Computational Tools and Bioinformatics Strategies for Defining the Taxonomic Domain of Applicability of Adverse Outcome Pathways

Abstract

This article provides a comprehensive guide for researchers and drug development professionals on the critical task of defining the taxonomic domain of applicability (tDOA) for Key Event Relationships (KERs) within Adverse Outcome Pathways (AOPs). The tDOA determines across which species a mechanistic toxicity pathway is biologically plausible, a cornerstone for reliable cross-species extrapolation in regulatory safety assessment. The content spans from foundational AOP and KER concepts to advanced methodologies employing bioinformatics tools like SeqAPASS and G2P-SCAN for tDOA expansion [citation:1][citation:2][citation:5]. It addresses common troubleshooting challenges in establishing taxonomic conservation and details systematic, evidence-based approaches for validation. By synthesizing current best practices and case studies, this article aims to equip scientists with the knowledge to enhance the confidence and regulatory utility of AOPs for protecting both human and ecological health under a One Health framework [citation:1][citation:10].

Core Concepts: Understanding KERs and the Imperative for Taxonomic Domain of Applicability (tDOA)

Defining Key Event Relationships (KERs) as the Causal Backbone of AOPs

This whitepaper establishes Key Event Relationships (KERs) as the fundamental causal and predictive linkages that define the mechanistic structure of Adverse Outcome Pathways (AOPs). Within the AOP framework, KERs are the connections between measurable, sequential biological steps, leading from an initial molecular interaction to an adverse outcome relevant to risk assessment [1]. Their rigorous definition and quantitative characterization are paramount for transforming AOPs from qualitative narratives into predictive tools for toxicology and drug development. This document provides an in-depth technical guide on the anatomy, evidence assessment, and quantitative modeling of KERs, framed within the critical research imperative of understanding their taxonomic conservation—the extent to which these causal biological relationships are consistent across species. Mastery of KERs enables researchers to extrapolate data across levels of biological organization, enhance the weight-of-evidence for AOPs, and support the development of targeted testing strategies that reduce reliance on conventional animal studies [2] [3].

Foundational Concepts: AOPs and the Centrality of KERs

The Adverse Outcome Pathway (AOP) Framework

An Adverse Outcome Pathway (AOP) is a structured, linear representation of existing knowledge that describes a logical chain of causally linked biological events. This chain begins with a Molecular Initiating Event (MIE), where a chemical or stressor interacts with a specific biological target, and concludes at the level of an Adverse Outcome (AO) that is of direct relevance to risk assessment for human health or ecological systems [1]. The primary utility of the AOP framework lies in its ability to organize mechanistic information, facilitating the extrapolation of data measured at lower levels of biological organization (e.g., molecular, cellular) to predict outcomes at higher levels (e.g., organ, organism, population) [2].

The AOP framework has been formally adopted by the Organisation for Economic Co-operation and Development (OECD), which maintains a collaborative AOP Knowledge Base (AOP-KB). This platform allows the scientific community to develop, share, and review AOPs, ensuring that knowledge about key events and their relationships can be reused and built upon across multiple pathways [1].

Key Event Relationships (KERs): Definition and Role

Key Event Relationships (KERs) are the explanatory links that form the causal spine of an AOP. Each KER explicitly describes the directional and causal relationship between a pair of sequential Key Events (KEs)—an upstream KE (the cause) and a downstream KE (the effect) [2]. The relationship articulated in a KER provides the biological rationale for why a perturbation in the upstream event is expected to lead to a change in the downstream event.

The formal elements of a KER, as structured in the AOP-KB, include [2]:

• Identifier and Title: A unique ID and a descriptive phrase defining the linked KEs and their sequence.
• Upstream and Downstream Events: Specification of the causing (upstream) and responding (downstream) Key Events.
• Biological Plausibility: The mechanistic rationale supporting the causal connection.
• Empirical Support: Citable evidence demonstrating the linkage.
• Quantitative Understanding: Information on the response-response relationship, time-scale, and modulating factors.

By deconstructing a complex adverse outcome into a series of linked KERs, the framework provides a transparent, evidence-based map of toxicity pathways. This structure is essential for identifying knowledge gaps, designing relevant in vitro or in chemico tests, and supporting integrated approaches to testing and assessment (IATA) [3].

The Anatomy of a KER: From Qualitative Description to Quantitative Understanding

Essential Components for KER Description

A robust KER description moves beyond simple assertion to a comprehensive evidence package. The core components, as defined by the AOP Wiki, are summarized below [2].

Table 1: Core Descriptive Components of a Key Event Relationship

Component	Description	Purpose
Biological Plausibility	The biological, biochemical, or mechanistic rationale for the connection.	Establishes theoretical credibility based on established scientific knowledge.
Empirical Support	Direct, citable experimental evidence showing that a change in the upstream KE leads to a change in the downstream KE.	Provides observational or experimental proof of the linkage.
Uncertainties & Inconsistencies	Acknowledgment of conflicting data, knowledge gaps, or contextual factors that weaken the relationship.	Ensures transparency and identifies areas for further research.
Applicability Domain	Definition of the taxonomic, life stage, and sex contexts for which the KER is believed to hold true.	Critical for defining the boundaries and confidence in extrapolation.

Quantitative Characterization of KERs

The transition from qualitative to quantitative AOPs (qAOPs) hinges on the quantitative understanding of individual KERs [1]. This involves defining the functional relationship between the measurable changes in linked Key Events.

Table 2: Elements for the Quantitative Understanding of a KER

Element	Description	Example Data/Analysis
Response-Response Relationship	The mathematical function describing how the magnitude/timing of the downstream KE change depends on the upstream KE change.	Dose-response curves, kinetic models, linear/non-linear regression outputs (e.g., EC50, slope).
Time-Scale	The temporal dynamics (lag/lead time) between the perturbation of the upstream KE and the observable change in the downstream KE.	Time-course study data, kinetic rate constants.
Known Modulating Factors	Factors (e.g., age, sex, genotype, diet, co-exposure) that alter the strength, sensitivity, or dynamics of the KER.	Data showing different dose-response curves in different sub-populations or conditions.
Known Feedback Loops	Descriptions of positive or negative feedback mechanisms that may amplify or dampen the relationship, including their homeostatic limits.	Evidence of compensatory mechanisms or feed-forward signaling.

Quantitative analysis methods are crucial for deriving these relationships. Descriptive statistics (mean, variance) summarize experimental data, while inferential statistics are used to establish and model the linkage. Key techniques include regression analysis (to define response-response functions), correlation analysis (to measure association strength), and comparative tests like ANOVA or t-tests (to evaluate the impact of modulating factors across groups) [4].

Diagram: Quantitative Linkage Between Key Events. This model depicts a KER where an upstream KE perturbation drives a downstream KE response via a quantifiable relationship, which is subject to modulation by intrinsic/extrinsic factors and potential feedback mechanisms.

Experimental Protocols for KER Development and Validation

A Generic Protocol for Establishing Empirical Support

The following protocol outlines a standardized approach for generating empirical evidence to support a hypothetical KER.

1. Objective: To experimentally test whether a defined perturbation in an upstream Key Event (KEup) causes a predictable and measurable change in a downstream Key Event (KEdown). 2. Experimental Design:

• Model System: Select an appropriate in vitro (cell line, primary cells, microphysiological system) or in vivo model that is relevant to the biological domain of the KER.
• Perturbation Agent: Use a specific tool (e.g., a chemical inhibitor, agonist, siRNA, CRISPR-mediated gene edit) known to directly modulate KE_up.
• Dose/Concentration Range: Establish a range that includes sub-effective, effective, and supra-effective levels to define the response gradient.
• Time-Course: Include multiple time points to capture the dynamics of KEup change and the subsequent KEdown response.
• Controls: Include vehicle/negative controls and, if possible, a positive control that affects KE_down via an independent pathway.
• Replication: Biological and technical replicates are essential for statistical power. 3. Endpoint Measurement:
• KE_up Measurement: Quantify using a specific, validated assay (e.g., enzyme activity, receptor occupancy, protein phosphorylation, gene expression).
• KE_down Measurement: Quantify using a distinct, validated assay relevant to the downstream event. 4. Data Analysis:
• Perform statistical analysis (e.g., ANOVA with post-hoc test) to compare treated groups to controls.
• Conduct correlation or regression analysis (e.g., linear, logistic, Hill slope modeling) to define the relationship between the magnitude of KEup change and KEdown change across doses/time. 5. Interpretation: Determine if the data supports a causal linkage based on strength of association, temporality, and dose-response consistency. Document any uncertainties or inconsistent findings [2].

Protocol for Investigating Taxonomic Applicability of a KER

This protocol is designed to test the conservation of a KER across species, a core aspect of KER taxonomic conservation research.

1. Objective: To evaluate whether a well-supported KER in a reference species (e.g., human, rat) is conserved in one or more alternative species (e.g., zebrafish, nematode). 2. Experimental Design (Comparative Approach):

• Species Selection: Choose phylogenetically diverse species relevant to the intended application (e.g., ecological risk vs. human health).
• Conserved Perturbation: Employ an orthologous tool (e.g., a chemical that binds the conserved target) or technique (e.g., knockdown of the gene ortholog) to perturb KE_up in each species.
• Assay Alignment: Develop and optimize functional assays for KEup and KEdown in each species. The assays should measure the same biological function, though the specific method may differ.
• Standardized Conditions: Match experimental conditions (exposure duration, endpoint timing, vehicle) as closely as possible across species to isolate biological differences. 3. Data Analysis:
• Quantitatively compare the dose-response relationships (e.g., EC50, slope, maximal response) for the KER across species.
• Use statistical models (e.g., extra sum-of-squares F-test) to determine if response curves are significantly different. 4. Interpretation: Define the taxonomic applicability domain of the KER. Evidence of conserved response dynamics strongly supports the KER's broader relevance and enables confident cross-species extrapolation [5]. A lack of conservation highlights critical species-specific biology that must be accounted for.

Diagram: Workflow for Assessing KER Taxonomic Conservation. This protocol outlines a systematic approach to test the universality of a KER by comparing quantitative response relationships across different species.

KERs in the Context of Taxonomic Conservation Research

The assessment of a KER's taxonomic applicability is not a secondary consideration but a foundational research question with significant implications for the utility of an AOP. The core thesis of KER taxonomic conservation research posits that the fidelity and quantitative parameters of a KER may be conserved, modified, or absent across different species, depending on the evolutionary conservation of the underlying biological pathway [2].

Research in this domain systematically investigates whether a KER established in a model organism (e.g., rat) reliably predicts the same causal relationship in other species of interest (e.g., human, fish, or bird). This involves comparative studies, as outlined in Section 3.2, which mirror the methodologies used in large-scale ecological research to identify biases and gaps in evidence. For instance, a systematic map of meta-analyses in agricultural biodiversity revealed significant geographical and taxonomic biases, with certain groups (arthropods) over-studied and others (annelids, vertebrates) under-represented [5]. Similar systematic mapping of KER evidence across the taxonomical spectrum is essential to identify:

• Evidence Clusters: KERs that are overwhelmingly supported in a narrow set of model species (e.g., rodents).
• Taxonomic Gaps: KERs with little to no empirical support in taxonomically distant but ecologically or toxicologically important species.
• Conservation Patterns: Insights into which types of KERs (e.g., those involving highly conserved nuclear receptors vs. more lineage-specific immune pathways) are more likely to be broadly applicable.

Addressing these gaps is critical for building AOP networks that are robust for both human health and ecological risk assessment. It ensures that predictions are not erroneously extrapolated beyond their valid biological domain and guides the targeted generation of new data where it is most needed [5].

The Scientist's Toolkit: Essential Reagents and Methods for KER Research

This table details critical reagents, tools, and methodological approaches essential for investigating and characterizing Key Event Relationships.

Table 3: Research Toolkit for KER Investigation

Tool/Reagent Category	Specific Examples	Function in KER Research
Perturbation Agents	Selective chemical agonists/antagonists, siRNA pools, CRISPR-Cas9 gene editing kits, neutralizing antibodies.	To selectively modulate the upstream Key Event in a controlled manner to test its causal effect on the downstream event.
Activity/Quantification Assays	ELISA kits, phospho-specific antibodies, enzymatic activity assays (e.g., luminescence-based), qRT-PCR probes, reporter gene assays.	To quantitatively measure the changes in molecular or cellular key events (both upstream and downstream) with high specificity and sensitivity.
High-Content Screening (HCS)	Automated fluorescence microscopy, image analysis software (e.g., CellProfiler).	To capture complex phenotypic downstream KEs (e.g., cytotoxicity, morphological changes) in a quantitative, high-throughput manner.
Omics Technologies	RNA-Seq, targeted mass spectrometry proteomics, metabolomics platforms.	To explore unknown intermediates in a KER, identify novel modulating factors, or provide comprehensive evidence for pathway perturbations.
Data Analysis & Modeling Software	R/Bioconductor packages, Python (SciPy, Pandas), GraphPad Prism, specialized qAOP modeling platforms.	To perform statistical analysis, derive response-response models (regression), and visualize quantitative KER data [4].
AOP/KB Management Tools	OECD AOP-KB Wiki, AOP modeling software (e.g., AOPXplorer).	To formally document KERs according to OECD guidelines, link them to AOPs, and explore network relationships [2] [1].

Diagram: KERs as the Supported Causal Links in an AOP Network. An AOP is a chain of Key Events, but its predictive power resides in the well-supported KERs (evidence blocks) that causally link them together.

The Role of the Taxonomic Domain of Applicability (tDOA) in Predictive Toxicology

In predictive toxicology, the Taxonomic Domain of Applicability (tDOA) defines the biological taxa to which a given toxicity pathway or prediction is confidently applicable [6]. Its formalization is critical for moving beyond assumptions and providing evidence-based boundaries for extrapolating toxicological findings, particularly within frameworks like the Adverse Outcome Pathway (AOP) and its core unit, the Key Event Relationship (KER) [7]. A KER describes a causal, mechanistic link between two measurable Key Events (KEs) within an AOP. The central thesis of modern KER research asserts that the confidence in extrapolating a KER across species is predicated on the evolutionary conservation of the underlying biological mechanism [6]. Consequently, tDOA is not a peripheral descriptor but a foundational element that determines the utility of AOPs and KERs in regulatory decision-making and safety assessment for untested species [6].

The traditional development of AOPs often relies on empirical data from a single or a handful of model species. While biological plausibility may suggest broader relevance, the tDOA remains narrowly defined without explicit evidence [6]. This limitation is a significant hurdle in ecological risk assessment, where protecting diverse species is paramount, and in drug development, where translation from preclinical models to humans is critical [8]. Defining the tDOA systematically transforms an AOP from a species-specific narrative into a generalized, portable template for prediction. This guide details the mechanistic basis, computational assessment methodologies, and practical integration of tDOA to enhance the reliability and scope of predictive toxicology.

The Mechanistic Basis of tDOA: Structural and Functional Conservation

The tDOA of a KER or an AOP is supported by two pillars of evidence: structural conservation and functional conservation [6]. Structural conservation evaluates whether the essential biological components (e.g., genes, proteins, receptors, tissues) are present and measurably similar in the taxa of interest. Functional conservation assesses whether those components perform the same role within the proposed pathway in different species.

• Molecular Initiating Event (MIE) Conservation: The highest confidence for cross-species extrapolation exists at the MIE, typically a chemical interaction with a specific protein target. If the protein's ligand-binding domain is conserved, susceptibility to the chemical stressor is likely shared [6].
• KER and KE Conservation: Confidence for extrapolating downstream KERs depends on the conservation of the biological processes linking the events. This includes conserved signaling cascades, metabolic pathways, and cellular stress responses [9].

A seminal tool for evaluating structural conservation is the U.S. EPA's Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool [6]. It operates via a hierarchical, three-level analysis:

• Level 1 (Primary Sequence): Identifies potential orthologs across species by evaluating whole protein sequence similarity.
• Level 2 (Functional Domains): Assesses the conservation of specific protein domains known to be critical for function.
• Level 3 (Critical Residues): Examines the preservation of individual amino acid residues essential for protein-ligand interaction, protein-protein interaction, or catalytic activity [6].

The output provides a line of evidence for structural conservation, which, when combined with empirical data for functional conservation, forms a weight-of-evidence basis for defining the tDOA [9].

Experimental Protocol: SeqAPASS Analysis for tDOA Definition

The following protocol, derived from published case studies [6] [9], outlines the steps for using SeqAPASS to inform the tDOA of an AOP or KER.

1. Identify Query Proteins:

• Extract all protein targets involved in the AOP's KEs from the AOP-Wiki or literature. This includes the MIE target and proteins integral to intermediate KEs.
• Obtain reference protein sequences (primary amino acid sequences) for these targets from a well-characterized species (e.g., human, rat, Apis mellifera for bee AOPs) using databases like UniProt or NCBI Protein.

2. Perform SeqAPASS Analysis:

• Level 1 Analysis: Input the reference sequence into the SeqAPASS tool. Set the similarity threshold (often ≥70-80% for ortholog prediction). The tool generates a list of putative orthologs across a wide taxonomic range.
• Level 2 Analysis: For the reference sequence, define the functional domains (e.g., via Pfam). Run Level 2 analysis to evaluate the conservation of these specific domains in the orthologs identified in Level 1.
• Level 3 Analysis: Using published site-directed mutagenesis or crystallography data, identify the critical amino acid residues for the protein's relevant function (e.g., ligand binding). Run Level 3 analysis to check for residue conservation in the target species list.

3. Data Integration & tDOA Assignment:

• Compile results into a matrix. A species receives higher confidence for inclusion in the tDOA if its ortholog passes all three levels of analysis.
• Integrate SeqAPASS structural evidence with any available empirical in vitro or in vivo data from the literature demonstrating functional conservation of the KE or KER.
• Define the tDOA in the AOP-Wiki with clear, evidence-based statements (e.g., "Plausible for Hymenoptera, based on high conservation of nAChR ligand-binding domains and empirical acute toxicity data").

Diagram 1: Workflow for Evidence-Based tDOA Assessment (94 characters)

Computational Assessment of tDOA: Integrating Pathways and Tools

Beyond single-protein analysis, modern approaches integrate pathway-level conservation to strengthen tDOA predictions. Tools like Unilever's Genes to Pathways – Species Conservation Analysis (G2P-SCAN) map human gene sets to biological pathways and evaluate their conservation across common model species [9]. When combined with SeqAPASS, this provides a multi-layered, consensus view of biological conservation.

Integrated Protocol: SeqAPASS & G2P-SCAN for Enhanced tDOA [9]

• Chemical & Target Identification: Select a chemical stressor and identify its primary protein target(s) using ToxCast, ChEMBL, or literature.
• SeqAPASS Analysis: Perform full three-level SeqAPASS analysis on the primary target to predict direct susceptibility across a broad taxonomic range.
• Pathway Mapping & G2P-SCAN: For the primary target, identify the broader biological pathway(s) it participates in (e.g., via Reactome, KEGG). Input the core gene set of this pathway into G2P-SCAN to assess its conservation across key model species (human, rat, mouse, zebrafish, etc.).
• AOP Alignment & tDOA Expansion: Compare the conserved pathway to relevant AOPs. The consensus from both tools (protein target conservation + pathway conservation) provides strong evidence to expand the biologically plausible tDOA of the AOP to additional species.

Diagram 2: Three-Level SeqAPASS Analysis (79 characters)

Quantitative Performance and Case Study Data

The following tables summarize quantitative outcomes from key studies applying these methodologies.

Table 1: SeqAPASS Conservation Levels for Case Study Proteins (AOP 89: nAChR Activation to Colony Death) [6]

Protein Target	Role in AOP	Conservation in Apis mellifera (Honey Bee)	Conservation in Bombus spp. (Bumble Bee)	Conservation in Drosophila melanogaster (Fruit Fly)
nAChR alpha1	Molecular Initiating Event	Reference Species (100%)	High (Levels 1, 2, 3 Pass)	High (Levels 1, 2, 3 Pass)
nAChR beta1	Molecular Initiating Event	Reference Species (100%)	High (Levels 1, 2, 3 Pass)	Moderate (Levels 1, 2 Pass)
PLCgamma	Intracellular Signaling Key Event	Present	High Conservation Predicted	High Conservation Predicted
PKC	Intracellular Signaling Key Event	Present	High Conservation Predicted	High Conservation Predicted

Table 2: Summary of Integrated Tool Performance in Cross-Species Predictions [9]

Case Study Target	Tool Used	Primary Prediction	Key Outcome for tDOA
PPARα	SeqAPASS	High conservation across vertebrates, low in invertebrates.	Supported vertebrate-specific tDOA for PPARα-mediated AOPs.
PPARα	G2P-SCAN	Fatty acid metabolism pathway highly conserved in human, rat, mouse, zebrafish.	Corroborated pathway-level relevance in standard test species.
ESR1 (Estrogen Receptor)	SeqAPASS	High conservation in jawed vertebrates; absent in arthropods, mollusks.	Clearly defined tDOA boundary between vertebrates and invertebrates.
GABRA1	SeqAPASS & G2P-SCAN	High protein & pathway conservation across vertebrates and some invertebrates.	Provided evidence to expand tDOA for GABA-gated chloride channel AOPs.

Case Studies in tDOA Application

Case Study 1: Defining tDOA for a Pollinator AOP

AOP: Activation of the nicotinic acetylcholine receptor (nAChR) leading to colony death/failure in honey bees (Apis mellifera) [6]. Challenge: The AOP was developed for honey bees, but regulatory protection is needed for thousands of other bee species. tDOA Assessment: SeqAPASS analysis was performed on nine proteins in the pathway, from nAChR subunits to neuronal proteins. Results demonstrated high structural conservation of the MIE (nAChR) across Apis and non-Apis bees, providing strong evidence to expand the plausible tDOA to other bee genera like Bombus (bumble bees) [6]. This computational evidence can guide targeted empirical testing on key species.

Case Study 2: KER-Focused tDOA for Human Health

KER: Decreased all-trans retinoic acid (atRA) levels in developing ovaries leads to disrupted meiotic entry of oogonia [7]. Context: This KER is part of a potential AOP for reduced female fertility. Its utility depends on the conservation of atRA's role in meiosis across mammals. tDOA Rationale: The KER description explicitly reviews comparative biological evidence, showing the role of atRA in initiating meiosis is conserved across studied mammalian species [7]. This functional conservation, rooted in developmental biology, defines the KER's (and future AOP's) tDOA as "mammals," providing clear boundaries for extrapolation from rodent models to human health risk assessment.

Diagram 3: Logic Flow for KER Taxonomic Conservation (98 characters)

Table 3: Research Reagent Solutions for tDOA and KER Conservation Studies

Tool / Resource	Type	Primary Function in tDOA Research	Access / Example
SeqAPASS	Bioinformatics Tool	Evaluates structural conservation of protein targets across species via three-tiered sequence analysis. Provides direct line of evidence for tDOA.	https://seqapass.epa.gov/seqapass/ [6]
G2P-SCAN	Bioinformatics Tool	Maps gene sets to biological pathways and evaluates pathway conservation across a defined set of model species.	Available from Unilever; complementary to SeqAPASS [9].
AOP-Wiki	Knowledge Base	Central repository for AOPs and KERs. Platform for publishing and viewing defined tDOA based on assembled evidence.	https://aopwiki.org/ [6]
Comparative Toxicology Databases	Data Resource	Provide empirical toxicity data across species (e.g., ECOTOX, PubChem). Essential for anchoring/validating computational tDOA predictions.	US EPA ECOTOXicology Knowledgebase
Ortholog Prediction Databases	Data Resource	Provide pre-computed ortholog groups (e.g., OrthoDB, Ensembl Compara). Useful for rapid initial assessment of gene conservation.	https://www.orthodb.org/
Reactome / KEGG	Pathway Database	Provide curated biological pathways. Used to identify the broader context of a molecular target for pathway-level conservation analysis.	https://reactome.org/ [9]

The taxonomic domain of applicability is a critical, evidence-driven component that determines the real-world utility of predictive toxicology frameworks. By anchoring tDOA definitions in the systematic assessment of KER conservation—through integrated bioinformatics tools like SeqAPASS and G2P-SCAN, and empirical data—scientists can transform AOPs from descriptive models into reliable extrapolation tools. This rigorous approach directly addresses core challenges in ecological risk assessment and translational drug safety, ensuring protective measures and predictions are grounded in evolutionary biology. Future advancements in comparative 'omics and systems biology will further refine tDOA precision, solidifying its role as the bedrock of credible cross-species prediction.

Abstract Defining the taxonomic domain of applicability (tDOA) is critical for the confident use of Adverse Outcome Pathways (AOPs) in regulatory decision-making, particularly for the protection of untested species. This technical guide establishes structural and functional conservation as the two foundational pillars for extrapolating Key Event Relationships (KERs) across species. We present integrated bioinformatics and empirical methodologies to evaluate these pillars, supported by case studies on neurotoxicants in pollinators and disrupted retinoid signaling in mammalian fertility. The proposed framework enables the systematic expansion of AOP applicability, transforming tDOA from a static assumption into a dynamic, evidence-driven construct essential for predictive toxicology and chemical safety assessment.

The Adverse Outcome Pathway (AOP) framework organizes mechanistic knowledge into a causal sequence linking a Molecular Initiating Event (MIE) to an Adverse Outcome (AO) via intermediate Key Events (KEs) and Key Event Relationships (KERs) [10]. While AOPs are developed based on data from specific test species, their ultimate utility in ecological and human health risk assessment hinges on reliable extrapolation to broader taxa. The taxonomic domain of applicability (tDOA) defines the species for which a given AOP is considered valid [6]. Historically, tDOA has often been narrowly or ambiguously defined, limiting confidence in cross-species predictions [6].

This gap underscores the need for a rigorous, evidence-based approach. As articulated in OECD guidance, evaluating structural conservation (the presence and similarity of a biological entity) and functional conservation (the preservation of its biological role) forms the scientific basis for extrapolating KEs and KERs [6]. This guide posits that plausible tDOA is built upon these dual pillars. By leveraging publicly accessible bioinformatics tools to assess structural conservation and designing targeted in vitro and in vivo assays to confirm functional conservation, researchers can systematically expand and defend the tDOA of critical AOPs [6] [11]. This approach is fundamental to advancing the AOP framework from a descriptive exercise to a predictive, regulatory-ready tool.

Foundational Concepts: KERs, tDOA, and the Pillars of Conservation

Key Event Relationships (KERs) as the Core Unit of Knowledge

Within an AOP, a Key Event Relationship is a scientifically supported, causal link between an upstream and a downstream Key Event [10]. It is the KER that enables predictive inference: the state of a downstream KE can be inferred from the measured state of an upstream KE. Recent proposals argue that KERs, which encapsulate these causal hypotheses, should be recognized as the core modular building blocks of the AOP knowledge base [11]. This modularity is essential for the taxonomic extrapolation of AOPs. Establishing the tDOA for an entire AOP first requires establishing the tDOA for its constituent KERs, which in turn depends on the conservation of the KEs they connect.

