Controlled Vocabularies in Ecotoxicology: Building a Foundation for Reliable Data and Research

Abstract

This article provides a comprehensive guide to controlled vocabularies (CVs) in ecotoxicology, tailored for researchers and drug development professionals. It explores the foundational role of CVs in organizing complex toxicity data, as exemplified by major resources like the ECOTOX Knowledgebase. The content details methodological approaches for implementing CVs, addresses common challenges in data curation and integration, and presents frameworks for validating data reliability. By establishing a clear understanding of how standardized terminology enhances data findability, interoperability, and reuse, this article aims to support more robust environmental risk assessments and chemical safety evaluations.

What Are Controlled Vocabularies and Why Are They Critical in Ecotoxicology?

In the realm of ecotoxicity data research, the standardization of terminology is not merely a convenience but a fundamental requirement for data integrity, interoperability, and reuse. A controlled vocabulary is an authoritative set of terms selected and defined based on the requirements set out by the user group, used to ensure consistent indexing or description of data or information [1]. These vocabularies do not necessarily possess inherent structure or relationships between terms but serve as the foundational layer for creating standardized knowledge systems.

The critical importance of controlled vocabularies becomes apparent when dealing with complex data extraction processes, such as in systematic reviews of toxicological end points from primary sources. Primary source language describing treatment-related end points can vary greatly, resulting in large labor efforts to manually standardize extractions before data are fit for use [1]. In ecotoxicity research, where data informs critical public health and regulatory decisions, this consistency is paramount. Without standardized annotation, divergent language describing study parameters and end points inhibits crosstalk among individual studies and resources, preventing meaningful synthesis of data across studies and ultimately compromising the FAIR (Findable, Accessible, Interoperable, and Reusable) principles that govern modern scientific data management [1].

Core Concepts and Terminology

Hierarchical Organization of Knowledge Systems

Knowledge organization systems exist on a spectrum of complexity and structure, each serving distinct purposes in information management:

• Term Lists: The simplest form, consisting of authorized terms with limited relationships
• Taxonomies: Hierarchical classification systems that organize concepts into categories and subcategories
• Thesauri: Controlled vocabularies that explicitly specify semantic relationships between concepts, including equivalence, hierarchical, and associative relationships
• Ontologies: Complex knowledge representations that define concepts and their relationships with formal logic, enabling computational reasoning and inference [1]

Foundational Terminology Table

Table 1: Core Terminology in Controlled Vocabulary Development

Term	Definition	Application in Ecotoxicity
Controlled Vocabulary	An authoritative set of standardized terms used to ensure consistent data description [1]	Standardizing terms for toxicological end points such as "hepatocellular hypertrophy"
SKOS (Simple Knowledge Organization System)	A W3C standard to support the use of knowledge organization systems within the Semantic Web framework [2]	Representing ecotoxicity thesauri in linked data formats
Indexing Language	The set of terms used in an index to represent topics or features of documents [3]	Cataloging developmental toxicity study outcomes
Crosswalk	A mapping that shows how terms in different vocabularies correspond to each other [1]	Aligning UMLS, OECD, and BfR DevTox terms for data harmonization
Precoordination	Combining multiple concepts into a single term (e.g., "head_small") [1]	Describing complex morphological abnormalities in developmental studies
Compositionality	The degree to which terms are formed by combining reusable semantic components [3]	Building complex toxicity findings from basic anatomical and effect terms

SKOS Standards Framework

The Simple Knowledge Organization System (SKOS) is a W3C-developed area of work producing specifications and standards to support the use of knowledge organization systems such as thesauri, classification schemes, subject heading lists, and taxonomies within the framework of the Semantic Web [2]. SKOS provides a standardized, machine-readable framework for representing controlled vocabularies, enabling them to be shared and linked across the web.

SKOS became a W3C Recommendation in August 2009, representing a significant milestone in bridging the world of knowledge organization systems with the linked data community [2]. This standard brings substantial benefits to libraries, museums, government portals, enterprises, and research communities that manage large collections of scientific data, including ecotoxicity resources. The alignment between SKOS and the ISO 25964 thesaurus standard further enhances its utility as an international framework for vocabulary representation [2].

The core SKOS data model organizes knowledge through several fundamental properties and relationships. Concepts are labeled using preferred, alternative, and hidden terms, while semantic relationships are established through broader, narrower, and related associations. Additionally, SKOS supports documentation through scope notes, definitions, and examples, as well as grouping concepts into concept schemes and collections for enhanced organization.

