis, and invasion. These KEs promote an increased tumor growth, angiogenesis and migration which leads to breast cancer metastasis and breast cancer related death. The evidence of the proposed AOP was weighted using the tailored Bradford Hill criteria and the OECD guidelines. The confidence in our AOP was considered strong. An in vitro validation must be carried out, but our review proposes a strong relationship between AhR activation and breast cancer-related death with an innovative use of an artificial intelligence literature search.

Investigation of a derived adverse outcome pathway (AOP) network for endocrine-mediated perturbations

An adverse outcome pathway (AOP) is a compact representation of the available mechanistic information on observed adverse effects upon environmental exposure. Sharing of events across individual AOPs has led to the emergence of AOP networks. Since AOP networks are expected to be functional units of toxicity prediction, there is current interest in their development tailored to specific research question or regulatory problem. To this end, we have developed a detailed workflow to construct an endocrine-relevant AOP (ED-AOP) network based on the existing information available in AOP-Wiki. We propose a cumulative weight of evidence (WoE) score for each ED-AOP based on the WoE scores assigned to key event relationships (KERs) by AOP-Wiki, revealing gaps in AOP development. Connectivity analysis of the ED-AOP network comprising 48 AOPs reveals 7 connected components and 12 isolated AOPs. Subsequently, we apply standard network measures to perform an in-depth analysis of the two largest connected components of the ED-AOP network. Notably, the graph-theoretic analyses led to the identification of important events including points of convergence or divergence in the ED-AOP network. These findings can suggest potential adverse outcomes and facilitate the development of new endpoints or assays for chemical risk assessment. Detailed analysis of the largest component in the ED-AOP network gives insights on the systems-level perturbations caused by endocrine disruption, emergent paths, and stressor-event associations. In sum, the derived ED-AOP network can provide a broader view of the biological events disrupted by endocrine disruption, as well as facilitate better risk assessment of environmental chemicals.

The Exposome and Toxicology: A Win-Win Collaboration

The development of the exposome concept has been one of the hallmarks of environmental and health research for the last decade. The exposome encompasses the life course environmental exposures including lifestyle factors from the prenatal period onwards. It has inspired many research programs and is expected to influence environmental and health research, practices, and policies. Yet, the links bridging toxicology and the exposome concept have not been well developed. In this review, we describe how the exposome framework can interface with and influence the field of toxicology, as well as how the field of toxicology can help advance the exposome field by providing the needed mechanistic understanding of the exposome impacts on health. Indeed, exposome-informed toxicology is expected to emphasize several orientations including (1) developing approaches integrating multiple stressors, in particular chemical mixtures, as well as the interaction of chemicals with other stressors, (2) using mechanistic frameworks such as the adverse outcome pathways to link the different stressors with toxicity outcomes, (3) characterizing the mechanistic basis of long-term effects by distinguishing different patterns of exposures and further exploring the environment-DNA interface through genetic and epigenetic studies, and (4) improving the links between environmental and human health, in particular through a stronger connection between alterations in our ecosystems and human toxicology. The exposome concept provides the linkage between the complex environment and contemporary mechanistic toxicology. What toxicology can bring to exposome characterization is a needed framework for mechanistic understanding and regulatory outcomes in risk assessment.

Emerging Bioinformatics Methods and Resources in Drug Toxicology

Assessing the drug safety at an early stage of a drug discovery program is a critical issue. With the recent advances in molecular biology and genomic, massive amounts of generated and accumulated data by advanced experimental technologies such as RNA sequencing or proteomics start to be at the disposal of the scientific community. Innovative and adequate bioinformatic methods, tools, and protocols are required to analyze properly these diverse and extensive data sources with the aim to identify key features that are related to toxicity observations. Furthermore, the assessment of drug safety can be performed across multiple scales of complexity from molecular, cellular to phenotypic levels; therefore, the application of network science contributes to a better interpretation of the drug's exposure effect on human health. Here, we review databases containing toxicogenomics and chemical-phenotype information, as well as appropriated bioinformatics approaches that are currently used to analyze such data. Extension to others methods such as dose-responses, time-dependent processes, and text mining is also presented giving an overview of suitable tools available for a best practice of drug safety analysis.