Defining the Dual Pillars

The OECD identifies two primary considerations for defining the tDOA of a KE: structural conservation and functional conservation [6].

• Structural Conservation: Evidence that the biological entity (e.g., protein, receptor, enzyme) central to the KE is present and conserved in the taxa of interest. This includes conservation of primary amino acid sequence, functional protein domains, and specific amino acid residues critical for activity or ligand binding [6].
• Functional Conservation: Evidence that the conserved biological entity plays an analogous functional role in the pathway within the taxa of interest. This means that a perturbation (e.g., chemical inhibition) elicits a qualitatively and quantitatively similar response, leading to the same downstream biological effect [6].

These pillars are hierarchical and interdependent. Structural conservation is a prerequisite for—but does not guarantee—functional conservation. Functional conservation provides the definitive evidence for plausibility but is more resource-intensive to establish. Therefore, a robust assessment begins with broad screening for structural conservation to prioritize candidate taxa for focused empirical testing of functional conservation.

The AOP-Wiki and Standardized Development

The AOP-Wiki serves as the central repository for AOP knowledge [10]. The OECD AOP Developer's Handbook provides a standardized template for AOP development, emphasizing the need for clear documentation of the evidence supporting KERs and their taxonomic applicability [10]. Incorporating lines of evidence for structural and functional conservation into the AOP-Wiki is essential for transparently defining and expanding the tDOA [6].

Methodological Framework: Assessing Structural and Functional Conservation

A tiered strategy that integrates computational bioinformatics with empirical validation provides the most efficient and defensible pathway to establish plausible tDOA.

Tier 1: Bioinformatics Assessment of Structural Conservation

Primary Tool: Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) The SeqAPASS tool is a publicly accessible web-based platform designed to evaluate cross-species protein conservation through a hierarchical, three-level analysis [6].

Table 1: Hierarchical Analysis Levels of the SeqAPASS Tool [6]

Level	Analysis Focus	Interpretation & Output
Level 1	Primary amino acid sequence similarity.	Identifies putative orthologs across species. Provides a percent identity score and generates a taxonomic tree visualizing similarity.
Level 2	Conservation of known functional domains (e.g., ligand-binding domains, catalytic sites).	Determines if the core functional regions of the protein are present. Output indicates domain preservation or loss.
Level 3	Conservation of specific amino acid residues critical for function (e.g., chemical binding, protein-protein interaction).	Assesses if residues known to be essential for the MIE or KE are identical. Highest specificity for predicting susceptibility.

Protocol 1: Conducting a SeqAPASS Analysis for tDOA

• Identify Query Protein(s): Determine the specific protein(s) implicated in the MIE and each KE of the AOP (e.g., the nicotinic acetylcholine receptor for an MIE of nAChR activation).
• Gather Reference Sequences: Obtain the full-length amino acid sequences for the query protein from the well-studied "source" species (e.g., Apis mellifera). Use unique accession numbers from UniProt or NCBI.
• Execute Level 1 Analysis: Input the reference sequence into SeqAPASS. Set appropriate parameters (e.g., BLAST e-value cutoff). The tool performs alignments against a broad taxonomic database and returns a list of orthologs with similarity scores.
• Refine with Level 2 & 3 Analyses: For orthologs of interest, perform Level 2 analysis using defined domain models (e.g., PFAM). For Level 3, input the positions of known critical residues from the source species to check for identity in orthologs.
• Data Synthesis: Integrate results across all three levels to categorize species: high-confidence (strong conservation at all levels), moderate-confidence (conservation at Levels 1-2), or low-confidence (weak or no conservation).

Tier 2: Empirical Assessment of Functional Conservation

Bioinformatics provides a plausible hypothesis of conservation, which must be tested empirically. Functional assays confirm that the conserved structure leads to conserved function within the biological pathway.

Protocol 2: In Vitro Assay for Functional Conservation of an MIE (e.g., Receptor Activation)

• Objective: To test if a chemical stressor elicits a similar functional response (e.g., receptor activation, enzyme inhibition) in proteins from a source species and a putative conserved target species.
• Materials:
  • Recombinantly expressed target protein (e.g., receptor) from source and target species.
  • Cell line or membrane preparation for functional readout.
  • Reference agonist/antagonist and the stressor of interest.
  • Functional assay kit (e.g., calcium flux, cAMP, ligand binding).
• Method:
  • Establish concentration-response curves for a reference agonist using the source species' protein to validate the assay system.
  • Expose the target species' protein to the same reference agonist to confirm baseline functionality.
  • Expose both proteins to the stressor chemical across a range of concentrations.
  • Measure the functional endpoint (e.g., % receptor activation, % enzyme inhibition).
• Data Analysis: Compare half-maximal effective/inhibitory concentration (EC50/IC50) and maximal response (efficacy) between species. Statistically similar parameters provide strong evidence of functional conservation for the MIE.

Protocol 3: In Vivo or Ex Vivo Assay for Functional Conservation of a KER

• Objective: To test if modulating the upstream KE leads to the expected downstream KE in a whole organism or tissue from a target species.
• Materials:
  • Organisms/tissues from source and target species.
  • Stressor chemical or a tool compound known to modulate the upstream KE.
  • Equipment for measuring the upstream and downstream KEs (e.g., PCR for gene expression, histology for morphological change, ELISA for protein level).
• Method:
  • Administer the stressor to organisms/tissues from the target species at a range of doses/time points.
  • Measure the magnitude and temporal sequence of the upstream KE.
  • Measure the subsequent occurrence and magnitude of the downstream KE.
  • Compare the dose-response and temporal relationships with established data from the source species.
• Data Analysis: Establish if a predictable, causal relationship between the KEs exists in the target species. Evidence includes a dose-dependent progression, a consistent temporal sequence, and a strong correlation between the magnitudes of the two events.

Case Studies in Applied tDOA Assessment

Case Study 1: Neurotoxicity in Bees – Expanding an AOP Network

Background: An AOP network linking the activation of the nicotinic acetylcholine receptor (nAChR) to colony death/failure was developed for the honey bee (Apis mellifera) [6]. The tDOA for non-Apis bees (e.g., bumblebees, solitary bees) was uncertain.

Application of the Dual-Pillar Framework:

• Structural Conservation: Researchers used SeqAPASS to analyze nine proteins involved in the AOP. For the primary target, nAChR, Level 1-3 analyses confirmed high sequence similarity, preservation of ligand-binding domains, and identity of key binding-site residues across multiple bee genera [6].
• Functional Conservation: Empirical toxicity studies from the literature were synthesized. These showed that neonicotinoid insecticides (nAChR agonists) caused similar sub-lethal effects (impaired foraging, learning) and lethal toxicity in various non-Apis bees, supporting conserved function [6].
• tDOA Conclusion: The strong evidence for both structural and functional conservation allowed the proposed tDOA to be expanded from A. mellifera to include other bees (e.g., Bombus, Osmia spp.), with the weight of evidence being strongest for proteins directly involved in the early KEs (MIE and molecular KEs).

Case Study 2: Retinoid Signaling and Mammalian Fertility – Building a KER

Background: A KER linking decreased all-trans retinoic acid (atRA) levels in the fetal ovary to disrupted meiotic entry of oogonia is a component of a proposed AOP for reduced fertility [11]. The initial evidence was primarily from mouse models.

Application of the Dual-Pillar Framework:

• Structural Conservation: Key proteins in the pathway (ALDH1A enzymes for atRA synthesis, STRA8 receptor) are well-conserved across mammals. Bioinformatics analysis would show high conservation of functional domains and critical residues from mice to humans.
• Functional Conservation: Empirical evidence was gathered to support the KER's function:
  • Essentiality: Vitamin A deficient diets in rats block meiosis, mimicking the effect of atRA depletion [11].
  • Dose-Response: In vitro ovarian cultures show dose-dependent stimulation of meiosis markers by atRA [11].
  • Temporal Concordance: The peak of atRA synthesis coincides with meiotic onset in both mice and human fetal tissues [11].
• tDOA Conclusion: The KER demonstrates functional conservation across mammals, allowing the plausible tDOA to be defined as "mammals," which is critically informative for human-relevant chemical risk assessment. This case also highlights the practice of developing and peer-reviewing individual KERs as standalone units of knowledge [11].

Table 2: Summary of Case Study Evidence for tDOA

Case Study	AOP/KER Focus	Source Species	Structural Evidence	Functional Evidence	Plausible tDOA Conclusion
Bee Neurotoxicity [6]	AOP 89: nAChR activation -> colony failure	Apis mellifera (Honey bee)	High SeqAPASS scores for nAChR & other proteins across bee genera.	Similar neonicotinoid toxicity profiles in multiple bee species.	Expanded to include other bees (e.g., Bombus).
Mammalian Fertility [11]	KER 2477: ↓ atRA -> disrupted meiosis	Mouse (Mus musculus)	High sequence conservation of ALDH1A & STRA8 across mammals.	Conserved dose-response & essentiality in rats & human tissue data.	Mammals.

The Scientist's Toolkit: Essential Research Reagents & Solutions

Table 3: Key Research Reagents and Tools for tDOA Assessment

Item / Solution	Function in tDOA Assessment	Example/Provider
SeqAPASS Tool	Publicly accessible bioinformatics platform for Tier 1 structural conservation analysis across three hierarchical levels.	U.S. EPA SeqAPASS (https://seqapass.epa.gov/seqapass/)
UniProt / NCBI Protein Databases	Source of reliable reference protein sequences and functional annotations for query setup in SeqAPASS.	UniProtKB (https://www.uniprot.org/), NCBI Protein (https://www.ncbi.nlm.nih.gov/protein)
Recombinant Protein Expression Systems	Enables production of target proteins from source and test species for in vitro functional assays (Tier 2).	Baculovirus (insect cells), HEK293 (mammalian cells), or cell-free systems.
Functional Assay Kits	Provide standardized, optimized methods to measure protein activity (e.g., receptor activation, enzyme inhibition).	Calcium flux assays (FLIPR), cAMP detection kits, luciferase reporter assays.
Target-Specific Reference Chemicals	Well-characterized agonists/antagonists used as positive controls to validate functional assays across species.	e.g., Acetylcholine for nAChR, all-trans Retinoic Acid for retinoid receptors.
Custom RNA/DNA Probes/Primers	For measuring species-specific gene expression changes as molecular KEs in in vivo or ex vivo studies.	Designed from target species' sequenced genomes.
AOP-Wiki (aopwiki.org)	The central knowledge base for publishing AOPs, KEs, and KERs, including documented evidence for tDOA.	Managed by the OECD.

The establishment of a plausible taxonomic domain of applicability is a non-negotiable requirement for the credible use of AOPs in protecting human health and the environment. This guide demonstrates that a rigorous, tiered framework—grounded in the dual assessment of structural conservation via bioinformatics and functional conservation via empirical testing—provides a systematic and defensible pathway to achieve this goal.

Future advancements will depend on tighter integration between computational predictions and high-throughput empirical screening. The expansion of high-quality genomic and proteomic databases will enhance the resolution of tools like SeqAPASS. Concurrently, the development of standardized, cross-species in vitro assays (e.g., using conserved cell lines or tissue models) will improve the efficiency of functional conservation testing. By embracing this integrated approach, the toxicology community can transform tDOA from a statement of assumption into a dynamic, evidence-based conclusion, significantly strengthening the predictive power and regulatory applicability of the AOP framework.

The paradigm of chemical risk assessment is undergoing a fundamental transformation, driven by the regulatory and ethical imperative to reduce reliance on animal testing and to develop faster, more mechanistic safety evaluations. This shift is encapsulated in the development of Next Generation Risk Assessment (NGRA), defined as an exposure-led, hypothesis-driven approach that integrates in silico, in chemico, and in vitro New Approach Methodologies (NAMs) [12]. A critical challenge within this framework is ensuring the human and ecological relevance of NAM-based predictions. This necessitates a rigorous understanding of the Taxonomic Domain of Applicability (tDOA) for the Adverse Outcome Pathways (AOPs) and their constituent Key Event Relationships (KERs) that form the backbone of mechanistic risk assessment. Defined tDOA specifies the range of species for which a KER is biologically plausible, based on the conservation of molecular targets and pathways. The demand for its explicit definition is a direct regulatory driver, essential for justifying the use of NAM data in safety decisions, enabling credible cross-species extrapolation, and fulfilling the core NGRA principles of being relevant to humans and preventing harm [12]. This whitepaper provides a technical guide to defining tDOA through the lens of KER taxonomic conservation research, detailing the experimental and computational protocols that underpin this emerging standard in regulatory science.

The NGRA Imperative and the Centrality of AOPs/KERs

The traditional risk assessment paradigm, heavily dependent on apical endpoint data from animal studies, is increasingly viewed as resource-intensive, low-throughput, and limited in its mechanistic insight. In response, regulatory bodies worldwide are promoting NGRA. The U.S. EPA's NexGen program, initiated over a decade ago, exemplifies a long-standing effort to incorporate advances in molecular and systems biology into risk assessment [13]. The modern consensus defines NGRA as a tailored, iterative, and tiered process that moves away from prescribed animal tests toward a hypothesis-driven integration of diverse data sources [14] [12].

Central to the NGRA paradigm is the Adverse Outcome Pathway (AOP) framework. An AOP is a structured, linear representation of a toxicological mechanism, linking a Molecular Initiating Event (MIE) through a series of measurable Key Events (KEs) to an Adverse Outcome (AO) at the organism or population level. The causal linkages between two adjacent KEs are termed Key Event Relationships (KERs). KERs are recognized as the fundamental building blocks of toxicological knowledge within the AOP knowledge base [7]. Their quality, supported by empirical evidence and mechanistic understanding, determines the predictive utility and regulatory acceptance of an AOP.

The modularity of the AOP framework is its strength, allowing for the assembly of pathways and networks based on conserved biological processes. However, a pathway described in a model organism (e.g., Caenorhabditis elegans) is only relevant to human or ecological risk assessment if the underlying KERs are operative across the species of concern. This is where tDOA becomes critical. For a given KER, the tDOA defines the set of taxa for which there is established or inferred biological plausibility that the relationship holds, based on the conservation of the proteins, signaling pathways, and cellular functions involved.

Defining tDOA: A Multifaceted Technical Challenge

The tDOA is not a single data point but a conclusion derived from a weight-of-evidence analysis. Its definition rests on two pillars: 1) empirical evidence from experiments in multiple species, and 2) in silico inference based on evolutionary conservation. Regulatory demand for a "defined" tDOA means moving from vague statements (e.g., "likely applicable to vertebrates") to a well-justified, evidence-based taxonomic scope.

Table 1: Core Concepts in tDOA Definition for KERs

Concept	Definition	Role in NGRA	Source of Evidence
Molecular Initiating Event (MIE)	The initial interaction between a stressor and a biomolecule within an organism.	Starting point for mechanistic prediction; high conservation increases tDOA breadth.	In vitro binding/activity assays, structural biology.
Key Event Relationship (KER)	A scientifically supported causal or associative link between two Key Events.	Core unit of predictive knowledge; the primary entity for tDOA assessment.	Empirical dose-response/temporal data, mechanistic studies.
Taxonomic Domain of Applicability (tDOA)	The range of taxa for which a KER is considered biologically plausible.	Justifies extrapolation of NAM data from test systems to target species (human/wildlife).	Cross-species empirical data, in silico sequence/pathway conservation analysis.
Empirical tDOA	tDOA based on direct experimental observation of the KER in listed species.	Provides highest confidence but is limited by the scope of tested species.	Published in vivo or in vitro studies across multiple taxa.
Inferred tDOA	tDOA extrapolated using computational tools analyzing evolutionary conservation.	Enables expansion of tDOA beyond empirically tested species; essential for broad screening.	SeqAPASS, G2P-SCAN, phylogenetic analysis.

The regulatory driver for defined tDOA is clear: without it, the use of an AOP for decision-making lacks a defined boundary of relevance, introducing unacceptable uncertainty. For instance, an AOP for reproductive toxicity developed in nematodes must have its tDOA rigorously defined to assess its utility for predicting risk to mammals or fish [15].

Experimental & Computational Protocols for tDOA Determination

Defining tDOA is a multi-step process that integrates data curation, empirical analysis, and computational prediction. The following protocols, drawn from a seminal case study on extending the tDOA for an AOP involving silver nanoparticle (AgNP)-induced reproductive toxicity, provide a replicable blueprint [15].

Protocol I: Building a Cross-Species AOP Network and Quantifying KERs

Objective: To assemble empirical evidence from multiple species into a unified AOP network and quantitatively assess the confidence in each KER.

Workflow:

• Data Collection and Curation: Systematically gather literature where the stressor of interest (e.g., AgNPs) has been studied in various model systems. For each study, extract relevant endpoints (e.g., ROS production, gene expression changes, apical outcomes like reproduction) and map them to standardized KE terms from the AOP-Wiki [15].
• AOP Network Assembly: Construct a network where KEs are nodes and evidence-supported KERs are directed edges. This network will integrate data from different species and biological levels (e.g., human cell in vitro, invertebrate in vivo).
• Key Event Relationship Quantification via Bayesian Network (BN) Modeling:
  • Rationale: A probabilistic BN approach is preferred for quantifying KERs as it handles biological variability and uncertainty inherent in cross-species data better than deterministic regression models [15].
  • Method: Translate the qualitative AOP network into a BN structure. Use conditional probability tables to represent the strength and uncertainty of each KER (e.g., the probability of observing KE "Y" given the state of KE "X"). Parameterize these tables using the collected empirical dose-response and temporal data.
  • Output: A quantitative model that provides confidence metrics for each KER and allows for probabilistic prediction of AOs based on MIEs. This model validates the overall coherence of the assembled cross-species evidence.

Table 2: Example Dataset for Cross-Species AOP Network Construction [15]

Ecological Compartment	Initial Empirically Tested Species	Number of Studies Integrated	Key MIE/KEs Mapped
Terrestrial	Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens (in vitro)	17	ROS generation, MAPK activation, oxidative damage, reproductive output.
Aquatic	Chironomus riparius, Daphnia magna, Oryzias latipes	8	Oxidative stress, genotoxicity, growth inhibition, mortality.

Protocol II:In SilicoExtension of Biologically Plausible tDOA

Objective: To extrapolate the tDOA of KERs beyond empirically tested species using computational tools that assess the conservation of molecular targets and pathways.

Workflow:

• Identify Molecular Targets: For each KER in the network, define the essential proteins or functional domains involved (e.g., the specific NADPH oxidase complex for an MIE of "ROS production").
• Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS):
  • Tool: A web-based platform from the US EPA.
  • Method: Input the amino acid sequence or functional domain of the primary molecular target (e.g., a key enzyme in the MIE). SeqAPASS performs pairwise alignment against the proteomes of species in its database, generating a similarity score [15].
  • Analysis: Establish a threshold of conservation (e.g., >80% sequence similarity in the functional domain). Species surpassing this threshold are considered to have a "highly plausible" tDOA for that MIE or KE.
• Genes-to-Pathways Species Conservation Analysis (G2P-SCAN):
  • Tool: An R package designed to assess the conservation of entire biological pathways [15].
  • Method: Input a set of genes representing a core pathway linking two KEs (e.g., the p38 MAPK signaling pathway). G2P-SCAN evaluates the presence, completeness, and orthology of these genes across a wide taxonomic range.
  • Analysis: Determine if the pathway is functionally complete in a given taxon. This provides stronger evidence for tDOA than single-target conservation, as it confirms the biological context for the KER to operate.
• Integrated tDOA Conclusion: Synthesize results from SeqAPASS (target conservation) and G2P-SCAN (pathway conservation) with the empirical BN model. The final tDOA for a KER is defined as the union of taxa with empirical support and taxa for which strong in silico evidence predicts biological plausibility.

Table 3: In Silico Tools for tDOA Extension

Tool	Primary Function	Input	Output for tDOA	Regulatory Relevance
SeqAPASS	Protein sequence/functional domain conservation analysis.	Amino acid sequence or PFAM domain.	List of species with conserved molecular target; informs on MIE/KE applicability.	Justifies extrapolation of molecular interactions. Supported by US EPA.
G2P-SCAN	Biological pathway and gene set conservation analysis.	Set of human genes (Entrez IDs) representing a pathway.	Assessment of pathway completeness across >400 species.	Confirms functional biological context for KERs, strengthening tDOA.
Bayesian Network Modeling	Probabilistic quantification of KER strength and uncertainty.	Empirical dose-response and temporal data for KEs.	Conditional probability tables for KERs; validates network logic.	Provides transparent, quantitative confidence metrics for use in weight-of-evidence assessments.

Visualizing the Workflow and KER-tDOA Integration

The following diagrams, created using Graphviz DOT language, illustrate the core workflows and conceptual relationships described in this guide.

Diagram 1: Workflow for Defining KER tDOA in NGRA

Diagram 2: Structure of a Quantitative AOP Network with KERs

Diagram 3: Extending tDOA via In Silico Conservation Analysis

Table 4: Research Toolkit for KER and tDOA Investigation

Category	Item/Resource	Function in tDOA Research	Example/Supplier
Data & Knowledge Bases	AOP-Wiki (aopwiki.org)	Central repository for curated AOPs, KEs, and KERs; provides standardized ontology for mapping data.	OECD-hosted database.
	US EPA SeqAPASS Tool	Web-based platform for performing protein sequence conservation analysis to predict molecular target applicability.	https://seqapass.epa.gov/seqapass/
	G2P-SCAN R Package	Tool for assessing conservation of human biological pathways across a wide range of species.	Available via Bioconductor.
Software & Modeling	Bayesian Network Software (e.g., Netica, AgenaRisk, R packages bnlearn, gRain)	Platform for constructing, parameterizing, and performing probabilistic inference on quantitative KER/AOP network models.	Commercial and open-source options.
	Molecular Visualization & Alignment Software (e.g., PyMOL, Clustal Omega)	Visualizes protein structures and performs multiple sequence alignments to support SeqAPASS analysis and threshold determination.
Experimental Models	Phylogenetically Diverse Model Organisms	Provide empirical data for KER validation across taxa (e.g., C. elegans, D. rerio, X. laevis).	Strain centers and commercial suppliers.
	Recombinant Proteins & Cell Lines	Express conserved molecular targets from human and non-human species for in vitro comparative assays to validate conservation predictions.	Commercial cDNA clones, ATCC cell lines.
Reference Materials	Chemical Stressors with Known MIEs (e.g., reference agonists/antagonists)	Positive controls for establishing KERs in novel test systems (e.g., a specific kinase inhibitor).	Sigma-Aldrich, Tocris.
	Conserved Pathway Antibodies	Immunodetection tools that cross-react with orthologous proteins in multiple species, enabling comparative KE measurement.	Commercial antibody suppliers with cross-reactivity data.

The regulatory demand for defined tDOA is not an abstract scientific ideal but a practical necessity for the implementation of NGRA. It provides the scientific boundary conditions for applying mechanistic, NAM-derived data to protect human health and the environment. As demonstrated, defining tDOA is a rigorous process that moves from curated empirical evidence to quantitative KER modeling and finally to computational inference of conservation. The integration of tools like SeqAPASS and G2P-SCAN with probabilistic AOP networks represents the cutting edge of this field, enabling scientists to confidently extrapolate mechanistic toxicological knowledge across the tree of life [15]. For researchers and drug development professionals, mastering these protocols is essential for building NGRA packages that meet the evolving standards of regulatory agencies, ultimately supporting the transition to a more predictive, efficient, and animal-free safety assessment paradigm.

Methodological Toolkit: Bioinformatics and Computational Strategies for tDOA Expansion

The Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool is a computational framework developed by the U.S. Environmental Protection Agency (EPA) to address a central challenge in toxicology and chemical safety: predicting chemical susceptibility across thousands of species for which empirical toxicity data are absent [16] [17]. In the context of research on Key Event Relationships (KER) within the Adverse Outcome Pathway (AOP) framework, SeqAPASS provides a critical methodology for assessing the taxonomic conservation of Molecular Initiating Events (MIEs). The tool operates on the principle that a species' intrinsic susceptibility to a chemical is largely determined by the conservation of the protein targets with which that chemical interacts [16] [18].

By leveraging publicly available protein databases, SeqAPASS allows researchers to extrapolate known chemical-protein interactions from well-studied model organisms (e.g., humans, rats, zebrafish) to non-target species, including plants, wildlife, and endangered species [16]. This capability is increasingly vital in a regulatory landscape moving towards New Approach Methodologies (NAMs) that reduce reliance on whole-animal testing, while simultaneously demanding broader ecological risk assessments [17] [19]. The tool's hierarchical design, which progresses from primary sequence to three-dimensional structural analysis, offers a tiered weight-of-evidence approach for evaluating protein conservation, making it a powerful asset for defining the domain of applicability for KERs across taxa and strengthening the scientific basis for cross-species extrapolation in modern toxicology [18] [19].

The Hierarchical Analytical Framework of SeqAPASS

SeqAPASS is structured around a multi-level analytical hierarchy, each level providing an increasingly refined line of evidence regarding protein conservation and predicted chemical susceptibility [20]. This design allows users to tailor the analysis based on the depth of available knowledge about the chemical-protein interaction of interest.

Level 1: Primary Amino Acid Sequence Comparison

This foundational level performs a whole-protein sequence alignment using the Basic Local Alignment Search Tool (BLASTp) algorithm. It compares the primary amino acid sequence of a query protein from a known sensitive species against sequences from all species within the National Center for Biotechnology Information (NCBI) protein database [17]. The result is a broad prediction of potential susceptibility across species, based on overall sequence similarity. This level is most useful for initial screening when detailed knowledge of the protein's functional domains or chemical interaction residues is limited [20].

Level 2: Functional Domain Comparison

Level 2 refines the analysis by focusing on conserved functional domains. Using tools like COBALT for multiple sequence alignment, this level evaluates whether the specific domains responsible for a protein's function (e.g., a ligand-binding domain, an active site) are preserved in other species [17]. Conservation of these domains suggests the protein's core function is retained, providing stronger evidence for a conserved chemical interaction than whole-sequence similarity alone.

Level 3: Critical Amino Acid Residue Comparison

The most sequence-specific level involves evaluating the conservation of individual critical amino acid residues known to be essential for the chemical-protein interaction [17] [20]. Users input the specific residue positions (e.g., from a solved crystal structure or site-directed mutagenesis studies). SeqAPASS then checks for their preservation across species. A match at these precise locations offers high-confidence evidence that the specific molecular interaction is conserved, even if other parts of the protein sequence vary.

Level 4: Protein Structural Evaluation (Advanced)

Introduced in SeqAPASS v7.0, Level 4 is an advanced feature that generates protein structural models for cross-species comparison [21]. It employs the Iterative Threading ASSEmbly Refinement (I-TASSER) tool to create 3D structural predictions from amino acid sequences [19]. Users can then align these predicted structures to a reference structure to assess structural conservation. This level provides a direct, biophysical line of evidence and yields models suitable for downstream applications like molecular docking or dynamics simulations [22] [19].

The following diagram illustrates this hierarchical workflow and its role in supporting KER-based research.

Core Methodologies and Experimental Protocols

Protocol for Executing a SeqAPASS Analysis

The standard workflow for a SeqAPASS analysis, as detailed in peer-reviewed protocols, involves the following steps [17]:

• Account Creation and Login: Access the tool via the EPA SeqAPASS website using a Chrome browser. Researchers must create a free account to run, store, and customize jobs.
• Query Definition: Identify a protein target and a sensitive "query" species (e.g., human estrogen receptor alpha) through literature review or using integrated links to resources like the CompTox Chemicals Dashboard or AOP-Wiki.
• Job Submission (Level 1):
  • On the request page, enter the query protein using its NCBI protein accession number or FASTA sequence.
  • Select the query species and configure analysis parameters (e.g., expectation value (E-value) threshold for BLAST).
  • Submit the job. The backend system mines the NCBI database (containing over 153 million proteins from >95,000 organisms) and performs the BLASTp alignment [16].
• Interpretation and Refinement (Levels 2 & 3):
  • After reviewing Level 1 results (displayed as a table and interactive taxonomy plot), users can proceed to Level 2.
  • For Level 2, the tool automatically extracts functional domain information from the Conserved Domain Database (CDD). Users review multiple sequence alignments of these domains.
  • For Level 3, users input the positions of critical amino acid residues. SeqAPASS extracts these residues from the alignment for all species and displays results in a customizable heatmap, predicting susceptibility based on exact residue matching [17].
• Data Synthesis and Export: Utilize built-in visualization tools (box plots, heatmaps) and generate a downloadable Decision Summary Report (.pdf) that synthesizes evidence from all levels for publication or assessment purposes [17].

Integrated Workflow for Structural Analysis (Level 4)

For advanced users, the Level 4 workflow extends SeqAPASS into structural bioinformatics [22] [19]:

• Structure Generation: Using protein sequences from a SeqAPASS output (e.g., species predicted as susceptible), generate 3D structural models with I-TASSER or import structures from AlphaFold or the Protein Data Bank (PDB) [21] [19].
• Structural Alignment: Align the generated models to a reference protein structure (e.g., a solved structure with a bound ligand) using alignment algorithms like TM-align to assess structural conservation [19].
• Downstream Molecular Modeling: Export structures for molecular docking to predict binding poses and affinities, or for molecular dynamics (MD) simulations to study the stability and quantitative dynamics of the chemical-protein interaction across species [22].

The diagram below integrates this structural workflow with the core SeqAPASS hierarchy.

Application in KER and Taxonomic Conservation Research

SeqAPASS is explicitly designed to evaluate the conservation of Molecular Initiating Events (MIEs)—a specific type of KER—across species, thereby defining the taxonomic applicability of an AOP [18]. For instance, if an MIE is defined as "Chemical X binding to Androgen Receptor (AR) leading to antagonism," SeqAPASS can predict which vertebrate species possess a conserved AR ligand-binding domain, suggesting they are potentially susceptible to this MIE.

This application is demonstrated in several published case studies:

• Endocrine Disruption: Scientists used SeqAPASS to translate data on chemical activation of the human estrogen receptor to non-mammalian vertebrates (fish, amphibians, birds), aiding the EPA's Endocrine Disruptor Screening Program in prioritizing testing for thousands of chemicals [16].
• Insecticide Specificity: The tool was used to compare the ecdysone receptor in the tobacco budworm (a target pest) to non-target species like honey bees and earthworms. The analysis helped explain species-specific toxicity by revealing sequence and structural differences in the receptor [16].
• Pollinator Risk Assessment: SeqAPASS evaluated conservation of the nicotinic acetylcholine receptor in honey bees and other insects to predict potential chemical susceptibility implicated in colony decline [16].
• PFAS Research: A recent study integrated SeqAPASS with MD simulations to investigate the interaction of perfluorooctanoic acid (PFOA) with transthyretin (TTR) across species. SeqAPASS identified hundreds of species with conserved TTR sequences, and subsequent simulations provided quantitative evidence that key binding interactions were conserved across vertebrate groups [22].

Performance Data and Tool Evolution

Since its initial release in 2016, SeqAPASS has undergone significant feature enhancements, driven by user feedback and advances in bioinformatics [17]. The following table summarizes its version history and key developments.

Table 1: Evolution of the SeqAPASS Tool and Its Capabilities [17]

Version	Release Date	Key Features and Updates
1.0	Jan 2016	Initial public release with Levels 1 & 2 (primary sequence and domain comparison).
2.0	May 2017	Introduction of Level 3 for critical amino acid residue comparison.
3.0	Mar 2018	Added interactive data visualization capabilities for Levels 1 & 2.
4.0	Oct 2019	Enhanced interoperability: links to ECOTOX Knowledgebase, AOP-Wiki, and summary reports.
5.0	Dec 2020	Introduced customizable heatmaps for Level 3 and a downloadable Decision Summary Report.
6.0	Sep 2021	Added widget to directly query the ECOTOX Knowledgebase from SeqAPASS results.
7.0	Sep 2023	Introduced Level 4 for protein structural evaluation using I-TASSER [21].
8.0	2025	Current version; allows submission of sequences to generate protein structures across species [16].

The tool's performance is benchmarked by its ability to efficiently process massive datasets. A single Level 1 query compares the query sequence against the entire NCBI protein database, facilitating predictions for thousands of species within a short timeframe [16]. Quantitative outputs from case studies demonstrate its predictive scale. For example, in the PFOA-TTR study, SeqAPASS predicted 952, 976, and 750 species as susceptible at Levels 1, 2, and 3, respectively [22]. The integration of structural modeling (Level 4) and MD simulations provided quantitative binding metrics (e.g., binding free energy, root-mean-square deviation) that confirmed the interaction's conservation across a subset of these species, validating the sequence-based predictions with biophysical data [22].

Table 2: Example Quantitative Output from a SeqAPASS Case Study (PFOA-TTR Interaction) [22]

Analysis Level	Number of Species Predicted as Susceptible	Key Output Metrics
Level 1	952	Primary sequence similarity threshold met.
Level 2	976	Functional TTR domains (binding pockets) conserved.
Level 3	750	Critical lysine residue (K15) for PFOA binding conserved.
Level 4 (MD Simulation)	Subset of above	Quantitative confirmation: No significant difference in predicted binding affinity (ΔG) or interaction stability across tested vertebrate species.

Successful application of the SeqAPASS framework, especially in advanced workflows, relies on a suite of integrated databases and computational tools.

Table 3: Key Research Reagent Solutions for SeqAPASS-Based Analysis

Item / Resource	Primary Function in SeqAPASS Workflow	Source / Availability
NCBI Protein Database	The primary source repository for protein sequence data against which all SeqAPASS queries are compared. Contains over 153 million sequences [16].	Publicly available from the National Library of Medicine.
BLASTp Algorithm	Executes the primary amino acid sequence alignments for Level 1 analysis, identifying homologous sequences across species [17].	Integrated into the SeqAPASS backend.
Conserved Domain Database (CDD)	Provides curated information on protein functional domains used for Level 2 comparative analysis [17].	Integrated into SeqAPASS from NCBI.
I-TASSER (Iterative Threading ASSEmbly Refinement)	The primary engine for de novo protein structure prediction from amino acid sequences in Level 4 analysis [21] [19].	Open-source, publicly available tool.
AlphaFold Protein Structure Database	A source of highly accurate, pre-computed protein structure predictions that can be imported into SeqAPASS for Level 4 structural comparisons [19].	Publicly available from DeepMind/EMBL-EBI.
Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB)	The global archive for experimentally determined 3D structures of proteins. Provides reference structures for alignment, docking, and simulation studies [21].	Publicly available.
Molecular Docking Software (e.g., AutoDock, GOLD)	Used downstream of SeqAPASS to predict the binding orientation and affinity of a chemical to protein models generated from diverse species [22] [19].	Various commercial and open-source packages.
Molecular Dynamics (MD) Simulation Software (e.g., GROMACS, AMBER)	Used to run physics-based simulations that quantify the stability and dynamics of chemical-protein complexes across species, providing free energy calculations and residue interaction profiles [22].	Various commercial and open-source packages.

The SeqAPASS tool represents a mature and critically important hierarchical framework for bridging comparative genomics and predictive toxicology. By providing a systematic, publicly accessible method to evaluate protein conservation from sequence to structure, it directly addresses the core challenge of taxonomic applicability in KER and AOP research. Its integration with molecular modeling and simulation techniques marks the frontier of next-generation risk assessment, moving beyond qualitative predictions to generate quantitative, biophysical lines of evidence for cross-species extrapolation [22] [19].

Future development of SeqAPASS will likely focus on deeper automation of the advanced workflow, potentially embedding simplified docking or simulation modules within the web interface. Furthermore, as databases like AlphaFold continue to expand the universe of available protein structures, the accuracy and ease of Level 4 structural comparisons will improve dramatically. For researchers and drug development professionals, SeqAPASS offers a powerful, validated platform to efficiently assess potential chemical risks across the tree of life, prioritize testing, and ultimately build more credible and defensible safety assessments for both human health and ecological systems.

G2P-SCAN (Genes-to-Pathways Species Conservation Analysis) is a computational pipeline designed to assess the conservation of human biological pathways across multiple species by integrating data on gene orthologs, protein families, and pathway entities [23]. This tool addresses a critical need in modern toxicology and safety assessment: the ability to extrapolate biological effects and chemical susceptibility across species with confidence [23] [24]. Its development is a direct response to the global regulatory shift towards New Approach Methodologies (NAMs) that reduce reliance on animal testing [23].

The operational context for G2P-SCAN is firmly embedded within the Adverse Outcome Pathway (AOP) framework, which structures toxicological knowledge into causal sequences from a Molecular Initiating Event (MIE) to an Adverse Outcome (AO) [10]. The core, inferential units of an AOP are Key Event Relationships (KERs), which describe the biologically plausible and empirically supported causal link between an upstream and a downstream Key Event (KE) [11] [10]. A fundamental challenge in AOP development and application is defining its Taxonomic Domain of Applicability (tDOA)—the range of species for which the described causal pathway is biologically plausible [10] [24].

G2P-SCAN directly informs KER taxonomic conservation research by providing a systematic, evidence-based method to evaluate whether the molecular and cellular processes underpinning a KER are conserved in non-human species. For example, a KER linking decreased retinoic acid levels to disrupted meiosis in oocytes relies on a conserved signaling pathway [11]. G2P-SCAN can analyze the core genes in this pathway (e.g., ALDH1A1, STRA8) to objectively assess its conservation in model organisms like mouse or zebrafish, thereby strengthening or limiting the tDOA claim for that KER [11] [24]. This moves beyond simple sequence similarity of individual proteins to a functional, pathway-level assessment of conservation, which is more relevant for predicting the propagation of a toxicological perturbation along an AOP [23] [25].

Core Methodology of the G2P-SCAN Pipeline

The G2P-SCAN pipeline is implemented as an R package and functions as a structured workflow that queries, synthesizes, and analyzes data from multiple established biological databases [23] [26].

The primary analytical workflow of G2P-SCAN consists of four main stages: Pathway Mapping, Orthology Identification, Functional Analysis, and Data Synthesis [26].

• Figure 1: G2P-SCAN Core Analytical Workflow. The pipeline processes a set of human input genes through four consecutive stages to produce a quantitative report on pathway conservation across specified species [26].

Stage 1: Human Pathway Mapping

The pipeline begins by mapping the user-provided human gene symbols (e.g., ESR1, PPARA) to biological pathways using the Reactome knowledgebase via the InterMineR API [26]. It retrieves all human pathways containing the input genes. Users can specify the desired level of the Reactome hierarchy for analysis: "terminal" (most specific pathways), "parental" (broad parent pathways), or "intermediate" [26].

Stage 2: Orthology Identification Across Species

For every human gene contained within the mapped pathways, G2P-SCAN identifies orthologous genes in the selected target species. This step also utilizes the InterMineR API to access orthology data [26]. The pipeline supports analysis for six key model organisms: mouse (Mus musculus), rat (Rattus norvegicus), zebrafish (Danio rerio), fruit fly (Drosophila melanogaster), roundworm (Caenorhabditis elegans), and yeast (Saccharomyces cerevisiae) [23] [26]. An orthology filter ("ALL" or "LDO"-Least Divergent Ortholog) can be applied to refine the results [26].

Stage 3: Functional Analysis via Protein Families

To move beyond gene-level lists and assess functional conservation, the pipeline queries the UniProt API to obtain protein identifiers for each human gene and its orthologs [26]. These protein identifiers are then submitted to the InterPro API to map each protein to one or more protein families or domains (e.g., "Nuclear hormone receptor," "Zinc finger") [23] [26]. Protein families serve as a proxy for conserved functional units, offering a more biologically meaningful metric than gene counts alone [23].

Stage 4: Data Synthesis and Output

In the final stage, G2P-SCAN compiles multiple quantitative metrics for each pathway and species:

• Gene & Protein Counts: The number of human genes and their identified orthologs/proteins.
• Protein Family Coverage: Counts of assigned and unassigned protein families.
• Pathway Entity & Reaction Counts: Utilizing pre-fetched data from Reactome, the pipeline extracts the counts of molecular entities (proteins, complexes, chemicals) and reactions (biochemical events) that are annotated for a given pathway in each species [26].

All results are organized into two primary outputs: a "counts" summary (tabular quantitative data) and a "data" file (the underlying gene, protein, and family lists) [26].

Quantitative Data on Pathway Conservation

Application of G2P-SCAN provides multi-dimensional metrics for evaluating conservation. The following table summarizes hypothetical output for two pathways central to case studies in the literature [24].

Table 1: Example G2P-SCAN Conservation Metrics for Selected Pathways and Species

Pathway (Human Input Gene)	Species	Human Genes in Pathway	Ortholog Count	Proteins Mapped	Protein Families Assigned	Pathway Entities (Species)	Pathway Reactions (Species)
Estrogen Signaling (ESR1)	Homo sapiens (Ref)	15	15 (Self)	15	22	150	95
	Mus musculus	15	15	15	22	148	94
	Danio rerio	15	14	13	18	132	82
	Drosophila melanogaster	15	1	1	1	25	10
PPARα Signaling (PPARA)	Homo sapiens (Ref)	12	12 (Self)	12	18	110	70
	Rattus norvegicus	12	12	12	18	109	69
	Danio rerio	12	11	10	16	105	65
	Caenorhabditis elegans	12	0	0	0	15	5

Note: Data is illustrative, based on case study descriptions [24]. Actual counts vary by database version and analysis parameters.

The power of G2P-SCAN is amplified when integrated with complementary tools. A pivotal study combined G2P-SCAN with the SeqAPASS (Sequence Alignment to Predict Across Species Susceptibility) tool [24] [25]. SeqAPASS performs deep sequence and structural analysis of specific molecular targets (like a receptor) to predict cross-species chemical susceptibility [24] [25]. The integration provides a weight-of-evidence approach: SeqAPASS assesses the conservation of the Molecular Initiating Event (MIE) target, while G2P-SCAN evaluates the conservation of the downstream cellular pathway context [24] [25].

Table 2: Combined G2P-SCAN and SeqAPASS Analysis for Chemical Targetsa

Chemical Class	Primary Molecular Target (MIE)	SeqAPASS Prediction (Target Conservation)	G2P-SCAN Analysis (Pathway Conservation)	Enhanced Inference for AOP tDOA
Fibrate drugs	PPARα	High confidence in mammals; moderate in zebrafish; low in invertebrates.	PPAR signaling pathway largely conserved in mammals & zebrafish; fragmented in Drosophila; absent in C. elegans.	Strong support for mammal & fish tDOA; suggests limited applicability to invertebrates.
Environmental Estrogens	Estrogen Receptor (ESR1)	High confidence in vertebrates; no orthologs identified in insects or nematodes.	Core estrogen signaling pathway conserved in vertebrates; highly divergent in Drosophila.	Corroborates vertebrate-specific tDOA for ESR1-mediated AOPs.
Pyrethroid insecticides	GABA-A Receptor (GABRA1)	Subunit orthologs present in insects, fish, and mammals with varying sequence similarity.	GABA receptor signaling & neurotoxicity pathways show modular conservation across taxa.	Supports broad tDOA but highlights potential for species-specific differences in sensitivity.

a Based on combined methodology described in [24] [25].

• Figure 2: Integrated Workflow for KER Taxonomic Assessment. Combining target-specific (SeqAPASS) and pathway-centric (G2P-SCAN) analyses generates multiple lines of evidence to support definitive inferences about the taxonomic domain of applicability for an AOP's Key Event Relationships [24] [25].

Experimental Protocol: Implementing a G2P-SCAN Analysis

This section provides a detailed protocol for executing a G2P-SCAN analysis using the R package, based on the official documentation and case studies [23] [26].

Software Installation and Setup

• Prerequisites: Ensure R (≥4.0.0) and RStudio are installed. The devtools package is required for installation.

• Install G2P-SCAN: Install the package directly from GitHub using the following R commands:

• Load Libraries: Load the G2P-SCAN package and the parallel package to enable faster processing.

Executing the Core Analysis

The primary wrapper function runGenes2Pathways() executes the entire pipeline. Below is an annotated example call analyzing the acetylcholinesterase genes ACHE and BCHE.

Output Interpretation and Downstream Analysis

• Output Files: The function creates two Excel files in the specified outputDir:
  • [prefix]_counts.xlsx: Contains quantitative summary tables.
  • [prefix]_data.xlsx: Contains the underlying lists of genes, orthologs, proteins, and families.
• Key Results Object: The function also returns a list object (results in the example) containing all structured data for programmatic access in R (e.g., results$all_counts).
• Conservation Assessment: Analyze the countSummary tab. High conservation for a pathway in a species is indicated by: a high proportion of human genes with orthologs, high protein family assignment overlap, and significant counts of conserved pathway entities and reactions (see Table 1 for example).
• Integration with AOP/KER Context: Map the analyzed pathways and their conservation metrics back to the specific Key Events in the AOP of interest. This provides empirical evidence to support, refine, or limit the proposed taxonomic domain of applicability for the KERs [11] [24].

Table 3: Key Computational Tools and Databases for Pathway Conservation Research

Tool/Resource Name	Type	Primary Function in Conservation Analysis	Relevance to G2P-SCAN/KER Research
G2P-SCAN R Package [23] [26]	Computational Pipeline	Core tool for integrated pathway-to-orthology analysis from human gene sets.	Directly executes the analysis framework described in this guide.
SeqAPASS [24] [25]	Computational Tool (Web-based)	Predicts chemical susceptibility across species via protein sequence/structure analysis of specific molecular targets.	Provides complementary, target-specific evidence to combine with G2P-SCAN's pathway evidence for robust tDOA assessment.
Reactome [23] [26]	Biological Pathway Knowledgebase	Provides curated human pathway data and orthology-projected pathway data for other species.	Primary source for pathway mapping and entity/reaction counts in G2P-SCAN.
InterPro [23] [26]	Protein Family/ Domain Database	Classifies proteins into families and domains based on sequence signatures.	Source for functional protein family assignments, a key conservation metric in G2P-SCAN.
UniProt [26]	Protein Sequence/Annotation Database	Provides authoritative protein identifiers and functional annotations.	Critical for accurately linking genes to protein sequences for subsequent family analysis.
AOP-Wiki [11] [10]	Knowledgebase	Central repository for published Adverse Outcome Pathways, Key Events, and Key Event Relationships.	Source of biological hypotheses (KERs) to be tested for taxonomic applicability using G2P-SCAN.
Orthology Data (via InterMine) [26]	Data Resource	Provides pre-computed orthology relationships across multiple species.	Foundational data source for the orthology identification step in the G2P-SCAN pipeline.

The assessment of chemical safety for both human and ecological health fundamentally depends on the accurate extrapolation of toxicological effects across diverse species. This process is central to the Key Event Relationship (KER) taxonomic conservation research within the Adverse Outcome Pathway (AOP) framework, which seeks to define the taxonomic domain of applicability (tDOA) for mechanistic toxicity pathways [9]. Historical reliance on in vivo vertebrate testing presents significant ethical, resource, and time constraints, creating an urgent need for efficient, non-animal New Approach Methodologies (NAMs) [9].

This guide details the synergistic integration of two pivotal computational NAMs: the U.S. EPA's Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool and Unilever's Genes to Pathways – Species Conservation Analysis (G2P-SCAN) tool [9] [25]. While SeqAPASS evaluates the conservation of primary protein targets (often Molecular Initiating Events) across species using sequence and structural similarity, G2P-SCAN analyzes the broader conservation of entire biological pathways triggered by chemical interaction [9] [23]. When used in combination, these tools generate complementary lines of evidence that significantly strengthen the weight of evidence (WoE) for cross-species extrapolation, thereby refining the tDOA of AOPs and supporting more confident safety decisions in chemical and pharmaceutical development [9].

Table 1: Core Tool Comparison for KER Conservation Analysis

Feature	SeqAPASS	G2P-SCAN
Primary Purpose	Predict intrinsic susceptibility by assessing conservation of specific protein targets/MIEs [18].	Infer biological pathway conservation from human gene sets across model species [23].
Core Methodology	Tiered sequence and structural alignment (Levels 1-4) [18].	Orthology mapping and pathway enrichment analysis using Reactome [9] [23].
Taxonomic Scope	Broad; any species with available protein sequence data [9].	Focused on 7 key model species: Human, Mouse, Rat, Zebrafish, Fruit Fly, Roundworm, Yeast [9].
Output for WoE	Binary susceptibility calls & structural models for molecular docking [9] [27].	Pathway conservation scores and lists of conserved/non-conserved reactions [9] [23].
Role in AOP/tDOA	Defines conservation of the MIE (AOP Key Event 1) [9].	Informs conservation of downstream key events and the overall pathway [9].

Foundational Tools: SeqAPASS and G2P-SCAN

The integrated workflow is built upon the distinct yet complementary functions of its two core tools.

SeqAPASS operates on the principle that susceptibility to a chemical is conferred by the presence and structural similarity of a specific molecular target. Its analysis progresses through four tiers:

• Level 1 (Primary Sequence Comparison): Assesses full-length protein sequence identity.
• Level 2 (Functional Domain Comparison): Focuses on the conservation of specific domains (e.g., a ligand-binding domain).
• Level 3 (Critical Amino Acid Residue Comparison): Evaluates the presence of amino acids known to be critical for chemical-protein interaction or protein function.
• Level 4 (3D Protein Structure Modeling): Generates and compares predicted protein structures across species using tools like I-TASSER [27]. This level provides a functional context for susceptibility predictions and enables advanced applications like cross-species molecular docking [27].

G2P-SCAN functions as an R package that translates a set of human genes (e.g., a ToxCast assay target or an AOP key event) into pathway-level information. It maps human genes to their orthologs in six other model species, retrieves associated biological pathways from the Reactome database, and performs a species conservation analysis for each pathway [9] [23]. Its output indicates whether the entire pathway, specific reactions within it, or the involved protein families are conserved, offering a higher-order biological context beyond the single protein target.

Diagram 1: Integrated SeqAPASS & G2P-SCAN Workflow for KER Conservation (760px)

Integrated Methodological Framework

The following protocol describes the stepwise integration of SeqAPASS and G2P-SCAN to build a WoE for KER taxonomic conservation.

Step 1: Molecular Target Identification. For the chemical of interest, identify its primary protein molecular initiating event (MIE) and associated human gene(s). Sources include:

• EPA's RefChemDB & ToxCast Data: For bioactivity data from high-throughput screening [9].
• RCSB Protein Data Bank (PDB): To find experimentally derived chemical-protein co-crystal structures [9] [27].
• Literature Review: Manual searches using gene names, chemical identifiers, and terms like "mechanism of action" [9].

Step 2: SeqAPASS Analysis for Target Conservation.

• Input the primary amino acid sequence of the human reference protein (e.g., NCBI Accession) into SeqAPASS (v8.0 or later) [18].
• Execute a Level 1 evaluation to obtain a preliminary list of species with a homologous protein.
• Conduct a Level 2 evaluation focusing on the relevant functional domain (e.g., ligand-binding domain for a receptor).
• Perform a Level 3 evaluation against known critical amino acid residues for chemical binding or protein function.
• (Optional) Generate Level 4 predicted 3D structural models for species of particular interest to enable deeper functional analysis [27].

Step 3: G2P-SCAN Analysis for Pathway Conservation.

• Prepare the input as a list of human gene symbols corresponding to the MIE and/or downstream key events.
• Run the G2P-SCAN R pipeline. The tool will:
  • Map human genes to orthologs in the six other model species.
  • Retrieve related biological pathways and reactions from Reactome.
  • Calculate a "Pathway Conservation Score" for each pathway, indicating the percentage of reactions that are plausibly conserved in each species [23].
• Interpret output tables and scores to determine if the perturbed biological pathway is conserved in the species of concern.

Step 4: WoE Integration & tDOA Refinement.

• Concordant Evidence (High Confidence): SeqAPASS predicts susceptibility (conserved target) and G2P-SCAN indicates high pathway conservation. This strongly supports the expansion of the AOP's tDOA to include that species [9].
• Discordant Evidence (Requires Scrutiny): SeqAPASS predicts susceptibility but G2P-SCAN shows low pathway conservation (or vice versa). This indicates a need for further investigation—the chemical may bind the target but not elicit downstream effects, or effects may proceed via an alternative pathway.
• Integrate these computational results with other existing lines of evidence (e.g., in vitro assay data, in silico molecular docking results [27]) to reach a final, confidence-rated conclusion on taxonomic applicability.

Table 2: Summary of Quantitative Outcomes from Case Studies [9]

Case Study Target	Example Chemical	SeqAPASS Prediction	G2P-SCAN Pathway Conservation Insight	Integrated WoE Conclusion
PPARα (Peroxisome Proliferator-Activated Receptor Alpha)	Various fibrates	High predicted susceptibility across mammals; variable in fish.	PPARα activation pathway reactions highly conserved in mammals, partially conserved in zebrafish.	Strong WoE for AOP applicability in mammals. Limited, plausible WoE for zebrafish requiring further investigation.
ESR1 (Estrogen Receptor 1)	Oxybenzone, Butylparaben	High conservation of ligand-binding domain across vertebrates.	Estrogen signaling pathway highly conserved in vertebrates; metabolism reactions show species-specific differences.	Supports tDOA for MIEs and early KERs across vertebrates. Downstream outcomes may vary due to metabolic differences.
GABRA1 (GABA-A Receptor)	Muscimol, Fipronil	Critical neurotransmitter-binding residues conserved from humans to insects.	GABA receptor activation pathway and neural signal transmission broadly conserved across bilaterians.	Provides strong mechanistic WoE for neurotoxic AOPs across a very broad taxonomic range.