Capturing a Comprehensive Picture of Biological Events From Adverse Outcome Pathways in the Drug Exposome

Background: The chemical part of the exposome, including drugs, may explain the increase of health effects with outcomes such as infertility, allergies, metabolic disorders, which cannot be only explained by the genetic changes. To better understand how drug exposure can impact human health, the concepts of adverse outcome pathways (AOPs) and AOP networks (AONs), which are representations of causally linked events at different biological levels leading to adverse health, could be used for drug safety assessment.

Methods: To explore the action of drugs across multiple scales of the biological organization, we investigated the use of a network-based approach in the known AOP space. Considering the drugs and their associations to biological events, such as molecular initiating event and key event, a bipartite network was developed. This bipartite network was projected into a monopartite network capturing the event–event linkages. Nevertheless, such transformation of a bipartite network to a monopartite network had a huge risk of information loss. A way to solve this problem is to quantify the network reduction. We calculated two scoring systems, one measuring the uncertainty and a second one describing the loss of coverage on the developed event–event network to better investigate events from AOPs linked to drugs.

Results: This AON analysis allowed us to identify biological events that are highly connected to drugs, such as events involving nuclear receptors (ER, AR, and PXR/SXR). Furthermore, we observed that the number of events involved in a linkage pattern with drugs is a key factor that influences information loss during monopartite network projection. Such scores have the potential to quantify the uncertainty of an event involved in an AON, and could be valuable for the weight of evidence assessment of AOPs. A case study related to infertility, more specifically to “decrease, male agenital distance” is presented.

Conclusion: This study highlights that computational approaches based on network science may help to understand the complexity of drug health effects, with the aim to support drug safety assessment.

AOP-helpFinder webserver: a tool for comprehensive analysis of the literature to support adverse outcome pathways development

MOTIVATION

Adverse outcome pathways (AOPs) are a conceptual framework developed to support the use of alternative toxicology approaches in the risk assessment. AOPs are structured linear organizations of existing knowledge illustrating causal pathways from the initial molecular perturbation triggered by various stressors, through key events (KEs) at different levels of biology, to the ultimate health or ecotoxicological adverse outcome.

RESULTS

Artificial intelligence can be used to systematically explore available toxicological data that can be parsed in the scientific literature. Recently, a tool called AOP-helpFinder was developed to identify associations between stressors and KEs supporting thus documentation of AOPs. To facilitate the utilization of this advanced bioinformatics tool by the scientific and the regulatory community, a webserver was created. The proposed AOP-helpFinder webserver uses better performing version of the tool which reduces the need for manual curation of the obtained results. As an example, the server was successfully applied to explore relationships of a set of endocrine disruptors with metabolic-related events. The AOP-helpFinder webserver assists in a rapid evaluation of existing knowledge stored in the PubMed database, a global resource of scientific information, to build AOPs and Adverse Outcome Networks supporting the chemical risk assessment.

AVAILABILITY AND IMPLEMENTATION

AOP-helpFinder is available at http://aop-helpfinder.u-paris-sciences.fr/index.php.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Identification of non-validated endocrine disrupting chemical characterization methods by screening of the literature using artificial intelligence and by database exploration

BACKGROUND

Exposure to endocrine disrupting chemicals (EDCs) represents a critical public health threat. Several adverse health outcomes (e.g., cancers, metabolic and neurocognitive/neurodevelopmental disorders, infertility, immune diseases and allergies) are associated with exposure to EDCs. However, the regulatory tests that are currently employed in the EU to identify EDCs do not assess all of the endocrine pathways.

OBJECTIVE

Our objective was to explore the literature, guidelines and databases to identify relevant and reliable test methods which could be used for prioritization and regulatory pre-validation of EDCs in missing and urgent key areas.