The Scientist's Toolkit: Essential Research Reagent Solutions

Table 3: Key Reagents, Databases, and Tools for Integrated Analysis

Item Name	Type	Primary Function in Workflow	Access/Source
RefChemDB	Chemical Database	Provides curated in vitro bioactivity data for target identification [9].	US EPA
ToxCast/Tox21 Data	Bioactivity Database	Supplies high-throughput screening data for chemical-protein interactions [9].	US EPA CompTox Dashboard
RCSB Protein Data Bank (PDB)	Structural Database	Source of experimental protein-chemical co-crystal structures for critical residue identification and docking reference [9] [27].	www.rcsb.org
SeqAPASS v8.0	Computational Tool	Performs cross-species sequence/structure alignment to predict protein target conservation and susceptibility [18].	EPA SeqAPASS Web Tool
G2P-SCAN R Package	Computational Tool	Analyzes conservation of biological pathways from human gene sets across model species [23].	R Package (Publication [23])
Reactome Database	Pathway Database	Provides curated biological pathways used by G2P-SCAN for conservation analysis [9] [23].	reactome.org
AOP-Wiki	Knowledgebase	Framework for organizing KERs and defining the initial tDOA for assessment [9].	aopwiki.org
I-TASSER/AlphaFold	Modeling Tool	Used within or alongside SeqAPASS for predicting 3D protein structures in Level 4 analysis [27].	Standalone Servers

Discussion: Implications for KER Research and NGRA

The combined application of SeqAPASS and G2P-SCAN directly addresses critical challenges in KER taxonomic conservation research. By simultaneously evaluating the conservation of the initial molecular target and the broader biological pathway, this approach moves beyond assumptions based solely on phylogenetic relatedness. It enables the generation of mechanistically grounded hypotheses about which species are likely to experience adverse outcomes along a defined AOP, thereby making the tDOA more biologically plausible and defensible [9].

This methodology is a cornerstone for the emerging Next Generation Risk Assessment (NGRA) paradigm. It exemplifies how multiple computational NAMs can be integrated in a WoE framework to reduce uncertainty and potentially replace animal testing for certain extrapolation questions [9]. Future developments, such as the direct integration of cross-species molecular docking outputs from SeqAPASS Level 4 models [27] with pathway conservation scores from G2P-SCAN, promise to add even deeper layers of functional understanding, further solidifying the role of integrated computational approaches in modern toxicology and drug development.

Systematic Review and Evidence Mapping for Transparent KER Support

This technical guide details a formalized methodology for conducting systematic reviews and evidence mapping to establish transparent support for Key Event Relationships (KERs) within the Adverse Outcome Pathway (AOP) framework. Framed within the critical research objective of defining and expanding the taxonomic domain of applicability (tDOA), this guide provides a step-by-step protocol for integrating diverse data streams—from in vivo and in vitro studies to in silico computational predictions [15]. The core workflow enables researchers to synthesize fragmented toxicological evidence, quantitatively assess KER confidence, and systematically extrapolate pathways across species. This process is foundational for robust, predictive toxicology and chemical safety assessment that aligns with the One Health perspective, bridging human and ecological risk assessment [15].

The Adverse Outcome Pathway framework provides a structured model for describing causal linkages between a Molecular Initiating Event (MIE) and an Adverse Outcome (AO) via intermediate Key Events (KEs) [15]. The scientific confidence in an AOP hinges on the empirical support for each causative Key Event Relationship (KER). However, evidence for KERs is often fragmented across studies employing different model species, experimental designs, and levels of biological organization. This creates significant challenges in assessing the taxonomic domain of applicability (tDOA)—the range of species for which the AOP is biologically plausible [9].

Conducting a systematic review and evidence map is no longer optional but essential for transparent KER support. This process addresses critical gaps and biases analogous to those identified in broader ecological research, such as the over-representation of certain taxa (e.g., vertebrates over invertebrates) and geographic regions [5] [28]. A systematic methodology ensures objectivity, reproducibility, and the identification of true knowledge gaps, thereby preventing skewed or incomplete AOP development that could misinform regulatory decisions [5]. This guide outlines a standardized protocol to meet this need.

Core Methodological Framework

The following integrated workflow (Figure 1) provides a comprehensive protocol for systematic evidence gathering, KER evaluation, and taxonomic domain expansion. This multi-step process transitions from qualitative data assembly to quantitative network analysis and finally to computational extrapolation.

Figure 1: Integrated Workflow for Systematic KER Review and tDOA Expansion

Phase 1: Evidence Synthesis

• Step 1: Protocol Development: Define the review's scope using a PICOC framework (Population, Intervention, Comparator, Outcome, Context) [5]. For KER development, this translates to specifying the chemical/stressor, the biological pathway of interest, the relevant KE terms, and the initial taxonomic focus.
• Step 2: Systematic Literature Review: Execute a comprehensive, reproducible search across multiple databases (e.g., Web of Science, Scopus, PubMed) using structured search strings [5]. Document the screening process (title/abstract, then full-text) against pre-defined eligibility criteria.
• Step 3: Evidence Collection & Mapping: Extract data from included studies into a structured map. Key fields include: study identifier, model organism, chemical/stressor tested, measured endpoints, reported effect size and direction, and the mapped KE(s) [15]. This map reveals clusters of evidence and critical gaps across species and biological levels [5].
• Step 4: Qualitative AOP Network Assembly: Synthesize the mapped evidence into a preliminary AOP network. Link KEs based on the collected evidence and established biological plausibility. This creates a visual hypothesis of the pathway's structure [15].

Phase 2: Quantitative Analysis & Extrapolation

• Step 5: Quantitative KER Analysis (Bayesian Network Modeling): Transform the qualitative network into a probabilistic Quantitative AOP (qAOP). Using tools like Bayesian Networks (BNs), model the strength and uncertainty of each KER based on the extracted experimental data [15]. This step quantifies confidence, identifies the most influential KEs, and allows for predictive simulation.
• Step 6: Taxonomic Domain Extension with Computational NAMs: Expand the tDOA beyond empirically tested species using two complementary in silico tools:
  • SeqAPASS: Analyzes the conservation of protein sequences (e.g., the MIE target) across species to predict potential susceptibility [9] [15].
  • G2P-SCAN: Evaluates the conservation of entire biological pathways (encompassing multiple KEs) across species from human gene inputs [9] [15]. The combined results provide a weight-of-evidence for pathway conservation, enabling a scientifically defensible expansion of the tDOA [9].

Detailed Experimental & Analytical Protocols

Protocol for Systematic Evidence Mapping

This protocol adapts established systematic review standards for the specific context of KER development [5].

• Search Strategy: Develop Boolean search strings combining terms for the stressor (e.g., "silver nanoparticle"), the pathway/effect (e.g., "oxidative stress", "reproduction"), and relevant model taxa. Search multiple databases to minimize bias [5].
• Screening & Eligibility: Use dual-reviewer screening at title/abstract and full-text levels. Eligibility criteria must explicitly define acceptable study designs (in vivo, in vitro, omics), required outcome metrics (e.g., must report a measurable endpoint linkable to a KE), and language restrictions [5].
• Data Extraction: Populate a standardized spreadsheet or database. Each row represents a unique experimental observation. Essential columns include: citation, test species, exposure regimen, quantitative result (mean, variance, n), calculated effect size (e.g., Hedge's g), and the specific KE(s) the evidence supports [15].

Protocol for Bayesian Network Modeling of KERs

This protocol details the quantitative assessment of KER confidence, moving beyond qualitative linkage [15].

• Data Preparation: Convert extracted quantitative data into states for each KE (e.g., "Low", "Medium", "High" for oxidative stress). Use expert judgment or data distribution thresholds to define these states.
• Network Structure Learning: Define the network structure (the direction of arrows between KEs) based on the biological plausibility established in the qualitative AOP network. The structure is not learned from data but is informed by it.
• Parameter Learning: Use the discretized experimental data to calculate Conditional Probability Tables (CPTs) for each node. These tables quantify the probability of a child KE being in a certain state given the state of its parent KE(s).
• Model Validation & Inference: Validate the BN using sensitivity analysis or hold-out data if available. Use the completed BN to perform probabilistic inference, such as predicting the probability of the AO given evidence of a specific MIE.

Protocol for Cross-Species Extrapolation using SeqAPASS & G2P-SCAN

This integrated computational protocol expands the tDOA [9] [15].

• Target Identification: From the AOP, identify the primary protein target of the MIE (for SeqAPASS) and the core set of genes/proteins involved in the pathway (for G2P-SCAN).
• SeqAPASS Analysis:
  • Input the reference protein sequence (e.g., human or model organism) into the SeqAPASS tool.
  • Perform sequential alignment filters (primary, secondary, tertiary) to assess structural and functional conservation across the species database.
  • Export predictions of susceptibility for hundreds of species based on sequence similarity thresholds.
• G2P-SCAN Analysis:
  • Input the list of human genes central to the AOP pathway into the G2P-SCAN tool.
  • The tool maps genes to Reactome pathways and assesses the conservation of these pathways across a defined set of species (e.g., human, mouse, zebrafish, nematode).
  • Analyze output for pathway completeness and confidence scores in each species.
• Evidence Integration: Synthesize results. High confidence from both tools provides strong evidence for tDOA inclusion. Interpret divergent results cautiously, considering the strengths of each method (SeqAPASS for specific protein target conservation, G2P-SCAN for broader pathway context).

Data Presentation: Quantitative Evidence Synthesis

Systematic reviews generate critical quantitative data that must be presented clearly. The following tables exemplify structured summaries for key outputs.

Table 1: Summary of Evidence Base for AOP Network Development This table catalogues the foundational studies, demonstrating the integration of data across testing modalities and species, a hallmark of a robust systematic review [15].

Ecological Compartment	Initial Taxonomic Domain (tDOA)	Number of Primary Studies	Key References (Examples)	Extended tDOA (Post-Analysis)
Terrestrial	Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens (in vitro)	17	Ahn et al. (2014); Eom & Choi (2010); Kim et al. (2009) [15]	Fungi (98 species), Birds (28), Rodents (1), Reptiles (1) [15]
Aquatic	Danio rerio (zebrafish), Daphnia magna	8	Choi et al. (2010); Kim et al. (2016) [15]	Fish (157 species), Amphibians (5), Aquatic invertebrates (11) [15]
Integrated	Multi-compartment synthesis	25	Collection of studies from 2009-2019 [15]	Total Extended: Over 100 taxonomic groups [15]

Table 2: Analysis of Evidence Gaps and Biases in KER Support Inspired by systematic maps in related fields, this table diagnoses the distribution of evidence, crucial for prioritizing future research and qualifying confidence in the AOP's tDOA [5] [28].

Analysis Dimension	Evidence Clusters (Well-Represented)	Evidence Gaps (Under-Represented)	Implication for KER Confidence
Taxonomic Focus	Arthropods (esp. insects), Microorganisms, Vertebrates (fish, rodents) [5]	Annelids, Amphibians, Reptiles, Plants [5]	KERs may be less certain for gap taxa; extrapolation required.
Biological Metric	Abundance, Species Richness, Mortality [5] [28]	Functional Diversity, Phylogenetic Diversity, Behavioral Endpoints [5] [28]	AOP supports population-level AOs but may miss ecosystem function impacts.
Geographic Origin	Studies from USA, China, Brazil, European countries [5]	Studies from tropical and Global South regions [5] [28]	tDOA may be biased towards species and conditions in well-studied regions.
Practice/Intervention	Fertilizer use, pesticide application, crop diversification [5]	Combined practice effects, landscape-level management [5]	AOPs for single stressors are stronger than for complex mixture or multi-stressor scenarios.

Visualizing Key Event Relationships and Pathways

Clear visualization of the AOP structure and the evidence supporting each KER is fundamental. The following diagram depicts a generalized AOP network for a chemical stressor, highlighting the strength of KERs based on systematic review output.

Figure 2: AOP Network with Mapped KER Evidence and tDOA Support

The Scientist's Toolkit: Essential Research Reagent Solutions

Successful execution of this systematic methodology relies on a suite of specific tools and databases. The following table details these essential "research reagents."

Table 3: Key Digital Tools and Databases for Systematic KER Review

Tool/Resource Name	Type	Primary Function in KER/tDOA Research	Access Link/Reference
Abstrackr	Screening Software	A semi-automated tool for accelerating the title/abstract screening phase of systematic reviews [5].	https://abstrackr.cebm.brown.edu/
SeqAPASS	Computational NAM	Predicts chemical susceptibility across species by analyzing protein sequence and structural conservation of molecular targets [9] [15].	https://seqapass.epa.gov/seqapass/
G2P-SCAN	Computational NAM	Estimates conservation of biological pathways across species from human gene inputs, providing pathway-level context for tDOA [9] [15].	Rivetti et al. (2023) [9]
AOP-Wiki	Knowledgebase	Central repository for collaborative AOP development. Essential for KE ontology and hosting published AOPs [15].	https://aopwiki.org/
CompTox Chemicals Dashboard	Chemistry Database	Provides curated chemical information, identifiers, and linked bioactivity data (e.g., ToxCast) for stressor characterization [9].	https://comptox.epa.gov/dashboard/
Reactome	Pathway Database	A curated, peer-reviewed knowledgebase of biological pathways. Used by tools like G2P-SCAN for pathway mapping [9].	https://reactome.org/
RCSB Protein Data Bank	Structural Database	Provides 3D structural data for proteins, useful for understanding MIEs and informing SeqAPASS analysis [9].	https://www.rcsb.org/
Bayesian Network Software (e.g., Netica, AgenaRisk)	Modeling Software	Enables the construction, training, and inference of Bayesian Network models for quantitative KER analysis [15].	Commercial & Open-Source Options

This case study is situated within a broader thesis investigating the conservation of Key Event Relationships (KERs) across taxonomic groups. The central premise is that the mechanistic toxicity pathways of chemicals, when structured as Adverse Outcome Pathways (AOPs), often rely on biological processes conserved through evolution. Therefore, validating an AOP in one model organism provides a powerful, hypothesis-driven framework for predicting toxicity in other species, bridging human and ecological toxicology under a One Health perspective [15] [29]. Silver nanoparticles (AgNPs) serve as an exemplary stressor for this research due to their widespread use, documented toxicity, and a well-characterized AOP (AOP 207) initiating from oxidative stress [30] [15]. This study demonstrates a methodological workflow for extending the taxonomic Domain of Applicability (tDOA) of an existing AOP by integrating ecotoxicological data, human toxicology data, and in silico cross-species extrapolation tools [15].

Background: AgNP Toxicity and the Foundation AOP

Silver nanoparticles (AgNPs) are defined as particles with at least one dimension between 1 and 100 nm [30]. Their extensive commercial use in textiles, cosmetics, food packaging, and medical products for their antimicrobial properties leads to potential exposure via ingestion, inhalation, and dermal contact [30] [31]. A primary mechanism of AgNP toxicity is the induction of oxidative stress. This can occur via the "Trojan-horse" mechanism, where intracellular dissolution of AgNPs releases Ag⁺ ions that impair thiol (SH)-containing antioxidants like glutathione, or through surface reactions generating reactive oxygen species (ROS) [30] [32].

The Adverse Outcome Pathway (AOP) framework organizes this knowledge into a causal sequence: a Molecular Initiating Event (MIE) leads to a series of measurable Key Events (KEs), culminating in an Adverse Outcome (AO) [30] [2]. AOP 207: "NADPH oxidase and P38 MAPK activation leading to reproductive failure in Caenorhabditis elegans" provides the foundation for this case study [30] [15]. Its KEs include oxidative stress (MIE), PMK-1/p38 MAPK activation, HIF-1 activation, mitochondrial damage, DNA damage, apoptosis, and finally, reproductive failure (AO) [30].

Core Methodology for Taxonomic Extension of an AOP

Extending the tDOA of an AOP requires a multi-step, integrative approach that moves from data collection to computational validation and prediction.

Data Collection and AOP Network Development

The first phase involves a systematic gathering of existing evidence from diverse studies to construct a putative cross-species AOP network.

• Source Identification: Peer-reviewed literature is systematically searched using databases (e.g., PubMed, Scopus). For AgNP reproductive toxicity, keywords include "silver nanoparticle," "reproductive toxicity," "oxidative stress," and specific KE terms like "apoptosis" or "mitochondrial damage" [30].
• Study Screening & Data Extraction: Identified studies are screened for quality (e.g., using ToxRTool [30]) and relevant endpoints. Data extracted includes model species (e.g., C. elegans, rodents, human cell lines), exposure conditions, measured biological endpoints, and quantitative results [30] [15].
• KE Matching and Network Building: Extracted endpoints are assigned to potential KEs (MIE, intermediate KE, AO) based on biological plausibility and expert judgment. Data from different species and biological levels (molecular, cellular, organismal) are integrated to form an initial AOP network that hypothesizes conserved KERs [15].

Quantitative Assessment of Key Event Relationships (KERs)

To evaluate the strength and confidence in the proposed AOP network, a probabilistic modeling approach is employed.

• Bayesian Network (BN) Modeling: BN analysis is used to quantify the causal relationships between KEs. This method is adept at handling uncertainty and variability in biological data [33] [15].
• Protocol:
  • Data Preparation: Compile a dataset where each observation represents a experimental condition, with columns for the measured states (e.g., fold-change, activity level) of each KE (node) in the AOP network.
  • Network Structure Learning: Use algorithms (e.g., from R packages like bnlearn) to infer the probabilistic dependency structure between KEs from the data, or impose the structure based on the hypothesized AOP [33].
  • Parameter Learning & Validation: Calculate the conditional probability tables for each node. Validate the model using sensitivity analysis or by testing its predictive accuracy on a subset of withheld data.
  • Inference: Use the validated BN to compute the probability of the AO given observations of upstream KEs, strengthening the weight of evidence for the KERs [33] [15].

In SilicoTaxonomic Extrapolation

Computational tools are used to predict the biological plausibility of the AOP across a wide range of species, formally extending its tDOA [15].

• Tool 1: Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS).
  • Protocol: Input the protein sequence (e.g., from NCBI) of a critical target involved in the MIE or an early KE (e.g., the NADPH oxidase dual oxidase BLI-3 in C. elegans for AOP 207). SeqAPASS performs pairwise alignments with sequences from other species in its database, assessing primary, secondary, and tertiary structural similarity to predict potential susceptibility [15].
• Tool 2: Genes-to-Pathways Species Conservation Analysis (G2P-SCAN).
  • Protocol: Input a set of genes (e.g., orthologs of genes involved in the p38 MAPK signaling or apoptosis pathways from the AOP). G2P-SCAN uses pre-computed orthology data to evaluate the conservation of the entire functional pathway or biological process across specified taxonomic groups [15].
• Integration: The outputs from SeqAPASS (target susceptibility) and G2P-SCAN (pathway conservation) are combined to generate evidence for the extended tDOA, predicting which taxonomic groups are likely to exhibit the AOP due to conserved molecular and pathway components [15].

Quantitative Data and Taxonomic Extension Outcomes

Organ-Specific Accumulation of AgNPs

The distribution of AgNPs, a modulating factor for toxicity, varies by particle size and organ.

Table 1: Size-Dependent Accumulation of Intravenously Administered AgNPs in Rat Organs [30]

Organ	20 nm AgNP Concentration (ng/g tissue)	80 nm AgNP Concentration (ng/g tissue)	110 nm AgNP Concentration (ng/g tissue)
Spleen	~80	~1,600	~1,600
Liver	169	539	1,077
Testes	Low and comparable for all sizes	Low and comparable for all sizes	Low and comparable for all sizes

Extension of Taxonomic Domain of Applicability (tDOA)

Applying the integrated in silico workflow to the AgNP reproductive toxicity AOP network successfully extended its biologically plausible tDOA far beyond the initial model organism.

Table 2: Extended Taxonomic Domain of Applicability for the AgNP Reproductive Toxicity AOP Network [15]

Ecological Compartment	Initial tDOA (Model Species)	Extended tDOA (Number of Species/Groups)
Terrestrial	Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens (in vitro)	Fungi (98), Birds (28), Rodents (1), Reptiles (1), Nematodes (1)
Aquatic	Danio rerio (zebrafish), Daphnia magna	Fish (154), Arthropods (43), Amphibians (17), Mollusks (8), Annelids (3)

Visualization of Pathways and Workflows

Putative AOP for AgNP-Induced Male Reproductive Toxicity [30]

Workflow for Cross-Species AOP Development and tDOA Extension [15]

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Reagents and Materials for AgNP AOP Research

Item	Function/Description	Relevance to AOP Research
Characterized AgNPs	Nanoparticles of defined size (e.g., 20, 50, 100 nm), coating (e.g., PVP, citrate), and charge.	The foundational stressor. Physicochemical properties dictate dissolution, uptake, and MIE potency [30] [31].
Thiol-containing Biomolecules	e.g., Glutathione (GSH), N-acetylcysteine (NAC), Cysteine.	Used to test the "thiol-scavenging" mechanism of Ag⁺ ions. Supplementation can rescue oxidative stress, confirming the MIE [30] [32].
ROS Detection Assays	e.g., DCFH-DA, DHE, MitoSOX.	Fluorescent probes to quantitatively measure intracellular or mitochondrial ROS generation, a core MIE/early KE [30] [32].
Mitochondrial Function Assays	e.g., JC-1 (membrane potential), MTT/XTT (metabolic activity), ATP luminescence kits.	Measure mitochondrial damage (a KE) resulting from oxidative stress [30] [31].
Apoptosis Detection Kits	e.g., Annexin V/PI staining, caspase-3/7 activity assays.	Quantify apoptotic cell death, a critical KE preceding tissue/organ dysfunction [30] [31].
AR Antagonism Assay	e.g., AR-CALUX, MDA-kb2 cell line.	Validated OECD test (No. 458) for detecting androgen receptor antagonism, a potential alternative MIE for reproductive toxicity AOPs [34].
BN Analysis Software	e.g., R packages (bnlearn, gRain), commercial BN software.	Essential for implementing the probabilistic quantitative assessment of KERs and building predictive qAOP models [33] [15].

Overcoming Challenges: Troubleshooting Common Pitfalls in Establishing KER Conservation

Within the broader thesis on Key Event Relationship (KER) taxonomic conservation research, a central challenge is the development of robust, predictive models when empirical data linking molecular events to adverse outcomes is sparse or confined to a narrow range of species. The adverse outcome pathway (AOP) framework has been widely adopted to structure this causal knowledge, but its predictive power for ecological and human health risk assessment depends on a clear understanding of a pathway's taxonomic domain of applicability (tDOA) [15]. Traditionally, establishing tDOA required extensive, costly, and ethically challenging in vivo testing across multiple species. Today, the convergence of New Approach Methodologies (NAMs)—spanning in vitro, in chemico, and in silico tools—provides a revolutionary strategy for addressing these data gaps [9]. This whitepaper details a core, integrative methodology that combines computational toxicology, bioinformatics, and probabilistic modeling to extrapolate KER confidence across the tree of life, thereby reducing reliance on novel animal testing and accelerating the application of mechanistic data in safety decision-making [15].

Core Methodological Framework

The proposed framework is an iterative, weight-of-evidence process that transforms limited empirical KER data into a predictive, cross-species AOP network (AOPN). It progresses from data collation to quantitative assessment and finally to taxonomic extrapolation.

Data Collection & Network Construction

The initial phase involves the systematic aggregation of all available evidence related to a molecular initiating event (MIE) and its downstream key events (KEs). Data sources must include:

• Empirical in vivo data from standard model organisms (e.g., C. elegans, zebrafish, rat).
• In vitro human and mammalian toxicity data, including high-throughput screening (HTS) results from platforms like ToxCast [9].
• Molecular interaction data from structural databases (e.g., Protein Data Bank) and the scientific literature on binding affinities, gene expression changes, and pathway perturbations [9].
• Existing AOP descriptions from the AOP-Wiki [15] [35].

Each study is deconstructed, and its endpoints are mapped to standardized KE terms within the AOP framework. This integrated evidence forms a putative qualitative AOPN, representing hypothesized causal linkages across biological scales [15].

Quantitative Assessment of Key Event Relationships

To move from qualitative linkage to quantitative prediction, the strength and uncertainty of each KER must be evaluated. A Bayesian Network (BN) modeling approach is recommended over deterministic regression for this purpose [15].

• Protocol: Using a BN software platform (e.g., Netica, AgenaRisk), define network nodes for each KE (including the MIE and AO). Parameterize the conditional probability tables for each node using the assembled empirical data. This process quantitatively encodes the probability of a downstream KE given the state of an upstream KE, explicitly handling data gaps and variability.
• Application: For an AOP describing reproductive toxicity via oxidative stress, BN analysis can quantify the probability of reproductive failure given a measured increase in reactive oxygen species (ROS), integrating data from both in vivo (e.g., C. elegans) and in vitro (human cell) studies [15]. This probabilistic model becomes the foundational quantitative AOP (qAOP) for extrapolation.

Cross-Species Extrapolation via Computational NAMs

This phase extends the biologically plausible tDOA of the qAOP using two complementary in silico tools.

• Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS): This tool predicts potential chemical susceptibility by comparing the primary through quaternary structures of a protein target (the MIE) across species [15] [9].

  • Protocol: Input the amino acid sequence of the well-characterized molecular target (e.g., a specific receptor or enzyme) from a reference species. SeqAPASS performs tiered sequence alignment analyses against the NCBI protein database. Results are interpreted as "sequence similarity," "functional site conservation," and "structural similarity" scores, generating a prediction of susceptibility for hundreds to thousands of species [9].

• Genes-to-Pathways Species Conservation Analysis (G2P-SCAN): This tool evaluates the conservation of entire biological pathways (the chain of KEs) beyond the initial molecular target [15] [9].

  • Protocol: Input a list of human genes corresponding to the KEs in the pathway. G2P-SCAN maps these to conserved biological pathways in Reactome and assesses their conservation across a defined set of model species (e.g., human, mouse, zebrafish, fruit fly, worm). The output indicates whether the downstream pathway logic, not just the initial target, is likely conserved in a given taxon [9].

The convergence of evidence from SeqAPASS (target-level susceptibility) and G2P-SCAN (pathway-level conservation) provides a robust, multi-layered rationale for extending the tDOA [15] [9].

Results & Data Integration

Applying this integrated methodology yields two primary quantitative outputs: a vastly expanded tDOA and a probabilistic qAOP model. The power of the approach is demonstrated through case studies, such as the extension of an AOP for silver nanoparticle (AgNP)-induced reproductive toxicity.

Table 1: Cross-Species Extrapolation Results for AgNP Reproductive Toxicity AOP (AOP 207) [15]

Ecological Compartment	Initial tDOA (Empirical Data)	Extended tDOA (In Silico Prediction)	Number of Taxonomic Groups
Terrestrial	C. elegans, D. melanogaster, H. sapiens (in vitro)	Fungi, Birds, Rodents, Reptiles, Nematodes	128+
Aquatic	C. riparius, D. rerio, H. sapiens (in vitro)	Fish, Amphibians, Crustaceans, Mollusks	100+
Combined Cross-Species AOPN	3 Model Species	>228 Taxonomic Groups	>228

The quantitative confidence in the KERs underpinning this network is derived from the BN model. The performance of the computational prediction tools themselves can be benchmarked.

Table 2: Performance Metrics of Computational KER Extrapolation Methods

Method	Primary Data Input	Output/ Prediction	Key Strength for KER Research	Typical Application Context
SeqAPASS [15] [9]	Protein sequence/structure	Taxonomic susceptibility for molecular target	High-throughput, broad taxonomic coverage	Establishing plausibility of MIE across species
G2P-SCAN [15] [9]	List of human genes	Pathway conservation across model species	Contextualizes MIEs within conserved biological processes	Supporting conservation of downstream KEs and pathways
Bayesian Network Modeling [15]	Empirical dose-response data	Probabilistic KERs with uncertainty quantification	Integrates disparate data types; handles variability and gaps	Building quantitative, predictive qAOPs for risk assessment
Agnolog Identification [36]	Transcriptomic & network data	Functionally equivalent genes/gene sets	Identifies conserved function beyond strict sequence homology	Mapping KERs in non-traditional or distantly related species

The Scientist's Toolkit: Research Reagent Solutions

Implementing this framework requires a suite of specific computational and data resources.

Table 3: Essential Research Toolkit for Cross-Species KER Analysis

Item	Function in Research	Key Resource / Example
AOP-KB (AOP-Wiki)	Central repository for accessing, developing, and sharing structured AOP knowledge, including KEs and KERs [35].	https://aopwiki.org/
SeqAPASS Tool	Web-based tool for predicting protein target conservation and chemical susceptibility across species via sequence alignment [9].	https://seqapass.epa.gov/seqapass/
G2P-SCAN R Package	Tool for evaluating conservation of biological pathways (Reactome) from human gene lists across model species [15].	R package G2P-SCAN
CompTox Chemicals Dashboard	Provides access to HTS data (ToxCast/Tox21), chemical properties, and bioactivity data to inform MIE and KE identification [9].	https://comptox.epa.gov/dashboard
Bayesian Network Software	Platform for constructing, parameterizing, and running probabilistic models to quantify KERs under uncertainty [15].	Netica, AgenaRisk, or R packages (bnlearn, gRain)
Protein Data Bank (PDB)	Repository for 3D structural data of proteins and complexes, critical for understanding MIEs at the atomic level [9].	https://www.rcsb.org/

Technical Visualizations

Diagram 1: Integrated workflow for cross-species KER analysis.

Diagram 2: Example AOP network for SSRI-induced feeding inhibition.

Diagram 3: Decision logic for extending the taxonomic domain of applicability.

Implementation & Challenges

Successfully implementing this strategy requires navigating several technical and philosophical challenges. A primary consideration is defining confidence thresholds for the in silico predictions; the "degree of similarity" in SeqAPASS or pathway coverage in G2P-SCAN that constitutes sufficient evidence for inclusion in the tDOA is context-dependent and must be justified [15] [9]. Furthermore, the field must grapple with the identification of "agnologs"—functionally equivalent genes or pathways that are not orthologous—which may be critical for transferring knowledge between evolutionarily distant species [36]. Finally, the validation of extrapolated KERs remains an iterative process. While this methodology minimizes animal testing, targeted in vitro assays in cells from predicted susceptible species or limited in vivo studies in non-traditional model organisms can provide crucial confirmatory evidence, strengthening the overall weight of evidence for regulatory application [15].

In the context of Key Event Relationship (KER) taxonomic conservation research, accurately distinguishing between true absence and lack of data is a fundamental challenge with significant implications for predictive model reliability and cross-species extrapolation. This distinction is critical within the Adverse Outcome Pathway (AOP) framework, where defining the taxonomic domain of applicability (tDOA) relies on precise understanding of whether a key event is genuinely not conserved in a taxonomic group or simply unmeasured [15]. The problem parallels challenges in ecological species distribution modeling, where models require information about where species are not found to accurately predict where they could exist [37] [38].

True absence in taxonomic analysis refers to confirmed non-conservation of a molecular initiating event, key event, or adverse outcome pathway across species, supported by empirical evidence or robust phylogenetic inference. In contrast, lack of data represents uncertainty stemming from insufficient investigation, where the taxonomic conservation status remains unknown due to limited research scope, methodological constraints, or inadequate detection methods [37]. This ambiguity directly impacts the confidence with which AOPs can be extrapolated for chemical safety assessment and drug development applications.

The consequences of misclassification are substantial. Overestimation of taxonomic applicability occurs when lack of data is misinterpreted as true absence, potentially leading to inappropriate cross-species predictions in toxicology or pharmacology. Conversely, underestimation of conservation patterns results when true absence is misclassified as lack of data, causing researchers to overlook legitimate taxonomic boundaries in pathway functionality. Within the broader thesis on KER taxonomic conservation, resolving this ambiguity enables more precise definition of AOP boundaries, enhances confidence in New Approach Methodologies (NAMs), and supports the development of reliable, taxonomically-aware predictive models for chemical and drug safety assessment [15] [9].

Defining the Spectrum: True Absence, Pseudo-Absence, and Lack of Data

Core Definitions and Comparative Analysis

The terminology surrounding absence in taxonomic analysis requires precise differentiation to avoid conceptual confusion in KER research. These definitions establish the foundation for methodological approaches to resolving ambiguity.

True Absence represents confirmed non-existence of a biological element within a defined taxonomic context. In KER research, this indicates that a molecular initiating event, key event, or entire pathway is genuinely not conserved in certain taxonomic groups, supported by either direct empirical evidence or robust phylogenetic inference. True absence data implies that "environmental conditions are unsuitable for a species to survive" in ecological terms [37], which translates to biological contexts where phylogenetic distance or evolutionary divergence has resulted in non-conservation of specific molecular pathways. The confidence in true absence designation increases with repeated, methodologically appropriate surveys that would detect the element if present [38].

Pseudo-Absence (in ecological modeling) or Inferred Non-Conservation (in taxonomic analysis) represents inferred rather than confirmed absence. This concept is derived from species distribution modeling where pseudo-absence points are generated in locations where a species has not been recorded but might potentially exist [37]. In taxonomic conservation research, this parallels situations where preliminary evidence suggests non-conservation, but confirmatory studies are lacking. Pseudo-absence serves as a practical substitute when true absence data is unavailable but comparative analysis requires contrast between presence and absence conditions [38].

Lack of Data constitutes genuine uncertainty stemming from insufficient investigation rather than biological reality. This occurs when taxonomic groups have been inadequately studied, detection methods are insufficiently sensitive, or research scope has been limited. Lack of data represents the "unknown" category that must be systematically addressed rather than assumed to represent either presence or absence. In ecological contexts, this would equate to unsurveyed areas where no information exists about species distribution [37].

Table 1: Comparative Analysis of Absence Classifications in Taxonomic Research

Classification	Definition	Confidence Level	Primary Source	Implications for KER Extrapolation
True Absence	Confirmed non-conservation supported by empirical evidence	High	Direct experimental evidence or robust phylogenetic analysis	Defines boundaries of tDOA; prevents over-extrapolation
Pseudo-Absence/Inferred Non-Conservation	Inferred absence based on available evidence but lacking confirmation	Medium to Low	Indirect evidence, preliminary data, or predictive modeling	Requires verification; useful for preliminary hypothesis generation
Lack of Data	Insufficient investigation to determine conservation status	Very Low	Gaps in research coverage or methodological limitations	Identifies research priorities; prevents erroneous conclusions

Methodological Implications for KER Research

The classification of absence types directly influences methodological choices in taxonomic conservation research. True absence data, when available, enables the most reliable definition of taxonomic domains of applicability for AOPs. However, comprehensive surveys to establish true absence are "time-consuming" and therefore rarely available across broad taxonomic ranges [37] [38]. This scarcity necessitates the development of alternative approaches that can differentiate between true absence and lack of data with reasonable confidence.

The prevalence ratio (proportion of occupied locations relative to absence points) significantly influences model accuracy in ecological contexts [37], suggesting analogous considerations in taxonomic analysis where the ratio of confirmed conservation to confirmed non-conservation across taxonomic groups affects the reliability of pattern identification. Furthermore, the method of generating pseudo-absence data—whether random, environmentally contrasted, or geographically constrained—affects outcomes in distribution modeling [37], indicating that the approach to handling uncertain taxonomic conservation similarly impacts conclusions in KER research.

Experimental Protocols and Methodological Framework

Integrated Workflow for Resolving Absence Ambiguity

A systematic, multi-tiered approach is required to differentiate true absence from lack of data in KER taxonomic conservation research. The following workflow integrates established ecological methods with novel computational toxicology approaches to address this challenge comprehensively.

Decision Framework for Differentiating Absence Types in KER Taxonomic Research

Phase I: Evidence-Based Assessment Protocol

Protocol 1: Comprehensive Taxonomic Survey Assessment

Objective: Determine whether sufficient investigation has been conducted to support true absence designation.

Materials: Taxonomic database access (NCBI, Ensembl), systematic review tools, phylogenetic analysis software.

Procedure:

• Systematic Literature Review: Conduct exhaustive search for studies investigating the specific KER or AOP across taxonomic groups using multiple databases and controlled vocabulary.
• Survey Comprehensiveness Evaluation: Apply criteria adapted from ecological methods [38]:
  • Number of independent studies per taxonomic group
  • Diversity of methodologies employed (molecular, physiological, in silico)
  • Temporal coverage (span of publication years)
  • Geographic/institutional diversity of research teams
• Detection Probability Assessment: For each taxonomic group, evaluate whether methods employed would detect the KER if present, considering:
  • Sensitivity of assays used
  • Developmental stages/life history phases examined
  • Environmental conditions tested
• Confidence Scoring: Assign confidence levels (high, medium, low) for true absence designation based on survey comprehensiveness.

Interpretation: Taxonomic groups with high confidence scores and consistent negative results across multiple studies may be classified as true absence. Groups with limited or methodologically inadequate investigation remain as lack of data.

Protocol 2: Phylogenetic Signal Analysis

Objective: Utilize evolutionary relationships to distinguish true absence from lack of data.

Materials: Phylogenetic trees, sequence alignment tools, ancestral state reconstruction software.

Procedure:

• Phylogenetic Framework Development: Construct or obtain robust phylogenetic tree encompassing taxa of interest.
• Character State Mapping: Map KER conservation status (present/absent/unknown) onto terminal taxa based on empirical evidence.
• Ancestral State Reconstruction: Apply maximum likelihood or Bayesian methods to infer conservation status at ancestral nodes.
• Phylogenetic Signal Quantification: Calculate metrics (D-statistic, Pagel's λ) to determine whether conservation status exhibits phylogenetic clustering.
• Predictive Modeling: Use phylogenetic comparative methods to predict conservation status in understudied taxa based on evolutionary relationships.

Interpretation: Strong phylogenetic signal with distinct clades showing conserved and non-conserved patterns supports true absence designation for non-conserved clades. Weak or absent phylogenetic signal suggests insufficient data or convergent evolution.

Phase II: Computational Integration Protocol

Protocol 3: Integrated SeqAPASS and G2P-SCAN Analysis

Objective: Leverage computational NAMs to extend taxonomic domain assessments and resolve ambiguity [15] [9].

Materials: SeqAPASS tool access, G2P-SCAN software, reference proteomes, pathway databases.

Procedure:

• Molecular Target Identification: Identify specific molecular targets (proteins, genes) involved in the KER or AOP.
• SeqAPASS Analysis [9]:
  • Input reference protein sequences for molecular targets
  • Perform pairwise sequence alignments across taxonomic groups
  • Assess conservation of functional domains and active sites
  • Generate susceptibility predictions based on sequence similarity thresholds
• G2P-SCAN Analysis [15] [9]:
  • Input human gene sets corresponding to KER pathways
  • Map to Reactome or KEGG pathways
  • Assess pathway conservation across model species (human, mouse, rat, zebrafish, fruit fly, roundworm, yeast)
  • Identify essential pathway components
• Integrated Interpretation:
  • Concordant results (both tools suggest non-conservation) provide stronger evidence for true absence
  • Discordant results require additional investigation
  • Limited data in either tool indicates lack of data rather than true absence

Interpretation: Computational evidence can upgrade classification from lack of data to inferred non-conservation (pseudo-absence) or, when combined with other evidence, support true absence designation.

Protocol 4: Bayesian Network Modeling of KERs

Objective: Quantitatively assess confidence in absence designations through probabilistic modeling of key event relationships [15].

Materials: Bayesian network software, empirical data on KERs, prior probability estimates.

Procedure:

• Network Structure Development: Construct Bayesian network representing causal relationships between molecular initiating events, key events, and adverse outcomes.
• Parameter Estimation: Use available empirical data to estimate conditional probability distributions for each relationship.
• Taxonomic Integration: Incorporate taxonomic conservation as nodes influencing probability of KER activation.
• Sensitivity Analysis: Assess how uncertainty in conservation status affects overall AOP plausibility.
• Predictive Application: Use network to predict KER functionality in taxa with uncertain conservation status based on related taxa.

Interpretation: Bayesian networks provide quantitative confidence estimates for absence designations and can identify which uncertain classifications most significantly impact model predictions.

Table 2: Summary of Experimental Protocols for Resolving Absence Ambiguity

Protocol	Primary Objective	Key Methodologies	Output Metrics	Strength for Absence Determination
Comprehensive Taxonomic Survey Assessment	Evaluate sufficiency of empirical investigation	Systematic review, detection probability assessment	Confidence scores, coverage metrics	Direct assessment of research gaps; identifies true lack of data
Phylogenetic Signal Analysis	Leverage evolutionary relationships to predict conservation	Ancestral state reconstruction, phylogenetic comparative methods	Phylogenetic signal metrics, ancestral state probabilities	Evolutionary context for absence patterns; distinguishes true absence from sampling artifacts
Integrated SeqAPASS and G2P-SCAN Analysis	Computational assessment of molecular and pathway conservation	Sequence alignment, pathway mapping, conservation scoring	Sequence similarity scores, pathway conservation metrics	Extends assessment beyond empirically studied taxa; provides mechanistic basis for absence
Bayesian Network Modeling of KERs	Quantitative probabilistic assessment of AOP functionality	Bayesian inference, sensitivity analysis, predictive modeling	Conditional probabilities, confidence intervals, sensitivity indices	Quantifies uncertainty in absence designations; identifies most impactful knowledge gaps

Research Reagent Solutions

Table 3: Research Reagent Solutions for KER Taxonomic Conservation Studies

Reagent/Tool Category	Specific Examples	Function in Absence Determination	Key Considerations for Use
Cross-Species Antibody Panels	Phospho-specific antibodies conserved across taxa, domain-targeted antibodies	Detection of protein expression/post-translational modifications across species	Validate cross-reactivity for each taxon; consider epitope conservation
Conserved Molecular Probes	Fluorescent in situ hybridization (FISH) probes for conserved gene regions, activity-based protein profiling probes	Visualization of gene expression/protein activity patterns across taxa	Design against most conserved regions; test specificity in each taxon
Taxonomic-Broad PCR Primers	Degenerate primers for conserved functional domains, universal primer sets for gene families	Amplification of target sequences from diverse taxa for comparative analysis	Optimize annealing temperatures for broad specificity; include positive controls
Pathway Activity Reporters	Conserved response element-driven luciferase constructs, pathway activation biosensors	Functional assessment of pathway activity/conservation across cell types from different species	Normalize for transfection efficiency/species-specific cellular properties
Reference Tissue Banks	Multi-species tissue collections, cell line repositories from diverse taxa	Provide biological materials for comparative studies across under-represented taxa	Ensure proper preservation methods; document taxonomic verification

Table 4: Computational Tools for Taxonomic Conservation Assessment

Tool Name	Primary Function	Application in Absence Determination	Access/Reference
SeqAPASS	Protein sequence comparison across species to predict chemical susceptibility [9]	Assess conservation of molecular targets across taxa; identify taxa likely lacking specific targets	https://seqapass.epa.gov/seqapass/ [9]
G2P-SCAN	Evaluate biological pathway conservation from human gene inputs across model species [15] [9]	Determine if entire pathways (not just individual components) are conserved across taxa	R package v0.0.1.0 [9]
AOP-Wiki	Collaborative repository of adverse outcome pathways with taxonomic applicability information [15]	Access existing knowledge on AOP conservation; identify knowledge gaps for specific taxa	https://aopwiki.org/
PhyloTree	Interactive visualization and analysis of phylogenetic relationships	Provide evolutionary context for absence patterns; identify clade-specific conservation	Multiple implementations available
Taxonomic Domain Mapper	Custom tool for visualizing tDOA across phylogenetic trees (conceptual)	Visual representation of conservation patterns and knowledge gaps across taxonomy	Development recommended based on [15]

Integration with AOP Development: A Case Study Framework

Workflow for Taxonomic Domain of Applicability Determination

The integration of absence determination methodologies into AOP development follows a systematic workflow that enhances the confidence in tDOA specification. This approach is exemplified by recent research extending the tDOA for AOP 207 involving reproductive toxicity of silver nanoparticles via oxidative stress in Caenorhabditis elegans [15].

Integrated AOP Development Workflow with Absence Ambiguity Resolution

Case Study Implementation: Silver Nanoparticle Reproductive Toxicity AOP

The application of absence determination methodologies is illustrated by the extension of AOP 207 (NADPH oxidase and P38 MAPK activation leading to reproductive failure in Caenorhabditis elegans) to a broader taxonomic domain [15].

Experimental Approach:

• Initial Data Integration: Collected and structured data from 25 mechanism-based toxicity studies on silver nanoparticles spanning ecotoxicology (multiple species), human toxicology (in vitro models), and the existing AOP 207 [15].
• Bayesian Network Assessment: Applied Bayesian network modeling to assess confidence in key event relationships, providing quantitative evaluation of relationship strength across different biological contexts [15].
• Absence Ambiguity Resolution: Applied Protocols 1-4 to distinguish true absence of pathway components from lack of data across taxonomic groups.
• Computational tDOA Extension: Utilized SeqAPASS and G2P-SCAN tools to extend the biologically plausible tDOA to over 100 taxonomic groups [15].

Key Findings:

• Integrated evidence from ecotoxicology and human toxicology strengthened confidence in KERs
• Bayesian approaches effectively managed uncertainty in cross-species extrapolation
• Computational NAMs successfully extended tDOA predictions beyond empirically studied taxa
• Absence determination protocols enabled differentiation between truly non-conserved pathways and those merely understudied

Quantitative Outcomes:

• Initial tDOA: Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens (in vitro)
• Extended tDOA: Fungi (98 species), birds (28 species), rodents (1 species), reptiles (1 species), nematodes (1 species) [15]
• Confidence increase: Bayesian network modeling provided quantitative confidence metrics for KERs across taxonomic groups

Table 5: Quantitative Results from AOP 207 Taxonomic Extension Study [15]

Taxonomic Group	Initial Evidence Base	Absence Determination Outcome	Extended tDOA Inclusion	Confidence Level
Fungi	Limited direct studies	SeqAPASS indicated protein conservation; G2P-SCAN suggested pathway functionality	Included (98 species)	Medium (computational evidence)
Birds	Some ecotoxicology data	Phylogenetic analysis suggested conservation; limited empirical confirmation	Included (28 species)	Medium-High
Fish	Substantial ecotoxicology literature	Strong evidence of pathway conservation; some species-specific variations	Included (not quantified in source)	High
Insects (beyond Drosophila)	Limited studies	Computational prediction suggested conservation; empirical data lacking	Conditionally included	Low-Medium

Applications in Drug Development and Toxicological Assessment

Strategic Implementation for Predictive Toxicology

The rigorous differentiation between true absence and lack of data provides substantial value throughout the drug development pipeline, particularly in safety assessment and species selection for toxicology studies.

Early Discovery Phase Applications:

• Target Conservation Assessment: Determine whether molecular drug targets are truly absent in non-target species to predict potential off-target ecological effects
• Pathway Interaction Predictions: Identify conserved pathways that might mediate unintended effects across species
• Model Selection Justification: Provide evidence-based rationale for selecting particular model organisms based on pathway conservation rather than tradition

Preclinical Development Applications:

• Species Selection for Toxicology Studies: Choose toxicology species that appropriately represent pathway conservation relevant to mechanism of action
• Risk Hypothesis Refinement: Develop more precise risk hypotheses regarding potential adverse outcomes based on pathway conservation patterns
• Biomarker Identification: Identify conserved biomarkers of pathway activation that can be monitored across species

Regulatory Submission Support:

• tDOA Justification: Provide robust evidence for the taxonomic domain of applicability in regulatory submissions
• Uncertainty Quantification: Explicitly quantify and justify uncertainties related to cross-species extrapolation
• Alternative Model Rationale: Justify use of alternative models (in vitro, in silico) based on pathway conservation evidence

Quantitative Framework for Decision Support

The integration of absence determination protocols enables development of quantitative decision frameworks for cross-species extrapolation in toxicological assessment.

Confidence Scoring System:

• Evidence Quality Metrics: Score empirical evidence based on study design, methodological appropriateness, and replication
• Taxonomic Coverage Index: Quantify breadth of taxonomic investigation relative to phylogenetic diversity
• Methodological Diversity Score: Assess variety of approaches used to investigate conservation
• Computational Corroboration Metric: Evaluate consistency between empirical findings and computational predictions

Decision Thresholds:

• High Confidence True Absence: ≥3 independent methodologically appropriate studies demonstrating non-conservation AND computational prediction of non-conservation AND phylogenetic pattern supporting absence
• Medium Confidence Inferred Non-Conservation: Limited empirical evidence BUT strong computational prediction OR clear phylogenetic pattern
• Lack of Data: Insufficient empirical investigation (<2 studies) AND inconclusive computational predictions

Implementation in Risk Assessment:

• Incorporate confidence scores into weight-of-evidence approaches for cross-species extrapolation
• Use absence determinations to define boundaries for read-across applications
• Apply probabilistic methods to account for uncertainty in absence classifications

Future Directions and Concluding Synthesis

Emerging Methodologies and Integration Opportunities

The field of absence determination in taxonomic analysis is evolving rapidly with several promising directions for methodological advancement:

High-Throughput Experimental Approaches:

• Cross-species transcriptomic profiling: Systematic comparison of pathway responses across multiple species under standardized conditions
• Protein interaction network mapping: Comparative analysis of protein-protein interactions relevant to AOPs across model organisms
• CRISPR-based functional screening: High-throughput assessment of gene essentiality in pathway contexts across cell lines from diverse species

Advanced Computational Integration:

• Machine learning classifiers: Development of algorithms trained on known presence/absence patterns to predict conservation status in understudied taxa
• Multi-omics data integration: Combined analysis of genomic, transcriptomic, proteomic, and metabolomic data to assess pathway functionality
• Phylogenetic comparative methods enhancement: Development of more sophisticated models incorporating gene duplication, loss, and functional divergence

Knowledge Synthesis Frameworks:

• Semantic integration platforms: Systems that automatically extract and synthesize conservation information from diverse literature sources
• Confidence-weighted knowledge graphs: Dynamic representations of AOP conservation with confidence metrics for each assertion
• Collaborative annotation tools: Crowdsourced approaches to evaluating and updating conservation status based on community expertise

Concluding Recommendations for KER Taxonomic Conservation Research

Based on the methodologies and applications presented, the following recommendations emerge for researchers addressing absence ambiguity in KER taxonomic conservation:

Methodological Recommendations:

• Adopt tiered assessment frameworks that progressively integrate evidence from empirical studies, phylogenetic analysis, and computational predictions
• Implement confidence scoring systems that transparently communicate uncertainty in absence determinations
• Standardize reporting of taxonomic coverage and methodological limitations in studies investigating KER conservation

Integration Recommendations:

• Embed absence determination protocols early in AOP development workflows rather than as retrospective additions
• Develop integrated databases that link AOP elements with taxonomic conservation evidence from multiple sources
• Create visualization tools that simultaneously represent phylogenetic relationships and conservation patterns for AOP components

Translational Recommendations:

• Establish best practices for using absence determinations in chemical safety assessment and drug development decision-making
• Develop regulatory guidance on acceptable evidence for true absence designations in toxicological context
• Create training resources to build capacity in absence determination methodologies across research communities

The systematic differentiation between true absence and lack of data represents more than a technical challenge—it constitutes a fundamental requirement for robust, reliable, and responsible extrapolation of adverse outcome pathways across taxonomic boundaries. By implementing the rigorous methodologies outlined in this framework, researchers can transform absence ambiguity from a source of uncertainty to a structured component of evidence-based decision-making in toxicological science and drug development.

The taxonomic domain of applicability (tDOA) defines the biological space—the species and taxa—within which a defined Key Event Relationship (KER) is considered biologically plausible [9]. Within the Adverse Outcome Pathway (AOP) framework, which structures mechanistic toxicological knowledge from a Molecular Initiating Event (MIE) to an Adverse Outcome (AO), accurately defining the tDOA is critical for reliable cross-species extrapolation in ecological and human health risk assessment [15]. A persistent and central challenge is managing the inherent tension between claiming a broad tDOA to maximize the utility of existing data for prediction and ensuring biological realism by acknowledging evolutionary divergence and taxonomic specificity.

Overly broad tDOA claims, while useful for screening-level assessments, risk generating false positives in toxicity predictions by assuming pathway conservation where it does not exist. Conversely, an overly narrow tDOA can lead to false negatives and a failure to protect susceptible species, unnecessarily complicating risk assessment and demanding extensive new animal testing [9]. This technical guide, framed within the broader thesis on KER taxonomic conservation, addresses this challenge. It provides a methodological roadmap for researchers and drug development professionals to systematically evaluate, evidence, and bound the tDOA of their KERs. The goal is to achieve a defensible balance that supports the use of New Approach Methodologies (NAMs) while maintaining scientific credibility and regulatory acceptance [15] [9].

Foundational Principles: Specificity, Conservation, and False Signals

The Spectrum of Specificity in KERs

A KER's tDOA exists on a spectrum. At one end are highly conserved relationships, often rooted in fundamental cellular processes (e.g., oxidative phosphorylation, DNA repair) shared across vast taxonomic groups. At the other end are taxon-specific KERs, dependent on unique anatomical features, receptor subtypes, or metabolic pathways found only in certain clades. The core task is to determine where a given KER falls on this spectrum. This requires moving beyond assumptions based solely on phylogenetic relatedness and towards evidence-based assessments of the conservation of the specific molecular targets and pathways involved [9].