METHODS

Abstracts of articles referenced in PubMed were automatically screened using an updated version of the AOP-helpFinder text mining approach. Other available sources were manually explored. Exclusion criteria (computational methods, specific tests for estrogen receptors, tests under validation or already validated, methods accepted by regulatory bodies) were applied according to the priorities of the French Public-privatE Platform for the Pre-validation of Endocrine disRuptors (PEPPER) characterisation methods.

RESULTS

226 unique non-validated methods were identified. These experimental methods (in vitro and in vivo) were developed for 30 species using diverse techniques (e.g., reporter gene assays and radioimmunoassays). We retrieved bioassays mainly for the reproductive system, growth/developmental systems, lipogenesis/adipogenicity, thyroid, steroidogenesis, liver metabolism-mediated toxicity, and more specifically for the androgen-, thyroid hormone-, glucocorticoid- and aryl hydrocarbon receptors.

CONCLUSION

We identified methods to characterize EDCs which could be relevant for regulatory pre-validation and, ultimately for the efficient prevention of EDC-related severe health outcomes. This integrative approach highlights a successful and complementary strategy which combines computational and manual curation approaches.

Bringing together scientific disciplines for collaborative undertakings: a vision for advancing the adverse outcome pathway framework

BACKGROUND

Decades of research to understand the impacts of various types of environmental occupational and medical stressors on human health have produced a vast amount of data across many scientific disciplines. Organizing these data in a meaningful way to support risk assessment has been a significant challenge. To address this and other challenges in modernizing chemical health risk assessment, the Organisation for Economic Cooperation and Development (OECD) formalized the adverse outcome pathway (AOP) framework, an approach to consolidate knowledge into measurable key events (KEs) at various levels of biological organisation causally linked to disease based on the weight of scientific evidence (http://oe.cd/aops). Currently, AOPs have been considered predominantly in chemical safety but are relevant to radiation. In this context, the Nuclear Energy Agency's (NEA's) High-Level Group on Low Dose Research (HLG-LDR) is working to improve research co-ordination, including radiological research with chemical research, identify synergies between the fields and to avoid duplication of efforts and resource investments. To this end, a virtual workshop was held on 7 and 8 October 2020 with experts from the OECD AOP Programme together with the radiation and chemical research/regulation communities. The workshop was a coordinated effort of Health Canada, the Electric Power Research Institute (EPRI), and the Nuclear Energy Agency (NEA). The AOP approach was discussed including key issues to fully embrace its value and catalyze implementation in areas of radiation risk assessment.

CONCLUSIONS

A joint chemical and radiological expert group was proposed as a means to encourage cooperation between risk assessors and an initial vision was discussed on a path forward. A global survey was suggested as a way to identify priority health outcomes of regulatory interest for AOP development. Multidisciplinary teams are needed to address the challenge of producing the appropriate data for risk assessments. Data management and machine learning tools were highlighted as a way to progress from weight of evidence to computational causal inference.

Development of an adverse drug event network to predict drug toxicity

Despite of their therapeutic effects, drug's exposure may have negative effects on human health such as adverse drug reaction (ADR) and side effects (SE). Adverse drug events (ADEs), that correspond to an event occurring during the drug treatment (i.e. ADR and SE), is not necessarily caused by the drug itself, as this is the case with medical errors and social factors. Due to the complexity of the biological systems, not all ADEs are known for marketed drugs. Therefore, new and effective methods are needed to determine potential risks, including the development of computational strategies. We present an ADE association network based on 90,827 drug-ADE associations between 930 unique drug and 6221 unique ADE, on which we implemented a scoring system based on a pull-down approach for prediction of drug-ADE combination. Based on our network, ADEs proposed for three drugs, safinamide, sonidegib, rufinamide are further discussed. The model was able to identify, already known drug-ADE associations that are supported by the literature and FDA reports, and also to predict uncharacterized associations such as dopamine dysregulation syndrome, or nicotinic acid deficiency for the drugs safinamide and sonidegib respectively, illustrating the power of such integrative toxicological approach.