The Peril of False Positives and Boundary Effects

The consequences of insufficient specificity are not merely theoretical. In population genetics, analogous problems arise when inferring range expansions from genetic data. Boundary effects in spatially structured populations—where genetic drift is stronger at distribution edges—can create clinal patterns in genetic indices (like the directionality index, ψ) that mimic the signatures of a true range expansion, leading to high false positive rates if not properly accounted for [39]. This is a powerful analogue for tDOA assessment: a superficial pattern (e.g., a toxic response in several tested species) can create a false signal of broad conservation. Just as population geneticists must normalize ψ against overall genetic structuring (FST) to identify true expansion signals [39], toxicologists must evaluate KER conservation against the background of known molecular and pathway divergence to avoid over-extrapolation.

Quantitative vs. Biological Plausibility

A tDOA can be supported by two complementary lines of evidence:

• Quantitative Empirical Evidence: Direct observation of the KER in a defined set of tested species.
• Biological Plausibility Evidence: In silico or in vitro data indicating the conservation of critical KER components (e.g., protein sequences, pathway nodes) in a wider set of species [9].

A robust tDOA description requires both. The biologically plausible tDOA, often wider than the empirically demonstrated one, must be carefully constructed and transparently documented to prevent overreach [15].

Methodological Framework for Assessing and Bounding tDOA

A systematic, multi-tool approach is essential to manage specificity. The following workflow integrates established computational NAMs to build a weight of evidence.

Core Computational New Approach Methodologies (NAMs)

1. Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS):

• Function: This tool predicts potential chemical susceptibility across species by comparing the primary through quaternary structure of a defined protein target (the MIE or KE target) [9].
• Protocol for tDOA Assessment:
  • Input: Obtain the amino acid sequence (or DNA sequence) of the critical protein (e.g., a specific nuclear receptor, enzyme) from a trusted source (e.g., UniProt, NCBI) for a reference species (typically human or a model organism where the KER is established).
  • Alignment & Tiered Analysis: The tool performs tiered assessments:
    • Tier 1: Compares primary amino acid sequence identity/similarity across species.
    • Tier 2: Assesses conservation of functional domains and motifs.
    • Tier 3: Evaluates conservation of key residues known for chemical interaction (e.g., ligand-binding pocket, active site) through homology modeling [9].
  • Output Interpretation: A "susceptibility prediction" is generated for species in the database. High conservation across all tiers provides strong evidence for inclusion in the tDOA. Divergence in critical functional residues provides a biologically grounded basis for exclusion.

2. Genes-to-Pathways Species Conservation Analysis (G2P-SCAN):

• Function: This tool moves beyond single proteins to assess the conservation of entire biological pathways. It maps a set of human genes (representing a KER pathway) to biological pathways in the Reactome database and evaluates their conservation across a core set of model species [15] [9].
• Protocol for tDOA Assessment:
  • Input: Compile the set of human genes implicated in the KER (from the MIE through intermediate KEs).
  • Pathway Mapping & Analysis: G2P-SCAN maps the gene set to Reactome pathways and calculates a Pathway Conservation Score (PCS) for each species. The PCS considers the proportion of pathway components present and their network topology.
  • Output Interpretation: A high PCS indicates the functional pathway is likely conserved, supporting a broader tDOA. A low PCS suggests pathway divergence or loss, arguing for a narrower tDOA. It provides a systems-level check on SeqAPASS predictions [9].

Integrating Evidence: The Weight-of-Evidence Synthesis

The outputs from SeqAPASS and G2P-SCAN must be integrated with existing empirical data and AOP knowledge.

• Consilience: Strong agreement between SeqAPASS (target conservation), G2P-SCAN (pathway conservation), and empirical data creates a robust, defensible tDOA.
• Discordance: If tools disagree (e.g., SeqAPASS shows target conservation but G2P-SCAN shows low pathway conservation), it flags a need for deeper investigation. The tDOA should be conservatively bounded until the discordance is resolved.
• Documentation: The tDOA statement must explicitly list the evidence source for included and excluded taxa (e.g., "Included in tDOA based on SeqAPASS Tier 3 conservation of ESR1 ligand-binding domain" or "Excluded from tDOA due to absence of pathway per G2P-SCAN analysis").

Case Study: Extending the tDOA for a Reproductive Toxicity AOP Network

A 2024 study provides a paradigm for this balanced approach. Researchers aimed to extend the tDOA of AOP 207, which describes silver nanoparticle (AgNP)-induced reproductive toxicity via oxidative stress in C. r elegans [15].

1. Initial Position: The empirical tDOA was narrow: primarily C. elegans and some limited in vitro human data.

2. Methodology Application:

• SeqAPASS: Used to evaluate conservation of critical targets (e.g., proteins involved in oxidative stress response like NADPH oxidases).
• G2P-SCAN: Used to assess conservation of the mapped Reactome pathways (e.g., "Cellular responses to stress") across species.
• Bayesian Network Modeling: Applied to the assembled cross-species data to quantitatively assess the confidence and strength of the KERs within the AOP network [15].

3. Outcome: The integrated computational analysis allowed the authors to propose a biologically plausible tDOA extending to over 100 taxonomic groups, including fungi, birds, and rodents, far beyond the empirically tested ones. This extension was not a blanket claim but was supported by specific evidence on pathway and target conservation [15].

Table 1: Key Quantitative Outcomes from the AgNP AOP Case Study [15]

Analysis Component	Tool/Method Used	Key Quantitative Output	Interpretation for tDOA
Molecular Target Conservation	SeqAPASS	High sequence identity (>80%) and functional domain conservation for oxidative stress targets across diverse taxa.	Supported inclusion of vertebrates and invertebrates in the plausible tDOA.
Pathway Conservation	G2P-SCAN	High Pathway Conservation Score (PCS > 0.7) for "Cellular response to stress" in core model species (zebrafish, fruit fly).	Induced that the overarching pathway is functionally conserved, strengthening cross-species plausibility.
KER Confidence	Bayesian Network	Probabilistic strength for key KERs (e.g., "Oxidative stress leads to apoptosis") was robust when integrating human in vitro and C. elegans in vivo data.	Provided quantitative confidence for extrapolating the KER structure across species within the proposed tDOA.

Table 2: Research Reagent Solutions for tDOA Investigations

Item / Resource	Category	Function in tDOA Management
SeqAPASS (Web Tool)	Computational NAM	Predicts protein target conservation and potential chemical susceptibility across species using sequence and structural data [9].
G2P-SCAN (R Package)	Computational NAM	Evaluates conservation of biological pathways (Reactome) from human gene sets across model species, providing a systems-level view [15] [9].
AOP Wiki	Knowledge Repository	Central database for published AOPs and KERs; provides the structured framework to which tDOA evidence must be anchored [15].
Comparative Tissue Biobanks	Biological Material	Provide preserved tissues from multiple species for in vitro or ex vivo assays (e.g., receptor binding, gene expression) to generate empirical conservation data.
Phylogenetic Analysis Software	Computational Tool	Allows construction of phylogenetic trees based on target gene sequences, visually contextualizing conservation data within evolutionary relationships.
Defined Reference Chemicals	Chemical Reagent	Chemicals with well-characterized, specific modes of action (agonists, antagonists) are essential for testing KER performance across different species' models.

Experimental Protocols for Generating tDOA Evidence

1In VitroCross-Species Assay Protocol

Purpose: To empirically test a specific Molecular Initiating Event (e.g., receptor activation) across species to bound the tDOA. Materials: Cell lines or primary cells from multiple species (human, rat, zebrafish, etc.), reference agonist/antagonist, reporter assay kit (e.g., luciferase), cell culture reagents. Procedure:

• Cell Preparation: Culture cells from each target species under optimal conditions.
• Transfection: Transfect cells with a reporter construct responsive to the target pathway (if applicable).
• Dosing: Expose cells to a logarithmic concentration series of the reference chemical.
• Measurement: Quantify the MIKE (e.g., receptor activation) using the reporter assay or a direct binding assay (e.g., SPR, competitive binding).
• Data Analysis: Generate dose-response curves. Calculate EC50/IC50 values. A difference in potency >1-2 orders of magnitude may indicate a meaningful taxonomic boundary for that specific MIE.

2In SilicotDOA Refinement Protocol

Purpose: To computationally propose a biologically plausible tDOA using SeqAPASS and G2P-SCAN. Materials: Protein sequence of the key molecular target (FASTA format), list of human genes comprising the KER pathway, access to SeqAPASS web tool and G2P-SCAN R package. Procedure:

• SeqAPASS Analysis:
  • Input the reference protein sequence into SeqAPASS.
  • Run the Tier 1, 2, and 3 analyses.
  • Export the list of species predicted "susceptible" (high conservation) and "not susceptible" (low conservation).
• G2P-SCAN Analysis:
  • Input the human gene list into G2P-SCAN.
  • Run the pathway mapping and conservation analysis for the core model species set.
  • Record the Pathway Conservation Score for each species.
• Evidence Integration:
  • Create a master table listing species of interest.
  • Populate columns with SeqAPASS prediction (Y/N) and G2P-SCAN PCS.
  • Apply a decision rule (e.g., include in tDOA if SeqAPASS = Y and PCS > 0.6). Manually review and justify any exceptions.

Balancing broad tDOA claims with biological realism is not a one-time exercise but a dynamic, evidence-driven process. By adopting the integrated methodological framework presented here—leveraging computational NAMs like SeqAPASS and G2P-SCAN within the AOP paradigm—researchers can replace assumption-based extrapolation with evidence-bounded extrapolation. The resulting tDOA statements are both more scientifically defensible and more useful for regulatory application. They enable confident use of data across species where justified, flag potential vulnerabilities in untested taxa, and strategically focus precious resources for empirical testing on true taxonomic boundaries. In doing so, they advance the core mission of KER taxonomic conservation research: to build a predictive toxicology capable of protecting biological diversity based on a deep understanding of biological unity and difference.

The expansion of the taxonomic domain of applicability (tDOA) for Adverse Outcome Pathways (AOPs) is a cornerstone for advancing next-generation, animal-sparing risk assessment. This requires the strategic integration of disparate evidence streams. This technical guide details a cohesive methodology for harmonizing in vitro bioactivity, in vivo phenotypic anchoring, and in silico cross-species extrapolation data. Framed within research on Key Event Relationship (KER) taxonomic conservation, we present a stepwise workflow from target identification to tDOA validation. The protocol leverages computational New Approach Methodologies (NAMs), including the SeqAPASS and G2P-SCAN tools, to build a weight-of-evidence for pathway conservation [9] [15]. A case study on antidiabetic phytochemicals demonstrates the quantitative correlation of in vitro enzyme inhibition (IC₅₀: 55.08–246.5 μg/mL) with in silico binding affinity and positive in vivo outcomes [40]. This integrative framework provides researchers and drug development professionals with a standardized, predictive approach for establishing ecologically and toxicologically relevant tDOAs.

The Adverse Outcome Pathway (AOP) framework provides a mechanistic bridge between a Molecular Initiating Event (MIE) and an Adverse Outcome (AO) through a series of Key Events (KEs). A critical, yet often poorly defined, element of an AOP is its Taxonomic Domain of Applicability (tDOA)—the range of species for which the described KERs are biologically plausible [15]. Explicitly defining the tDOA is essential for the reliable application of AOPs in chemical safety assessment across ecological and human health contexts under a One Health perspective [15].

Traditional tDOA definition relies on limited in vivo toxicity data from standard model organisms, creating significant uncertainty for extrapolation. Research into KER taxonomic conservation seeks to solve this by determining which key relationships in a toxicological pathway are conserved across phylogeny. This demands the integration of diverse data types:

• In vitro assays identify MIEs and measure bioactivity in controlled systems.
• In vivo studies anchor these molecular effects to apical outcomes in whole organisms.
• In silico analyses predict the conservation of molecular targets and pathways across the tree of life.

Harmonizing these disparate data streams into a cohesive evidentiary package is the key to robustly and confidently expanding tDOAs, thereby reducing dependency on animal testing and improving risk predictions for untested species [9] [15].

Core Methodologies: A Triangulated Approach

A robust tDOA assessment is built on three pillars of evidence, each with standardized protocols.

1In VitroBioactivity Profiling

In vitro systems provide the foundational data on chemical-target interactions.

• Objective: To quantitatively measure the potency of a stressor (e.g., chemical, nanoparticle) at a defined molecular target.
• Protocol (Enzyme Inhibition Assay) [40]:
  • Reaction Setup: In a 96-well plate, mix the target enzyme (e.g., α-amylase, α-glucosidase) with the test compound at a range of concentrations. Use a standard inhibitor (e.g., acarbose) as a positive control and a vehicle control.
  • Substrate Addition: Initiate the reaction by adding the specific enzyme substrate (e.g., soluble starch for α-amylase; p-nitrophenyl-α-D-glucopyranoside for α-glucosidase).
  • Incubation: Incubate the plate at a defined temperature (e.g., 37°C) for a precise period to allow the enzymatic reaction.
  • Reaction Termination & Measurement: Stop the reaction with a stopping reagent (e.g., 3,5-dinitrosalicylic acid for α-amylase; Na₂CO₃ solution for α-glucosidase). Measure the absorbance of the product using a microplate reader.
  • Data Analysis: Calculate percentage inhibition relative to the control. Plot dose-response curves and determine the half-maximal inhibitory concentration (IC₅₀) using nonlinear regression analysis.

2In SilicoCross-Species Extrapolation

Computational tools predict the conservation of molecular targets and pathways.

• Objective: To assess the potential for a molecular interaction (MIE) to occur in species beyond the tested model.
• Protocol (Integrated SeqAPASS & G2P-SCAN Analysis) [9] [15]:
  • Target Identification: Identify the specific protein target of the MIE (e.g., Peroxisome Proliferator-Activated Receptor Alpha, PPARα) using literature, ToxCast data, or molecular docking results [9].
  • SeqAPASS Analysis:
    • Input the amino acid sequence or accession number of the reference protein (e.g., human) into the SeqAPASS tool.
    • Perform sequence alignment across taxa. Assess conservation of critical amino acid residues known to be essential for chemical binding or protein function.
    • Generate a prediction of potential susceptibility across hundreds of species based on sequential and structural similarity.
  • G2P-SCAN Analysis:
    • Input the human gene symbol of the target into the G2P-SCAN tool.
    • The tool maps the gene to its associated biological pathways (e.g., via Reactome database).
    • It evaluates the conservation of the entire pathway across a core set of model species (human, mouse, rat, zebrafish, fruit fly, worm, yeast), providing a systems-level view of potential conservation.
  • Evidence Integration: Combine SeqAPASS (target-level) and G2P-SCAN (pathway-level) results to create a weight-of-evidence for taxonomic applicability. High conservation at both levels strongly supports a broader tDOA.

3In VivoPhenotypic Anchoring

In vivo studies confirm the pathway leading from the MIE to the AO.

• Objective: To empirically link molecular perturbations to adverse outcomes in a whole organism.
• Protocol (Rodent Model for Metabolic Disruption) [40]:
  • Model Induction: Induce a disease state (e.g., Type 2 diabetes) in experimental animals (e.g., mice) via intraperitoneal injection of Streptozotocin (STZ) at a dose of 50-60 mg/kg for multiple days.
  • Treatment: Administer the test compound (e.g., plant extract or purified phytochemical) to the treatment group via oral gavage daily for a defined period (e.g., 28 days). Maintain vehicle-control and disease-control groups.
  • Endpoint Monitoring: Regularly measure physiological endpoints such as blood glucose levels and body weight. Collect terminal blood samples for plasma insulin and lipid profile analysis.
  • Tissue Analysis: Harvest organs (liver, kidney, pancreas). Perform histopathological examination to assess tissue damage. Homogenize tissues to measure oxidative stress markers (e.g., MDA, SOD, CAT, GSH).
  • Statistical Correlation: Correlate the improvement in phenotypic endpoints (e.g., reduced hyperglycemia) with molecular biomarkers (e.g., increased antioxidant activity) to establish a quantitative KER.

Table 1: Quantitative Data from an Integrated Antidiabetic Study [40]

Assay Type	Target/Endpoint	Test Compound	Result (Mean ± SD or IC₅₀)	Control (Acarbose)
In Vitro	α-Amylase Inhibition	Cicer arietinum extract	55.08 μg/mL	196.3 ± 10 μg/mL
In Vitro	α-Amylase Inhibition	Hordeum vulgare extract	115.8 ± 5 μg/mL	196.3 ± 10 μg/mL
In Vitro	α-Glucosidase Inhibition	Cicer arietinum extract	100.2 ± 5 μg/mL	246.5 ± 10 μg/mL
In Vitro	α-Glucosidase Inhibition	Hordeum vulgare extract	216.2 ± 5 μg/mL	246.5 ± 10 μg/mL
In Silico	Molecular Docking (α-Amylase)	Medicagol	Strong binding affinity (specific score not provided)	N/A
In Vivo	Blood Glucose Reduction	C. arietinum extract	Significant reduction in STZ-mice	N/A
In Vivo	Antioxidant Activity (Liver)	C. arietinum extract	Increased SOD, CAT, GSH; decreased MDA	N/A

Strategy for Data Integration and tDOA Expansion

The individual data streams must be logically synthesized. A Bayesian network (BN) modeling approach is particularly effective for integrating heterogeneous data and managing uncertainty in KERs [15].

• AOP Network Development: Collate all evidence into a preliminary AOP network linking MIEs, KEs, and AOs. For example, AgNP-induced ROS generation (MIE) → MAPK activation (KE) → reproductive failure (AO) [15].
• Quantitative KER Modeling: Use BN software to construct a probabilistic model. Populate the model with conditional probabilities derived from the experimental data (e.g., probability of enzyme inhibition given a certain concentration, probability of hyperglycemia given enzyme inhibition).
• Sensitivity Analysis: The BN identifies which KERs have the strongest influence on the AO, highlighting the most critical pathways for conservation analysis.
• tDOA Expansion:
  • Use the in silico tools (SeqAPASS, G2P-SCAN) on the primary molecular targets identified in the most sensitive KERs.
  • The outputs provide a list of taxa where the target and pathway are conserved, thereby expanding the biologically plausible tDOA from a few model species to potentially over 100 taxonomic groups [15].
  • This expanded tDOA can be visually represented as a network, moving from a core of empirical data to a periphery of predicted applicability.

Table 2: tDOA Extension for a Hypothetical AOP [15]

Ecological Compartment	Initial tDOA (Empirical Data)	Extended tDOA (In Silico Prediction)
Terrestrial	Caenorhabditis elegans, Drosophila melanogaster	Fungi (98 species), Birds (28 species), Rodents, Reptiles
Aquatic	Danio rerio (zebrafish)	Bony fishes (multiple orders), Amphibians

Visualization of Workflows and Pathways

Integrated Data Workflow for tDOA

Cross-Species AOP Extrapolation via tDOA

Table 3: Key Research Reagent Solutions for Integrated tDOA Studies

Item	Category	Function in tDOA Research	Example/Supplier
α-Amylase/α-Glucosidase Assay Kits	In Vitro Reagent	Measures inhibitory potential of chemicals against carbohydrate-digesting enzymes, defining potency for an MIE [40].	Sigma-Aldrich, Global Scientific [40]
Streptozotocin (STZ)	In Vivo Reagent	Chemical inducer of diabetes in rodent models, used for phenotypic anchoring of metabolic disruptors [40].	Sigma-Aldrich [40]
Acarbose	In Vitro/In Vivo Control	Standard inhibitor drug used as a positive control in enzyme inhibition assays and in vivo studies [40].	Pharmaceutical grade
AutoDock Vina, GOLD	In Silico Software	Performs molecular docking to predict binding affinity and mode of a ligand to a protein target, informing the MIE [40].	Open Source / Commercial
SeqAPASS Web Tool	In Silico Tool	Predicts protein susceptibility and conservation across species via sequence alignment, core to tDOA expansion [9] [15].	US EPA (Publicly available)
G2P-SCAN R Package	In Silico Tool	Evaluates the conservation of entire biological pathways across model species, providing systems-level evidence [9] [15].	Publicly available
Bayesian Network Software	Data Analysis Tool	(e.g., Netica, GeNIe) Integrates probabilistic data from different streams to model KERs and quantify uncertainty [15].	Commercial & Open Source
Reference Protein Structures	In Silico Data	High-resolution 3D structures (e.g., from PDB ID 1B2Y for α-amylase) are essential for molecular docking studies [40].	RCSB Protein Data Bank

Discussion and Future Perspectives

The harmonization of in vitro, in vivo, and in silico evidence is not sequential but iterative. In silico predictions can prioritize in vitro testing on non-standard species cell lines, the results of which can refine computational models. The ultimate goal is a predictive, pathway-based framework where a well-defined MIE and its associated KERs, supported by strong conservation evidence, can be used to anticipate AOs in a wide range of species within the tDOA with high confidence.

Future advancements will depend on:

• Standardized Reporting: Adopting common data standards for all three evidence types to facilitate automated integration.
• High-Throughput In Vitro Phylogenetics: Developing scalable cell-based assays from diverse species to ground-truth computational predictions.
• Quantitative AOP (qAOP) Development: Moving from qualitative networks to quantitative models that can predict the magnitude of effect across species, which is essential for regulatory risk assessment.

By adopting the integrative framework outlined here, researchers can systematically build and expand the tDOA of AOPs, transforming them from descriptive models into powerful, predictive tools for chemical safety evaluation in the 21st century.

Validation and Confidence: Assessing Weight of Evidence and Comparative Case Studies

Applying Modified Bradford-Hill Criteria to Evaluate tDOA Evidence

Evaluating the taxonomic domain of applicability (tDOA) of an Adverse Outcome Pathway (AOP) is a fundamental challenge in modern regulatory toxicology and chemical safety assessment. The tDOA defines the range of species for which the causal relationships described within an AOP—a structured sequence of events linking a molecular perturbation to an adverse outcome—are considered biologically plausible and operative [41]. Establishing a robust tDOA is critical for cross-species extrapolation, a core component of the One Health approach that seeks to protect human, animal, and environmental health in an integrated manner [41].

The central building block of an AOP is the Key Event Relationship (KER), which describes a scientifically supported, causal link between an upstream and a downstream Key Event (KE) [11] [10]. The confidence in any AOP, and by extension its tDOA, hinges entirely on the collective weight of evidence for its constituent KERs [11] [10]. This technical guide proposes the application of a modified set of Bradford-Hill (BH) "viewpoints"—originally formulated for epidemiological causation—as a rigorous, structured framework to assess the evidence supporting the taxonomic conservation of KERs [42] [43]. By adapting this framework to the context of comparative biology and pathway conservation, researchers can systematically evaluate tDOA evidence, moving beyond assumptions based solely on phylogenetic proximity to a more mechanistic, evidence-based determination of applicable taxa.

Theoretical Foundation: Modern Interpretations of Bradford-Hill Viewpoints

Sir Austin Bradford Hill proposed nine "viewpoints" (often termed "criteria") to guide the assessment of whether an observed association might reflect a causal relationship [44]. He emphasized they were not a checklist but considerations to weigh [42] [45]. Modern causal thinking, built on the potential outcomes framework, has refined the application and interpretation of these viewpoints [42] [45]. For the specialized task of evaluating tDOA, a subset of these viewpoints is particularly relevant, and their interpretation requires modification to address questions of biological conservation across species.

The following table outlines the traditional BH viewpoints, their modern reinterpretation in light of contemporary causal inference frameworks like Directed Acyclic Graphs (DAGs) and Sufficient-Component Cause (SCC) models, and their proposed modification for application to tDOA assessment for KERs [42] [43] [45].

Table 1: Modification of Bradford-Hill Viewpoints for tDOA Assessment of Key Event Relationships

Bradford-Hill Viewpoint	Modern Interpretation & Role in Causal Inference	Modified Application to KER Taxonomic Conservation (tDOA)
Strength of Association	A strong association is less likely to be fully explained by unmeasured confounding. Statistical significance and effect size are considered [43].	The degree of evolutionary conservation of the molecular sequence (e.g., protein target) and the functional response of the intervening biological pathway across taxa. Strong, conserved sequence-structure-function relationships support a broader tDOA.
Consistency	Reproducible findings across different studies, locations, and populations. In modern practice, consistency is also sought across different types of evidence (e.g., epidemiological, in vitro, in vivo) [43].	Observation of the KER (upstream KE leads to downstream KE) across multiple, taxonomically diverse species. Consistency in the direction and essential nature of the relationship strengthens tDOA evidence.
Specificity	Considered rare in multifactorial disease etiology. A more useful modern concept is the use of "negative controls" or falsification analyses [42] [45].	Demonstration that the downstream KE does not occur in taxonomic groups where the upstream molecular target or essential pathway component is legitimately absent or non-functional. This helps define the boundaries of the tDOA.
Plausibility	Biological plausibility is informed by current knowledge. DAGs and SCC models help articulate plausible mediating pathways and component interactions [42].	Biological plausibility for conservation is based on established principles of evolutionary biology, comparative genomics, and the essentiality of the pathway for conserved physiological functions.
Coherence	The causal interpretation should not conflict with generally known facts of the natural history of the disease [44].	The hypothesized tDOA should be coherent with known phylogenetic relationships, life histories, and ecological/physiological adaptations of the species in question.
Experiment	Evidence from experimental interventions (e.g., randomized trials) provides the strongest support for causality [42].	Experimental evidence demonstrating that modulation of the upstream KE (e.g., via chemical inhibition, genetic knockout) prevents or alters the downstream KE in multiple species. This is a powerful line of evidence for KER essentiality and conservation.
Analogy	Reasoning based on similar, established cause-effect relationships [44].	Inference of KER conservation in a new taxon based on its established operation in a well-studied surrogate species, considering analogous anatomical structures, physiological processes, and molecular pathways.

Application: A Framework for Evaluating tDOA Evidence

Applying the modified BH viewpoints to tDOA evaluation involves a sequential, evidence-weighted process. The workflow begins with the definition of the KER of interest and proceeds through the assembly and assessment of evidence for the conservation of its biological underpinnings [10] [9].

Diagram 1: Workflow for Applying Modified BH Viewpoints to tDOA Assessment (Max. 760px)

Step 1: Identify Essential Molecular Target/Pathway The evaluation begins by deconstructing the KER to identify the essential molecular target(s) (e.g., a specific enzyme, receptor, or ion channel) and the biological pathway that mechanistically links the upstream and downstream Key Events. This is the foundational unit for conservation analysis [9].

Step 2: Assess Taxonomic Conservation of Target This step investigates the strength, plausibility, and coherence of target conservation. Computational New Approach Methodologies (NAMs) are critical here. The US EPA's SeqAPASS tool analyzes protein sequence and structural similarity across species to predict potential chemical susceptibility, providing a line of evidence for the conservation of the molecular initiating event [9]. Complementary tools like G2P-SCAN map human gene targets to biological pathways (e.g., Reactome) and assess the conservation of those entire pathways across a core set of model species [9]. High sequence similarity in critical functional domains and conservation of core pathway architecture support a broader tDOA.

Step 3: Evaluate Functional Conservation of KER This step gathers evidence for consistency and experiment. It involves reviewing empirical data demonstrating that the causal relationship described in the KER holds in multiple species. This includes in vivo or in vitro studies showing that perturbation of the upstream KE leads to the downstream KE in taxonomically diverse organisms [11] [43]. Dose-response data (biological gradient) within a species further strengthens the causal claim for that species, while consistent directional effects across species bolster the case for conservation.

Step 4 & 5: Integrate Evidence and Test Boundaries All evidence is integrated to propose a preliminary tDOA. The final, critical step is to apply the principle of specificity by actively seeking falsification evidence. Are there taxonomic groups related to those within the proposed tDOA that legitimately lack the molecular target or pathway? If the KER is claimed to be broadly conserved, evidence of its absence in a well-studied species (where confounding factors are ruled out) would sharply delineate the tDOA boundary [42].

Experimental and Computational Protocols

Generating and compiling evidence for tDOA requires both empirical biology and bioinformatics. The following table details key experimental and computational protocols relevant to assessing different BH viewpoints.

Table 2: Protocols for Generating tDOA Evidence Aligned with BH Viewpoints

BH Viewpoint	Experimental/Computational Protocol	Objective & Relevance to tDOA
Strength & Plausibility	SeqAPASS Analysis: Input the amino acid sequence of the protein target from a reference species (e.g., human). The tool performs tiered assessments (primary, secondary, tertiary) comparing sequence, domain, and active site conservation across species in its database [9].	Provides quantitative data on protein conservation. High percent identity/alignment scores in functional domains provide strength for the hypothesis of conserved molecular interaction. The biological plausibility of extrapolation is grounded in evolutionary biology.
Plausibility & Coherence	G2P-SCAN Pathway Analysis: Input human gene symbols for the molecular target and associated pathway components. The tool maps genes to Reactome pathways and evaluates the conservation of the pathway architecture and gene-content across seven core model organisms [9].	Moves beyond single-protein conservation to assess the plausibility of the entire intervening pathway being conserved. Results should cohere with known phylogenetic relationships and physiological adaptations.
Consistency & Experiment	Multi-Species In Vitro Assay: Employ standardized cell-based assays (e.g., reporter gene assays, high-content imaging) using primary cells or cell lines from multiple species to measure the downstream KE response to modulation of the upstream KE [43] [41].	Provides direct experimental evidence for the functional operability of the KER across species. Consistency in the response direction and potency across taxa is powerful supporting evidence.
Experiment & Analogy	Essentiality Testing (e.g., CRISPR/Cas9): Use genetic knockout or knockdown of the upstream KE target in embryo or adult models of multiple species (e.g., zebrafish, mouse) and assess the impact on the downstream KE and adverse outcome [11].	Provides the strongest possible experimental evidence for the KER's essential role. Successful analogy from one model organism to another is supported if the same intervention produces comparable phenotypic results.
Specificity	Negative Control / Falsification Analysis: Intentionally investigate species groups (e.g., insects, mollusks for a vertebrate-specific hormone receptor) where the target pathway is known to be absent or fundamentally different. Confirm the absence of the KER response [42].	Actively tests the boundaries of the tDOA. The absence of effect where the mechanism is absent provides high confidence in the specificity of the KER to taxa possessing the conserved mechanism.

The integration of computational and empirical data is best visualized in a converging lines-of-evidence model.

Diagram 2: Integration of Evidence Streams for tDOA Confidence (Max. 760px)

Successfully applying this framework requires a suite of specialized databases, software tools, and experimental resources.

Table 3: Research Toolkit for tDOA Evidence Evaluation

Tool/Resource Name	Type	Primary Function in tDOA Evaluation	Relevant BH Viewpoints
SeqAPASS (Sequence Alignment to Predict Across Species Susceptibility)	Computational Tool / NAM	Predicts potential chemical susceptibility across species by analyzing conservation of protein sequences, functional domains, and active sites [9].	Strength, Plausibility
G2P-SCAN (Genes to Pathways - Species Conservation Analysis)	Computational Tool / NAM	Maps human gene sets to biological pathways and evaluates the conservation of those pathways across seven model species, providing pathway-level context [9].	Plausibility, Coherence, Analogy
AOP-Wiki (aopwiki.org)	Knowledgebase	The central repository for published AOPs, KEs, and KERs. Provides the structured descriptions and existing evidence that form the starting point for tDOA analysis [11] [10].	All (Foundation)
Reactome (reactome.org)	Pathway Database	A curated, peer-reviewed database of human biological pathways. Serves as a reference for pathway architecture used by tools like G2P-SCAN to assess conservation [9].	Plausibility
OECD AOP Developers' Handbook	Guidance Document	Provides formal guidance on AOP development, including weight-of-evidence assessment and considerations for taxonomic applicability [10].	All (Framework)
In vitro Bioactivity Data (e.g., ToxCast/Tox21)	Empirical Data	High-throughput screening data showing chemical effects on molecular targets. Can be used to identify potential molecular initiating events and assess conservation of target response [9].	Experiment, Consistency
Ortholog Databases (e.g., Ensembl Compara, OrthoDB)	Bioinformatics Database	Provide predictions of orthologous genes (genes diverged after a speciation event) across species, which are crucial for correct cross-species comparisons [9].	Strength, Plausibility, Coherence

The determination of a toxicological pathway's taxonomic domain of applicability is a critical inference with major implications for ecological risk assessment, chemical regulation, and the reduction of animal testing through cross-species extrapolation. By adapting the time-tested Bradford-Hill viewpoints to the specific question of KER conservation, researchers gain a structured, transparent, and scientifically defensible framework for tDOA evaluation. This modified approach moves beyond qualitative guesswork, demanding convergent evidence from computational predictions of conservation (strength, plausibility), empirical demonstrations of functional operability across species (consistency, experiment), and deliberate testing of proposed boundaries (specificity). As the AOP knowledgebase expands and computational NAMs become more sophisticated, the systematic application of this framework will be essential for building confidence in pathway-based safety assessments and realizing the promise of 21st-century toxicology.

In scientific assessments for environmental conservation, human health, and drug development, decision-making is rarely supported by a single, definitive study. Instead, it relies on synthesizing multiple lines of evidence of varying types and quality [46]. The Weight of Evidence (WoE) approach is a structured process for integrating this diverse evidence to determine the relative support for possible answers to a scientific or risk assessment question [47]. Critically, a robust WoE argument does not choose between quantitative (quant) and qualitative (qual) data but strategically integrates both to leverage their complementary strengths [48] [49].

Quantitative data provides objective, numerical measurements that answer "how many," "how much," or "how often," enabling statistical analysis and generalization [50] [51]. Qualitative data provides descriptive, contextual information that explores "why" and "how," uncovering meanings, mechanisms, and subjective experiences [48] [49]. In the context of Key Event Relationship (KER) taxonomic conservation research—which investigates the relationships between stressors, biological key events, and adverse outcomes in species and ecosystems—this integration is paramount. Conservation decisions must consider not only population statistics (quantitative) but also behavioral observations, genetic purity, and ecological context (qualitative) [52] [53].

This guide outlines a framework for building a defensible WoE argument by systematically assembling, weighing, and integrating quantitative and qualitative evidence, with a focus on applications in taxonomic conservation and biomedical research.

Foundational Concepts: Quantitative vs. Qualitative Data

Understanding the inherent characteristics and appropriate applications of each data type is the first step in their integration.

Table 1: Core Characteristics of Quantitative and Qualitative Data and Research [48] [49] [50].

Characteristic	Quantitative Data & Research	Qualitative Data & Research
Nature of Data	Numerical, measurable, countable [50].	Descriptive, involving words, images, or observations [51].
Core Question	What? How many? How much? How often? [48].	Why? How? What is the experience? [48].
Research Goal	To test hypotheses, measure variables, establish patterns, and generalize [51].	To explore ideas, understand concepts, experiences, and generate deep insights [51].
Sample & Design	Large samples for statistical power; structured and predetermined design [51].	Small, focused samples for depth; flexible and iterative design [51].
Collection Methods	Surveys, experiments, structured observations, analysis of existing metrics [50].	In-depth interviews, focus groups, participant observation, open-ended surveys [51].
Analysis Approach	Statistical analysis to identify relationships, differences, and trends [51].	Thematic, content, or discourse analysis to identify patterns, themes, and narratives [51].
Output	Statistical significance, effect sizes, predictive models [51].	Detailed descriptions, conceptual frameworks, hypotheses, and illustrative quotes [51].

Advantages and Limitations: Quantitative data excels at providing objective, generalizable, and statistically testable evidence but may miss contextual nuance and underlying causes [50]. Qualitative data provides rich, explanatory depth and is ideal for exploring complex phenomena but is subject to researcher interpretation and is not statistically generalizable [51]. An integrated WoE approach mitigates these individual limitations by using each data type to address the gaps of the other.

Weight of Evidence Frameworks: Integrating Systematic Review with Expert Judgment

The WoE process is more than a simple tally of studies. It is a transparent and structured methodology for assembling and weighing diverse evidence [46]. Best practice integrates the rigorous, bias-minimizing approach of Systematic Review (SR) with the inferential judgment characteristic of traditional WoE [46].

The Integrated SR & WoE Process

The European Food Safety Authority (EFSA) guidance outlines a three-step WoE assessment: (1) assembling evidence, (2) weighing evidence, and (3) integrating evidence to reach a conclusion [47]. Integrating SR principles ensures the assembly phase is comprehensive and unbiased.

Table 2: Integrated SR & WoE Framework, Adapted from Classic Approaches [46] [47].

Assessment Phase	Integrated Activities & Considerations
1. Problem Formulation	Define the specific KER or assessment question. Determine the required lines of evidence (e.g., exposure, toxicity, ecological effect).
2. Assemble Evidence (SR-driven)	Conduct a systematic literature search and screening for all evidence types [46]. Extract data from quant studies (e.g., effect sizes) and qual studies (e.g., themes, mechanistic descriptions). Include grey literature, field data, and expert input where relevant [46].
3. Weigh Evidence	Evaluate each piece of evidence for reliability (methodological quality, risk of bias), relevance (directness to the KER), and consistency (agreement across studies) [47]. Use predefined scoring criteria or ranking (e.g., high, medium, low confidence).
4. Integrate Evidence	Triangulate findings across qualitative and quantitative lines of evidence. Examine if different data types converge (strengthens conclusion), are complementary (provides complete picture), or contradict (requires resolution). Use formal methods (e.g., meta-analysis for quant data) or structured expert judgment (e.g., Hill's criteria) to draw an inference [46].
5. Document and Conclude	Clearly state the conclusion (e.g., "The evidence is sufficient/insufficient to support the KER..."). Articulate the uncertainty and the relative contribution of qualitative and quantitative evidence to the conclusion.

Causal Considerations: Hill's Criteria

For assessing causal KERs (e.g., "Does chemical X cause population decline in species Y?"), Bradford Hill's criteria provide a qualitative-quantitative framework for weighing evidence [46]. These include strength of association (quantitative), consistency across studies (quantitative), specificity, temporality, biological gradient (dose-response, quantitative), plausibility (often qualitative, mechanistic evidence), coherence, experiment, and analogy [46]. Not all criteria must be met, but a WoE judgment considers the pattern across them.

Title: A Weight of Evidence Integration Framework Based on Hill's Criteria

Case Study: Taxonomic Conservation of the Australian Dingo

The conservation challenge of the Australian dingo (Canis familiaris dingo) threatened by hybridization with domestic dogs (C. f. familiaris) exemplifies the need for a WoE approach integrating multiple data types [52].

Assessment Question: What is the genetic purity and conservation status of a dingo population?

Assembling & Weighing Multiple Lines of Evidence

• Quantitative - Genetic Analysis:

  • Method: Collection of tissue/blood samples for DNA analysis. Use of microsatellite markers or single nucleotide polymorphisms (SNPs) to estimate the proportion of domestic dog ancestry in individuals [52].
  • Data & Weight: Provides high reliability and specificity for discriminating hybrid generations (e.g., F1, backcrosses) [52]. Weight is high for determining individual ancestry but may be limited by reference population validity [52].

• Quantitative - Morphometric Analysis:

  • Method: Precise measurement of skull morphology (e.g., cranial width, snout length) from specimens using calipers. Statistical comparison to historical "pure" dingo specimens [52].
  • Data & Weight: Provides objective, historical data. However, its weight is moderate to low because it poorly discriminates beyond pure vs. hybrid and requires dead specimens [52].

• Qualitative - Phenotypic (Coat Colour) Assessment:

  • Method: Field-based visual scoring of coat colour, presence of white markings, ticking, and body form against a "pure" dingo standard [52].
  • Data & Weight: Provides high practical relevance for rapid field assessment. Weight is low to moderate due to subjectivity and poor discrimination (e.g., some dog breeds resemble dingoes) [52].

• Qualitative - Behavioral & Ecological Observation:

  • Method: Long-term field studies observing pack structure, breeding cycles, hunting behavior, and territoriality.
  • Data & Weight: Provides critical contextual relevance about ecological function and reproductive isolation. Weight depends on observer expertise and study duration.

Integration for Conservation Decision-Making

A robust WoE argument for a management plan (e.g., removing hybrids from a conservation area) would not rely on coat colour alone. It would prioritize high-weight genetic evidence to definitively identify hybrids, use morphometric data from culled animals to validate genetic findings, and employ field observations to understand pack dynamics and the ecological impact of removal. The qualitative evidence provides the "why" for conservation actions (ecological role, cultural value), while the quantitative evidence provides the "how much" and "which ones" for tactical decisions.

Title: WoE for Dingo Conservation Integrating Quantitative and Qualitative Lines

Table 3: The Scientist's Toolkit for Dingo Hybridization Assessment [52].

Research Reagent / Tool	Primary Function	Evidence Type Generated
Microsatellite or SNP Panel	Genotyping to quantify proportional ancestry of dingo vs. domestic dog.	Quantitative (high reliability).
Digital Calipers / 3D Scanner	Precise measurement of skull and skeletal morphological traits.	Quantitative (moderate reliability).
Standardized Phenotype Scoring Sheet	Field guide for consistent visual assessment of coat colour, markings, and form.	Qualitative (low-moderate reliability).
GPS Collars & Camera Traps	Monitoring movement, pack interactions, and breeding behavior in situ.	Qualitative/Quantitative (high contextual relevance).
Pre-European Reference Specimens	Historical baseline (bones, skins) for genetic and morphological comparison.	Quantitative & Qualitative (high relevance, scarce).

Application in Biomedical Research & Drug Development

The WoE framework is equally critical in biomedical sciences. Assessing the therapeutic potential of a drug or the hazard of a chemical requires integrating evidence across in vitro assays, animal models, and human studies [46].

• Problem Formulation: Define the KER (e.g., "Compound A inhibits enzyme B, leading to reduced tumor growth in model C").
• Assemble Evidence: Systematically gather quantitative data (e.g., IC50 values, tumor volume measurements, clinical trial endpoints) and qualitative data (e.g., histopathology descriptions, patient-reported outcomes, mechanistic pharmacology).
• Weigh Evidence: Evaluate study quality (e.g., blinding, sample size), relevance (human vs. animal model), and consistency across experimental systems.
• Integrate Evidence: Use WoE to determine if mechanistic data (qualitative/quantitative) supports the biological plausibility of observed in vivo outcomes. A meta-analysis of quantitative clinical data can be combined with a systematic review of qualitative safety reports to form a complete risk-benefit profile [46].

This process moves beyond a single "key study" to build a convincing, holistic argument for regulatory submission or a conservation management plan, explicitly acknowledging the role and limitations of each type of scientific evidence.

The Adverse Outcome Pathway (AOP) framework is a structured representation that connects a Molecular Initiating Event (MIE), through a series of measurable Key Events (KEs), to an Adverse Outcome (AO) relevant to risk assessment [6]. A critical component for the regulatory application of an AOP is defining its Taxonomic Domain of Applicability (tDOA)—the range of species for which the described causal pathway is biologically plausible [6]. For most developed AOPs, the tDOA is narrowly defined, often limited to the single species (e.g., Apis mellifera, the European honey bee) used in the foundational empirical studies [6]. This presents a significant challenge for ecological risk assessment, which must protect a wide diversity of untested species.

Expanding the tDOA relies on evaluating the structural and functional conservation of KEs and their causal relationships (Key Event Relationships, KERs) across taxa [6]. Bioinformatics tools that leverage publicly available protein sequence data provide a powerful, efficient method to generate evidence for structural conservation. The Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool, developed by the U.S. Environmental Protection Agency, is explicitly designed for this purpose [6] [17]. It enables researchers to rapidly extrapolate knowledge of chemical-protein interactions and pathway components from a model species to thousands of others by analyzing protein sequence similarity at multiple levels.

This case study analysis details the process of using SeqAPASS to define the biologically plausible tDOA for a neurotoxic AOP of critical ecological concern: the pathway linking the activation of the nicotinic acetylcholine receptor (nAChR) to colony death/failure in bees (AOP 89) [6]. The analysis demonstrates how computational evidence for protein conservation strengthens the tDOA for individual KEs and KERs, thereby supporting the broader thesis that KER taxonomic conservation is fundamental to credible, widely applicable AOPs for ecological and translational toxicology.

Case Study Background: AOP 89 - nAChR Activation to Colony Failure

Neonicotinoid insecticides, which target the nAChR, have been implicated in the global decline of pollinator populations [6]. AOP 89 was developed to organize the mechanistic understanding of how the MIE (nAChR activation) leads, through intermediate KEs at cellular, organ, and organism levels, to the AO of colony death/failure in Apis mellifera [6].

The initial, empirically derived tDOA for this AOP was restricted primarily to A. mellifera. However, concerns extend to other managed bees (e.g., Apis cerana) and, importantly, to a wide array of non-Apis bees (e.g., bumble bees and solitary bees), which are also vulnerable to pesticide exposure [6]. To evaluate the potential applicability of this AOP across Hymenoptera, a bioinformatics-driven approach was employed to assess the conservation of nine proteins critical to the KEs and KERs within the pathway [6].

Methodology: The SeqAPASS Tool and Analysis Protocol

SeqAPASS is a freely available, web-based tool that performs a hierarchical, three-level evaluation of protein conservation to predict potential chemical susceptibility across species [17]. The following protocol, adapted from the tool's detailed methodology, was applied to the bee neurotoxic AOP case study [17].

SeqAPASS Analysis Protocol

Step 1: Protein Target Identification Nine proteins integral to the neurotoxic AOP were identified from the AOP-Wiki description (AOP 89). These included the primary molecular target (nAChR subunits) and downstream proteins involved in subsequent KEs [6].

Step 2: Sequence Acquisition and Query Submission For each protein, the primary amino acid sequence from Apis mellifera (the "sensitive" model species) was obtained using a standard NCBI protein accession number. This sequence was submitted as the query to the SeqAPASS tool [17].

Step 3: Tiered Evaluation of Conservation

• Level 1 Analysis (Primary Sequence): SeqAPASS compares the full-length query sequence against all available protein sequences in public databases using BLASTp. It calculates a percent identity and an alignment score (bitscore) for each match. A preliminary prediction of susceptibility (i.e., whether the ortholog is likely to interact with the same chemical) is made based on these metrics against a predefined threshold [17].
• Level 2 Analysis (Functional Domains): The tool aligns the conserved functional domains (from the NCBI Conserved Domain Database) of the query protein with those of putative orthologs. Conservation of these domains is critical for maintaining protein function within a pathway [6] [17].
• Level 3 Analysis (Critical Residues): This most refined level assesses the conservation of specific amino acid residues known to be essential for chemical-protein binding (e.g., neonicotinoid binding sites on nAChR) or protein-protein interaction. This level provides the highest taxonomic resolution for predicting susceptibility [6] [17].

Step 4: Data Synthesis and tDOA Inference Results from all three levels are synthesized. Conservation of primary sequence, functional domains, and critical residues in a non-target species provides evidence for structural conservation of that KE. When structural conservation is established for proteins across linked KEs, it supports the biological plausibility that the entire KER and AOP may be conserved, thereby expanding the proposed tDOA [6].

Research Reagent Solutions Toolkit

The following virtual "reagents" and resources are essential for executing this bioinformatics analysis.

Table 1: Research Reagent Solutions for SeqAPASS-driven tDOA Analysis

Item	Function/Description	Source/Example
SeqAPASS Web Tool	Core platform for conducting multi-level protein sequence comparisons and generating susceptibility predictions.	US EPA website (seqapass.epa.gov) [17]
Query Protein Sequence(s)	The reference amino acid sequence(s) from the model organism. Serves as the baseline for all cross-species comparisons.	NCBI Protein Database (e.g., Accession XP_016911190.1 for an A. mellifera nAChR subunit) [6]
NCBI Databases	Comprehensive, publicly archived repositories for protein sequences, conserved domains, and taxonomic information that form the backend data for SeqAPASS.	National Center for Biotechnology Information
AOP-Wiki	Collaborative knowledge base providing the detailed structure of the AOP (MIE, KEs, KERs) and identifying critical proteins for analysis.	aopwiki.org [6]
Critical Residue Data	Published empirical or structural data (e.g., from X-ray crystallography) identifying amino acids vital for chemical binding or protein function.	Scientific literature; referenced within AOP-Wiki KER descriptions [6]

Results & Data Analysis: Protein Conservation Across Bee Taxa

The SeqAPASS analysis of the nine AOP-relevant proteins generated quantitative data on their conservation across various bee species. The summary below illustrates the type of findings generated, which support inferences about the tDOA.

Table 2: SeqAPASS Analysis Summary for Key Proteins in the Bee Neurotoxic AOP [6]

Protein Target	Role in AOP	Level 1 Conservation (Primary Sequence)	Level 3 Conservation (Critical Residues)	Inference for tDOA
nAChR α1 Subunit	MIE: Chemical binding & receptor activation.	High (≥80% identity) across Apis and many non-Apis bees.	Key binding site residues fully conserved across all major bee families.	Strongly conserved. MIE is biologically plausible for a broad bee tDOA.
nAChR β1 Subunit	MIE: Part of receptor complex.	High across bees; moderate in more distant Hymenoptera.	Critical residues conserved in bees but not in all insects.	Conserved within bees. Supports bee-specific tDOA for MIE.
Voltage-Gated Sodium Channel	KE: Neuronal hyperexcitation.	High sequence similarity across all insects analyzed.	Functional residues critical for channel gating are universally conserved.	Widely conserved. This KE likely has a very broad tDOA (Insecta).
Acetylcholinesterase	KE: Synaptic signaling modulation.	High among bees; variable in other taxa.	Active site residues are conserved, but peripheral sites may differ.	Functionally conserved in bees. Supports KERs involving synaptic disruption.
Dopamine Receptor	KE: Altered behavior & learning.	Moderate to high among bees.	Binding pocket characteristics are maintained across Apis species.	Likely conserved in Apis. tDOA for behavior-based KERs may be narrower.

Visualization 1: SeqAPASS Tool Workflow for tDOA Analysis The following diagram illustrates the hierarchical, evidence-building workflow of the SeqAPASS tool as applied in this case study.

Visualization 2: The Neurotoxic AOP for Bees with tDOA Evidence Integration This diagram maps the essential structure of AOP 89, highlighting where SeqAPASS-derived evidence for protein conservation informs the tDOA of specific KEs and KERs.

Discussion: Enhancing KER Confidence and Expanding the tDOA

The case study demonstrates that SeqAPASS provides objective, scalable lines of evidence for the structural conservation of molecular KEs. For AOP 89, results strongly supported the conservation of the MIE (nAChR) across a broad range of bee species, thereby expanding its tDOA beyond Apis mellifera [6]. This directly strengthens the biological plausibility of the upstream KERs within the pathway for these additional species, a core objective of KER taxonomic conservation research.

However, the analysis also revealed nuances. While primary sequence (Level 1) was often highly conserved, critical residue comparisons (Level 3) provided definitive evidence for predicting functional interaction with neonicotinoids [6]. Furthermore, conservation varied among downstream proteins, suggesting that the tDOA might narrow for certain later-stage KERs (e.g., those involving specific behavioral receptors). This underscores that the tDOA is not necessarily uniform for an entire AOP but must be considered on a KE-by-KE and KER-by-KER basis.

Integrating SeqAPASS outputs with other New Approach Methodologies (NAMs), such as the G2P-SCAN tool for biological pathway analysis, can create a more robust weight-of-evidence for functional conservation [24]. This combined approach can further refine the biologically plausible tDOA, helping to fulfill the AOP framework's potential in predictive toxicology for both ecological and human health applications [24].

This analysis confirms that bioinformatics tools like SeqAPASS are indispensable for systematically defining the tDOA of AOPs. By providing evidence for the structural conservation of protein targets, the tool moves tDOA descriptions from assertions based on limited empirical data to defensible, evidence-based inferences. For the neurotoxic AOP in bees, SeqAPASS enabled the proposed expansion of the tDOA to include numerous non-Apis bees, directly informing ecological risk assessments for neonicotinoid insecticides. Ultimately, embedding such computational analyses into AOP development is crucial for building taxonomically broad, mechanistically credible pathways that can reliably support cross-species prediction in regulatory decision-making.

Comparative Evaluation of tDOA Across Different AOPs and Stressor Classes

The taxonomic domain of applicability (tDOA) is a critical, yet often narrowly defined, component of an Adverse Outcome Pathway (AOP) that determines the species for which the described biological pathway is relevant [6]. This evaluation is foundational for reliable use in regulatory decision-making, particularly when extrapolating knowledge to protect untested species [6]. This whitepaper provides a comparative technical evaluation of tDOA assessment methodologies, framed within the broader thesis that Key Event Relationships (KERs) represent the core, conserved units of AOPs [11]. We detail protocols for evaluating tDOA through bioinformatics and empirical approaches, using case studies from ecotoxicology (neonicotinoids and pollinators) and mammalian reproductive toxicology (retinoic acid signaling). The analysis underscores that a robust, comparative understanding of tDOA enhances confidence in AOP application for chemical safety assessment across diverse taxa and stressor classes [54].

The AOP framework organizes mechanistic knowledge into a causal chain from a Molecular Initiating Event (MIE) to an Adverse Outcome (AO), linked by measurable Key Events (KEs) and causal Key Event Relationships (KERs) [10]. While AOPs are often developed with specific model species, their utility in ecological and human health risk assessment depends on accurately defining their tDOA—the range of taxa for which the pathway is biologically plausible [6].

Recent conceptual advances posit that KERs are the fundamental building blocks of AOP knowledge [11]. This perspective shifts the focus of taxonomic conservation from the entire AOP to its constituent KERs. Evaluating tDOA, therefore, involves determining the conservation of the biological plausibility and empirical support for each causal link (KER) across species [10]. This requires evidence for both structural conservation (e.g., presence and similarity of proteins, receptors) and functional conservation (e.g., similar physiological role) of the entities involved in the KEs [6]. The integration of public bioinformatics tools with traditional toxicological data is essential for expanding tDOA definitions beyond the limited species for which empirical toxicity data exist [54].

Methodological Framework for tDOA Evaluation

Evaluating tDOA is a multi-evidence process combining computational and empirical lines of evidence. The following structured workflow is recommended.

Foundational Protocol: The SeqAPASS Bioinformatics Tool

The Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) tool is a publicly available web-based resource developed by the U.S. EPA to evaluate cross-species protein conservation [6]. Its hierarchical, three-level analysis provides key evidence for structural conservation.

• Level 1: Primary Sequence Similarity. The tool performs a BLAST-based comparison of the primary amino acid sequence of a query protein (e.g., a receptor identified as an MIE) against databases. It identifies putative orthologs across species and generates a similarity score, providing an initial line of evidence for the presence and broad conservation of the molecular target [6].
• Level 2: Functional Domain Conservation. This level assesses the conservation of specific protein domains known to be critical for function (e.g., ligand-binding domains, catalytic sites). Conservation across species at this level increases confidence that the protein performs a similar biochemical role [6].
• Level 3: Critical Residue Conservation. The most precise level evaluates the conservation of individual amino acid residues known to be essential for chemical-protein interaction (e.g., neonicotinoid binding site on nAChR) or protein function. A lack of conservation at this level can predict markedly different susceptibility [6].

Protocol Application: For a given AOP, identify all relevant proteins (MIE target, intermediate signaling molecules). Submit each as a query to SeqAPASS. The aggregated results across levels and proteins inform a biologically plausible tDOA for the KEs and KERs [6].

Protocol for Empirical KER Assessment

Complementing bioinformatics, the empirical assessment of a KER follows a standardized template to evaluate its strength and taxonomic anchors [11] [10].

• KER Definition: Precisely define the upstream and downstream Key Events.
• Weight of Evidence Analysis:
  • Biological Plausibility: Is the relationship consistent with established biological knowledge?
  • Empirical Support: What experimental evidence (e.g., co-occurrence, dose-response, temporal concordance) supports the causal link? Document the specific species used in these studies.
  • Essentiality: Does modulation (inhibition/activation) of the upstream KE alter the downstream KE?
• Taxonomic Anchoring: Explicitly list the species for which empirical evidence exists. Use bioinformatics (SeqAPASS) to evaluate structural conservation and propose a broader, biologically plausible tDOA [6].

Comparative Case Study Analysis

The following case studies illustrate the application of tDOA evaluation across different stressor classes (synthetic insecticides vs. endogenous signaling disruptors) and taxonomic groups (invertebrates vs. mammals).

Case Study 1: Neonicotinoid Activation of nAChR Leading to Colony Collapse (AOP 89)

This AOP, developed for the honey bee (Apis mellifera), links the MIE of nicotinic acetylcholine receptor (nAChR) activation to the AO of colony death/failure [6].

• tDOA Evaluation Methodology: Researchers used SeqAPASS to evaluate nine proteins involved in the AOP, including nAChR subunits and downstream neuronal signaling components [6].
• Key Findings: Level 1 and 2 analyses confirmed the broad conservation of nAChR subunits across Hymenoptera (bees, wasps) and other insects. Level 3 analysis of critical ligand-binding residues helped refine predictions about which bee species may be most susceptible to neonicotinoids, providing a data-driven tDOA that extends beyond the single model species [6].
• Quantitative Data Summary:

Table 1: tDOA Evaluation for AOP 89 (Neonicotinoid - nAChR - Colony Collapse)

Evaluation Aspect	Empirical tDOA (from Literature)	Bioinformatics (SeqAPASS) Inferred tDOA	Key Evidence & Confidence
MIE: nAChR Activation	Apis mellifera (Honey bee)	Likely all insects possessing conserved nAChR ligand-binding domain. Specificity within bees informed by residue analysis.	High confidence for insects; variable confidence within insects based on Level 3 residue conservation [6].
Intermediate KEs (Neuronal Hyperexcitation)	Primarily A. mellifera	Plausible for taxa with conserved neuronal physiology and target proteins.	Moderate confidence, dependent on functional conservation of downstream signaling pathways [6].
AO: Colony Death/Failure	A. mellifera (some evidence for Bombus spp.)	Limited to eusocial bees. Not applicable to solitary species.	Low extrapolation confidence; AO is highly dependent on social behavior, not just molecular conservation [6].

Case Study 2: Inhibition of Retinoic Acid Signaling Leading to Female Infertility (AOP 398 & KER 2477)

This developing AOP in mammals links inhibition of ALDH1A enzymes (MIE) to reduced fertility (AO). A core KER (2477) describes the link between decreased all-trans retinoic acid (atRA) in the fetal ovary and disrupted meiotic entry of oogonia [11].

• tDOA Evaluation Methodology: tDOA was assessed through comparative empirical studies across mammals (mice, rats, humans) and analysis of the conservation of the retinoic acid synthesis and signaling pathway (e.g., ALDH1A enzymes, STRA8 expression) [11].
• Key Findings: The core biological role of atRA in initiating meiosis is highly conserved across mammals. However, the source of atRA (gonadal vs. extra-gonadal) differs between species (e.g., mouse vs. human). This demonstrates that while the KER (the causal link) is conserved, the specific biological context of a KE (the cellular source of the signal) can vary within the tDOA [11].
• Quantitative Data Summary:

Table 2: tDOA Evaluation for KER 2477 (Reduced atRA → Disrupted Meiosis) within AOP 398

Evaluation Aspect	Empirical tDOA (from Literature)	Inferred Biological Plausible tDOA	Key Evidence & Confidence
Upstream KE: Decreased atRA in Ovary	Mouse, Rat, Rabbit, Human	All mammalian species.	High confidence based on conserved role of atRA in gonad development [11].
KER 2477: Link to Disrupted Meiosis	Strong evidence in mouse, rat, rabbit; observational in human.	Strongly plausible for therian mammals.	High biological plausibility. Essentiality shown via genetic knockout (Stra8-/-) and dietary vitamin A deficiency studies [11].
Downstream KE: Disrupted Meiotic Entry	Mouse, Rat, Human	All mammalian species.	High confidence; meiotic marker STRA8 is a direct target of atRA signaling and is conserved [11].

Diagram 1: Integrated Workflow for Evaluating tDOA (72 characters)

Diagram 2: AOP 89 tDOA Analysis with Evidence (58 characters)

Table 3: Research Reagent Solutions for tDOA and KER Conservation Studies

Tool/Resource Name	Type	Primary Function in tDOA Evaluation	Access/Source
SeqAPASS	Bioinformatics Tool	Evaluates protein sequence and structural conservation across species via three-tiered analysis to inform structural tDOA.	https://seqapass.epa.gov/ [6]
AOP-Wiki	Knowledgebase	Central repository for published AOPs, KEs, and KERs. Provides templates for development and captures tDOA information.	https://aopwiki.org/ [10]
EcoDrug	Database	Links human drug targets to orthologs in >600 eukaryotes, aiding in predicting pharmaceutical target conservation across species.	https://www.ecodrug.org/ [54]
OECD AOP Developers' Handbook	Guidance Document	Provides standardized methods and principles for AOP/KER development, including weight-of-evidence assessment for KERs and tDOA.	https://aopwiki.org/handbooks [10]
EcoToxChips	Experimental Tool	Cross-species qPCR arrays for measuring conserved transcriptional responses, providing functional evidence for KE activation.	Cited in literature [54]

Synthesis and Strategic Recommendations for tDOA Characterization

The comparative analysis reveals that a robust tDOA is not a binary designation but a gradient of confidence informed by multiple lines of evidence. The following strategic recommendations are proposed:

• Adopt a KER-Centric Approach: Evaluate and document tDOA for each KER individually, as they are the conserved core units. This modular approach allows for more precise and flexible AOP application [11].
• Mandatory Bioinformatics Integration: The use of tools like SeqAPASS to assess structural conservation should be a standard step in AOP development to explicitly define the biologically plausible tDOA beyond the empirical anchors [6].
• Transparent Evidence Reporting: Within the AOP-Wiki, tDOA descriptions must clearly distinguish between the empirical tDOA (species with direct experimental data) and the inferred tDOA (species predicted via bioinformatics or phylogenetic reasoning) [6] [10].
• Address Functional Divergence: Recognize that structural conservation does not guarantee identical function or toxicological outcome. Life-history traits (e.g., sociality in AOP 89) and system-level biology can constrain the AO's tDOA, even with conserved MIEs and early KEs [6] [54].

In conclusion, advancing the science of tDOA evaluation through the comparative, integrated methodologies outlined here is essential for realizing the promise of the AOP framework in predictive toxicology and fit-for-purpose chemical risk assessment for a wide range of species and stressor classes.

The Taxonomic Domain of Applicability (tDOA) is a foundational concept within the Adverse Outcome Pathway (AOP) framework, defining the range of species for which a described sequence of Key Event Relationships (KERs) is biologically plausible [9]. In the context of KER taxonomic conservation research, a predicted tDOA starts as a plausible hypothesis, often based on initial data from a single model organism. The critical scientific challenge is transitioning this plausible prediction to a confirmed and empirically validated tDOA, thereby expanding the utility of AOPs for cross-species chemical safety assessment and drug development without further animal testing [15].

This transition is not trivial. It requires a multi-faceted validation strategy that integrates quantitative causal analysis of KERs with computational cross-species extrapolation. Recent advancements in New Approach Methodologies (NAMs) have created a pathway for this empirical validation, combining probabilistic modeling of KER confidence with bioinformatic tools that assess the conservation of molecular targets and biological pathways across the tree of life [15] [9]. The validation of a tDOA thus becomes a confirmatory process, substantiating that the mechanistic toxicity described by an AOP is not an artifact of a single species but a conserved biological response with defined taxonomic boundaries.

Foundational Frameworks for tDOA Prediction and Expansion

The initial prediction and subsequent expansion of a tDOA rely on integrating data from multiple sources and lines of evidence. The process begins with a well-constructed AOP, typically developed from a model organism, and systematically seeks to extrapolate its KERs across broader taxonomic groups.

Table 1: Core Components for tDOA Prediction and Expansion

Component	Description	Role in tDOA Validation
Adverse Outcome Pathway (AOP) Network	A structured framework linking a Molecular Initiating Event (MIE) to an Adverse Outcome (AO) via intermediate Key Events (KEs) [15].	Provides the mechanistic KER sequence whose taxonomic conservation is being evaluated. Serves as the foundational hypothesis for the predicted tDOA.
Key Event Relationship (KER) Assessment	Quantitative evaluation of the causal, correlative, or predictive links between KEs, often using Bayesian networks [15].	Establishes confidence in the AOP's internal logic. A robust KER network within the source species strengthens the plausibility of its conservation.
Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS)	A bioinformatics tool that compares primary protein sequence, domain, and 3D structural similarity to extrapolate potential chemical susceptibility [15] [9].	Empirically tests the conservation of the MIE's molecular target (e.g., a specific receptor or enzyme) across diverse species, providing direct evidence for tDOA expansion.
Genes-to-Pathways Species Conservation Analysis (G2P-SCAN)	A computational tool that maps human genes to biological pathways and evaluates the conservation of those pathways across a defined set of species [15] [9].	Provides evidence for the conservation of the broader biological pathway downstream of the MIE, supporting the plausibility that the entire KER sequence could be conserved.

A pivotal case study demonstrating this framework involved extending the tDOA for AOP 207 (reproductive toxicity of silver nanoparticles via oxidative stress in C. elegans). Researchers integrated in vivo ecotoxicology data, in vitro human toxicology data, and in silico tools (SeqAPASS and G2P-SCAN) to build a cross-species AOP network. This approach extended the biologically plausible tDOA from a few model species to over 100 taxonomic groups, including fungi, birds, rodents, and reptiles [15].

Diagram Title: Integrated workflow for expanding tDOA from model organism data.

Quantitative and Empirical Validation Methodologies

Moving from a qualitative, plausible tDOA to a quantitative, confirmed one requires rigorous empirical validation methods. These methodologies assess both the strength of the underlying KERs and the performance of the tDOA prediction against independent data.

Bayesian Network Modeling for KER Confidence

A core step in validating the AOP itself is the quantitative assessment of KERs. Bayesian Network (BN) modeling is a probabilistic approach adept at managing the inherent uncertainty and variability in biological systems [15]. It is used to analyze the causal relationships between KEs based on experimental data.

• Process: Experimental data for each KE (e.g., protein activation, cellular response, organ-level effect) is structured into the network. Conditional probability tables are calculated to define the likelihood of a downstream KE given the state of an upstream KE.
• Output: The BN provides a quantitative measure of confidence for each KER and the overall AOP network. This statistical confidence in the causal chain within the source organism is a prerequisite for having confidence in its cross-species conservation [15].

Empirical Validation Metrics for Cross-Species Predictions

Once a tDOA is predicted computationally, its validity must be tested. Empirical validation strategies from mechanistic model evaluation can be adapted for this purpose [55].

• Bootstrapped Statistical Tests: Methods like the bootstrapped log-rank test or MaxCombo test can compare the predicted incidence of an adverse outcome (derived from the AOP) against observed toxicity data in a new species within the predicted tDOA [55].
• Prediction Interval Coverage: This metric assesses whether a defined percentage (e.g., 95%) of observed experimental outcomes from validation species fall within the model's prediction intervals, which account for biological variability and uncertainty [55].
• The Juncture Metric: A specialized metric for time-to-event data (e.g., time to reproductive failure) that accounts for uncertainty in the exact observation time of an event, providing a more realistic validation against real-world experimental data [55].

Table 2: Empirical Validation Metrics for Predicted tDOA Performance

Validation Metric	Application to tDOA Validation	Interpretation of a Successful Result
Bootstrapped Log-Rank/MaxCombo Test	Statistically compares the predicted vs. observed survival (or adverse outcome) curves in a validation species.	No significant difference (p > 0.05) suggests the AOP-derived prediction aligns with empirical data, supporting tDOA inclusion.
Prediction Interval Coverage	Checks if observed endpoint measurements (e.g., brood size, enzyme activity) fall within the predicted range.	High coverage (e.g., ≥95%) indicates the model accurately captures the variability of the response in the new species.
Juncture Metric	Evaluates the accuracy of predicting the timing of a key event (e.g., onset of pathology) in validation studies.	A low juncture error score indicates the model's temporal predictions are reliable for the new species.

Detailed Experimental Protocols for tDOA Validation

The following protocols outline a step-by-step pathway for the empirical validation of a predicted tDOA, integrating the frameworks and methods described above.

Protocol: IntegratedIn SilicotDOA Expansion

This protocol uses computational NAMs to generate a testable tDOA prediction [15] [9].

• Define the Molecular Initiating Event (MIE): Precisely identify the primary molecular target (e.g., specific protein receptor) from the source AOP.
• Run SeqAPASS Analysis:
  • Input the amino acid sequence or accession number of the MIE protein from the source species.
  • Set similarity thresholds (e.g., ≥80% primary sequence identity, conserved domain structure).
  • Execute the tool to generate a list of species possessing a ortholog meeting the criteria, indicating potential susceptibility.
• Run G2P-SCAN Analysis:
  • Input the human orthologs of genes involved in the AOP's key events.
  • Map genes to Reactome or KEGG biological pathways.
  • Analyze the tool's output to identify which species conserve the complete or partial biological pathway underlying the AOP.
• Synthesize and Propose tDOA: Integrate results from SeqAPASS (MIE conservation) and G2P-SCAN (pathway conservation). The proposed tDOA includes species with strong evidence from both analyses, forming a prioritized list for empirical validation.

Protocol: Empirical Validation Using a Bayesian Network AOP Model

This protocol validates the AOP's predictive power within the proposed tDOA [15].

• Construct the Quantitative AOP (qAOP) Network:
  • From the source AOP, define all KEs as variables (nodes).
  • Using experimental data from the source organism, quantify the dose-response or response-response relationships for each KER using regression models.
• Develop the Bayesian Network (BN):
  • Structure the BN with nodes matching the qAOP KEs.
  • Parameterize the BN by populating conditional probability tables using the quantified KER data and, if available, expert elicitation for gaps.
  • Calibrate the BN model to ensure it accurately reproduces the source species' toxicity data.
• Generate Predictions and Design Validation Study:
  • Use the calibrated BN to simulate expected outcomes (e.g., probability of adverse outcome given a specific MIE perturbation) for a candidate validation species within the proposed tDOA.
  • Design an in vivo or in vitro experiment with the validation species to measure the relevant KEs and AO.
• Execute Validation and Compare:
  • Conduct the toxicity study with the validation species.
  • Compare the observed data against the BN model predictions using the validation metrics from Table 2 (e.g., prediction interval coverage).
  • A successful validation provides strong empirical evidence to confirm the species' inclusion in the tDOA.

Diagram Title: Two-phase protocol for tDOA prediction and empirical validation.

Table 3: Research Reagent Solutions for tDOA Validation

Tool/Reagent Category	Specific Item	Function in tDOA Validation
Bioinformatics Software	SeqAPASS Tool (v6.1+) [9]	Provides empirical evidence for the conservation of the molecular initiating event (MIE) protein across species.
Bioinformatics Software	G2P-SCAN Tool (v0.0.1.0+) [15] [9]	Evaluates the conservation of the broader biological pathway implicated in the AOP across key model species.
Statistical Modeling Software	Bayesian Network Software (e.g., Netica, GeNIe, R packages bnlearn, gRbase)	Enables the construction, parameterization, and simulation of quantitative AOP networks for probabilistic prediction.
Reference Chemical	AgNO₃ or characterized AgNPs [15]	A positive control stimulus for validating AOPs involving oxidative stress and reproductive toxicity (e.g., AOP 207).
Reference Chemical	Prototypical Receptor Agonists/Antagonists (e.g., for PPARα, ESR1) [9]	Used in validation studies to directly perturb a specific MIE and test the downstream KER sequence in a new species.
Validated Assay Kits	ROS detection kits (e.g., DCFDA), Caspase-3 activity kits, Hormone ELISA kits	Provide standardized methods to quantitatively measure key events (KEs) such as oxidative stress, apoptosis, or endocrine disruption in validation studies.
Reference Genomic Material	cDNA or gDNA from species across the proposed tDOA	Essential for in vitro cloning and expression of putative orthologs to functionally test MIE-chemical interaction (e.g., in reporter gene assays).

The empirical validation of a predicted tDOA transforms an AOP from a species-specific model into a generalized tool for predictive toxicology. For drug development professionals, this has direct applications:

• De-risking Candidate Drugs: A confirmed tDOA identifies which non-target species (in environmental risk assessment) or patient subpopulations (via conserved human pathways) might be susceptible to a drug's off-target toxicity, informing early safety profiling [9].
• Guiding Animal Model Selection: It provides a mechanistic rationale for selecting the most relevant in vivo models for toxicity testing, ensuring the biological pathways of concern are conserved.
• Supporting New Approach Methodologies (NAMs): A robust, validated tDOA justifies the use of non-animal testing methods (e.g., human in vitro assays) for safety decisions on chemicals when the relevant AOP is confirmed to be conserved in humans [15].

The pathway from plausible to confirmed tDOA is emblematic of the evolving paradigm in toxicological sciences. It leverages computational power to generate hypotheses and empirical rigor to test them, ultimately strengthening the scientific confidence in cross-species extrapolation and enabling more efficient, ethical, and predictive safety assessments.

Conclusion

Defining the taxonomic domain of applicability for Key Event Relationships is not a peripheral task but a fundamental requirement for the credible application of AOPs in modern, animal-sparing toxicology. As demonstrated, the integration of computational bioinformatics tools like SeqAPASS and G2P-SCAN provides a powerful, evidence-based methodology to extrapolate mechanistic knowledge beyond the model organisms used in initial AOP development [citation:1][citation:2][citation:5]. Success hinges on a systematic, transparent approach that combines structural sequence analysis with functional pathway conservation, all framed within a rigorous weight-of-evidence assessment. The future of this field lies in the development of standardized, accepted workflows for tDOA definition and their integration into the AOP-Wiki framework, fostering consistency and collaboration. Initiatives like the International Consortium to Advance Cross-Species Extrapolation in Regulation (ICACSER) are pivotal in driving this harmonization forward [citation:10]. Ultimately, robust KER taxonomic conservation strengthens the predictive power of the AOP framework, accelerating its use in regulatory decision-making to achieve comprehensive chemical safety assessments for both human and environmental health under the unifying vision of One Health [citation:1][citation:10].

References

• 1. Adverse outcome pathway [https://en.wikipedia.org/wiki/Adverse_outcome_pathway]
• 2. Help: KE Relationships [https://aopwiki.org/info_pages/2/info_linked_pages/2]
• 3. Key Event Relationships → Area [https://lifestyle.sustainability-directory.com/area/key-event-relationships/]
• 4. Top 6 Visualizations for Quantitative Data Analysis Methods [https://chartexpo.com/blog/quantitative-data-analysis-methods]
• 5. Geographic, Taxonomic and Metric Gaps in Biodiversity ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12513691/]
• 6. Defining the Biologically Plausible Taxonomic Domain of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10100214/]
• 7. AOP Key Event Relationship report: Linking decreased ... [https://www.sciencedirect.com/science/article/pii/S2666027X22000068]
• 8. Advancing predictive toxicology: overcoming hurdles and ... [https://pubs.rsc.org/en/content/articlehtml/2025/dd/d4dd00257a]
• 9. Combination of Computational New Approach Methodologies ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10926110/]
• 10. Developers' Handbook 2.5 [https://aopwiki.org/handbooks/3]
• 11. AOP Key Event Relationship report: Linking decreased ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8957012/]
• 12. Principles underlying the next generation risk assessment approach (NGRA) – ToxMinds [https://toxminds.com/principles-underlying-the-next-generation-risk-assessment-approach-ngra/]
• 13. Next Generation Risk Assessment: Incorporation of Recent Advances in Molecular, Computational, and Systems Biology (External Review Draft) | Risk Assessment Portal | US EPA [https://cfpub.epa.gov/ncea/risk/recordisplay.cfm?deid=259936]
• 14. Next Generation Risk Assessment approaches for advanced nanomaterials: Current status and future perspectives - ScienceDirect [https://www.sciencedirect.com/science/article/pii/S2452074824000338]
• 15. Environmental Toxicology and Chemistry | Oxford Academic [https://academic.oup.com/etc/article/44/9/2511/7943011]
• 16. Sequence Alignment to Predict Across Species ... [https://www.epa.gov/comptox-tools/sequence-alignment-predict-across-species-susceptibility-seqapass-resource-hub]
• 17. Demonstration of the Sequence Alignment to Predict ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10758989/]
• 18. SeqAPASS User Guide [https://www.epa.gov/comptox-tools/seqapass-user-guide]
• 19. From Protein Sequence to Structure: The Next Frontier in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11265300/]
• 20. USEPA/seqapass-common [https://github.com/USEPA/seqapass-common]
• 21. Continued development of SeqAPASS [https://ntp.niehs.nih.gov/iccvamreport/2023/technology/comp-tools-dev/10-epa-seqapass]
• 22. Using molecular dynamics simulation to enhance ... [https://academic.oup.com/etc/article/44/9/2687/8171337]
• 23. Genes-to-Pathways Species Conservation Analysis [https://pubmed.ncbi.nlm.nih.gov/36861224/]
• 24. Combination of computational new approach ... [https://www.sciencedirect.com/science/article/abs/pii/S0048969723072017]
• 25. Combination of computational new approach ... [https://pubmed.ncbi.nlm.nih.gov/37981146/]
• 26. seacunilever/G2P-SCAN [https://github.com/seacunilever/G2P-SCAN]
• 27. Cross-species molecular docking method to support ... [https://www.sciencedirect.com/science/article/abs/pii/S2468111324000215]
• 28. Publication trends in global biodiversity research on ... [https://www.sciencedirect.com/science/article/pii/S0006320723000885]
• 29. Bridging the gap between human toxicology and ... [https://pubmed.ncbi.nlm.nih.gov/38980262/]
• 30. Putative adverse outcome pathways for silver nanoparticle ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9814206/]
• 31. Nano Silver-Induced Toxicity and Associated Mechanisms [https://pmc.ncbi.nlm.nih.gov/articles/PMC9056105/]
• 32. Synthesis, applications, toxicity and toxicity mechanisms of ... [https://www.sciencedirect.com/science/article/pii/S0147651323001409]
• 33. Proof of concept for quantitative adverse outcome pathway ... [https://www.nature.com/articles/s41598-024-55220-4]
• 34. AOP key event relationship report: Linking androgen ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9459418/]
• 35. Inhibition, 5-hydroxytryptamine transporter (5-HTT [https://aopwiki.org/aops/568]
• 36. Computational strategies for cross-species knowledge ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11343225/]
• 37. Absence Data [https://support.bccvl.org.au/support/solutions/articles/6000235254-absence-data]
• 38. Understanding Absence Data in Species Distribution Models [https://support.ecocommons.org.au/support/solutions/articles/6000255789-absence-data]
• 39. Boundary Effects Cause False Signals of Range Expansions ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11135943/]
• 40. Integrated in vitro, in silico, and in vivo approaches to ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11850774/]
• 41. Environmental Toxicology and Chemistry | Oxford Academic [https://academic.oup.com/etc/article/40/12/3226/7717070]
• 42. Assessing causality in epidemiology: revisiting Bradford ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8206235/]
• 43. Applying the Bradford Hill criteria in the 21st century [https://pmc.ncbi.nlm.nih.gov/articles/PMC4589117/]
• 44. Bradford Hill criteria [https://en.wikipedia.org/wiki/Bradford_Hill_criteria]
• 45. Assessing causality in epidemiology: revisiting Bradford ... [https://link.springer.com/article/10.1007/s10654-020-00703-7]
• 46. SYSTEMATIC REVIEW AND WEIGHT OF EVIDENCE ARE ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7551547/]
• 47. Guidance on the use of the weight of evidence ... - EFSA [https://www.efsa.europa.eu/en/efsajournal/pub/4971]
• 48. Qualitative vs. Quantitative Data & Why Data Matters to ... [https://www.swydo.com/blog/qualitative-vs-quantitative-data/]
• 49. Qualitative vs Quantitative Research: Complete Guide 2025 [https://www.risurveys.com/blog/simple-guide-to-choosing-quant-vs-qual]
• 50. Qualitative vs. Quantitative Data in Research: The Difference [https://www.fullstory.com/blog/qualitative-vs-quantitative-data/]
• 51. Qualitative vs Quantitative Research: What's the Difference? [https://www.simplypsychology.org/qualitative-quantitative.html]
• 52. (PDF) Assessing the taxonomic status of dingoes Canis familiaris... [https://web.archive.org/web/20210809115736/https://www.researchgate.net/publication/43453285_Assessing_the_taxonomic_status_of_dingoes_Canis_familiaris_dingo_for_conservation]
• 53. action based on threatened species capture Conservation ... taxonomic [https://pubmed.ncbi.nlm.nih.gov/25337861/]
• 54. Towards Precision Ecotoxicology: Leveraging Evolutionary ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11017229/]
• 55. Empirical methods for the validation of time-to-event ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10478282/]