Quantitative Comparison Metrics for Indexing Languages

Intra-Term Set Measurement Protocols

The quantitative characterization of indexing languages enables empirical, reproducible comparison between different vocabulary systems. These metrics are divided into two primary categories: intra-set measurements that describe the internal structure of a single term set, and inter-set measurements that compare overlaps between different term sets [3].

Table 2: Intra-Term Set Metrics for Vocabulary Analysis

Metric	Measurement Protocol	Interpretation in Ecotoxicity Context
Number of Distinct Terms	Count of syntactically unique terms in the set [3]	Indicates coverage and granularity of toxicity concepts
Term Length Distribution	Descriptive statistics (mean, median) of character counts per term [3]	Reflects specificity and precoordination level of end point descriptions
Observed Linguistic Precoordination	Categorization of terms as uniterms, duplets, triplets, or quadruplets+ based on syntactic separators [3]	Measures compositional structure in morphological abnormality terms
Flexibility Score	Fraction of sub-terms that also appear as uniterms [3]	Indicates reusability of semantic components in developmental toxicology
Compositionality	Number of terms containing another complete term as a proper substring [3]	Reveals semantic factoring in complex pathological findings

Inter-Term Set Comparison Methodology

The protocol for comparing different controlled vocabularies involves calculating overlap metrics that reveal the degree of alignment between systems:

• Term Set Preparation: Extract complete term lists from each vocabulary system to be compared
• Normalization: Apply consistent case-folding, punctuation removal, and stemming to enable fair comparison
• Exact Match Calculation: Compute the Jaccard similarity coefficient as the size of the intersection divided by the size of the union of the two term sets
• Semantic Similarity Assessment: Employ advanced natural language processing techniques to identify related terms beyond exact string matches
• Structural Alignment Analysis: Map hierarchical relationships and property structures between the vocabularies

Application Protocol: Implementing Controlled Vocabularies for Ecotoxicity Data

Experimental Protocol: Automated Vocabulary Mapping for Toxicological End Points

The following protocol details a proven methodology for standardizing extracted ecotoxicity data using automated application of controlled vocabularies, adapted from successful implementation in developmental toxicology studies [1].

Objective: To minimize labor efforts in standardizing extracted toxicological end points through an augmented intelligence approach that automatically applies preexisting controlled vocabularies.

Materials and Reagents:

Table 3: Research Reagent Solutions for Vocabulary Mapping

Item	Specification	Function
Source Data	Extracted end points from prenatal developmental toxicology studies (approx. 34,000 extractions) [1]	Provides raw terminology for standardization
Vocabulary Crosswalk	Harmonized mapping between UMLS, OECD, and BfR DevTox terms [1]	Serves as reference for standardized terminology
Annotation Code	Python 3 (version 3.7) scripts for automated term matching [1]	Executes the computational mapping process
Validation Dataset	Manually curated subset of extracted end points (≥500 terms)	Provides ground truth for performance evaluation

Procedure:

• Crosswalk Development Phase:

  • Create a harmonized controlled vocabulary crosswalk containing Unified Medical Language System (UMLS) codes, German Federal Institute for Risk Assessment (BfR) DevTox harmonized terms, and The Organization for Economic Co-operation and Development (OECD) end point vocabularies [1]
  • Establish semantic relationships between equivalent terms across the three vocabulary systems
  • Document hierarchy and mapping rules for complex term relationships

• Automated Mapping Phase:

  • Apply annotation code to match extracted end point language to controlled vocabulary terms
  • Implement fuzzy matching algorithms to handle spelling variations and synonyms
  • Generate confidence scores for each automated mapping decision

• Validation and Quality Control Phase:

  • Manually review a statistically significant sample of automated mappings (approximately 51% of total) for potential extraneous matches or inaccuracies [1]
  • Identify patterns in terms that resist automated mapping (typically overly general terms or those requiring human logic)
  • Calculate performance metrics including precision, recall, and mapping coverage

• Implementation Phase:

  • Apply standardized controlled vocabulary terms to the successfully mapped extractions
  • Document unmapped terms for future vocabulary expansion
  • Generate FAIR-compliant dataset for downstream analysis

Expected Outcomes:

• Automatic application of standardized controlled vocabulary terms to 75% of extracted end points from guideline studies [1]
• Significant reduction in manual standardization effort (estimated savings of >350 hours) [1]
• Production of computationally accessible, standardized developmental toxicity datasets

Visualizing Vocabulary Mapping Workflow

Diagram Title: Automated Vocabulary Mapping Process

Advanced Applications in Ecotoxicity Research

Protocol for Cross-Study Data Integration

The integration of ecotoxicity data across multiple studies and research domains requires sophisticated vocabulary alignment techniques. The following protocol enables semantic interoperability between disparate data sources:

• Source Vocabulary Analysis:

  • Apply intra-term set metrics to characterize each source vocabulary's size, term length distribution, and compositionality [3]
  • Identify structural patterns and semantic factoring within each terminology system

• Intersection Mapping:

  • Calculate pairwise Jaccard similarity coefficients between all vocabulary pairs
  • Identify core concept overlap and domain-specific extensions
  • Map hierarchical relationships across vocabulary boundaries

• SKOS Representation:

  • Convert aligned vocabulary to SKOS format using standardized predicates (skos:prefLabel, skos:broader, skos:narrower) [2]
  • Publish linked data vocabulary for Semantic Web applications

• Query Federation:

  • Implement SPARQL endpoints for each standardized vocabulary
  • Enable cross-database queries using aligned semantic concepts

Augmented Intelligence Implementation

The successful application of automated vocabulary mapping in developmental toxicology demonstrates the power of augmented intelligence approaches [1]. This methodology combines computational efficiency with human expertise through:

• Automated Processing: Handling routine, high-confidence mappings algorithmically
• Human Oversight: Applying expert judgment to complex cases and ambiguity resolution
• Continuous Improvement: Using manual corrections to refine automated algorithms
• Resource Generation: Producing reusable assets including controlled vocabulary crosswalks, organized related terms lists, and customizable code for implementation in other study types [1]

This approach has proven particularly valuable for standardizing legacy developmental toxicology datasets, where historical terminology variations present significant challenges for contemporary computational toxicology and predictive modeling applications.

The systematic implementation of controlled vocabularies, standardized through frameworks like SKOS and applied via rigorous protocols such as those described herein, represents a transformative methodology for ecotoxicity data research. By moving from ad hoc terminology to structured, computable knowledge organization systems, researchers can unlock the full potential of existing and future toxicity data. The quantitative metrics, automated mapping protocols, and visualization approaches detailed in these application notes provide researchers, scientists, and drug development professionals with practical tools for enhancing data interoperability, supporting validation of alternative methods, and ultimately strengthening the scientific foundation for chemical risk assessment and regulatory decision-making.

Application Note: The Scale and Diversity of Ecotoxicity Data

The evaluation of chemical safety relies on the systematic compilation and curation of ecotoxicity data. The volume and variety of this data present significant challenges, underscoring the critical need for standardized curation processes and controlled vocabularies to ensure reusability and interoperability.

Table 1: Scope and Scale of Publicly Available Ecotoxicity and Toxicology Databases

Database Name	Primary Focus	Number of Chemicals	Number of Records/Results	Key Data Types
ECOTOX [4]	Ecological toxicity	>12,000	>1,100,000 test results	Single-chemical ecotoxicity tests for aquatic and terrestrial species.
ToxValDB (v9.6.1) [5]	Human health toxicity	41,769	242,149 records	Experimental & derived toxicity values, exposure guidelines.
ToxRefDB [6]	In vivo animal toxicity	>1,000	Data from >6,000 studies	Detailed in vivo study data from guideline-like studies.
ADORE [7]	Acute aquatic toxicity (ML-ready)	Not Specified	Extracted from ECOTOX	Curated acute mortality data for fish, crustaceans, and algae, expanded with chemical & species features.

The Data Variety Challenge

The variety in ecotoxicity data manifests across multiple dimensions, necessitating robust controlled vocabularies for meaningful integration:

• Taxonomic Diversity: Databases like ECOTOX encompass over 14,000 species, requiring consistent taxonomic classification [4].
• Experimental Endpoints: A single effect, such as mortality, can be represented by different measured endpoints (e.g., LC50, EC50) which must be clearly defined and categorized [7].
• Experimental Variables: Critical factors such as exposure duration, test medium, and organism life stage introduce significant variability that must be captured with standardized terminology [4] [7].

Protocol: Systematic Data Curation and Integration Workflow

This protocol details a standardized procedure for curating ecotoxicity data from primary sources, emphasizing the use of controlled vocabularies to support computational toxicology and research.

Experimental Workflow for Data Curation

The following diagram illustrates the multi-stage pipeline for processing ecotoxicity data, from initial acquisition to final standardized output.

Reagent and Resource Solutions

Table 2: Essential Research Reagents and Computational Tools for Ecotoxicology

Item/Tool Name	Function/Application	Key Features
ECOTOX Knowledgebase [4]	Authoritative source for curated single-chemical ecotoxicity data.	Over 1 million test results; systematic review procedures; FAIR data principles.
CompTox Chemicals Dashboard [6]	Chemistry resource supporting computational toxicology.	Provides DSSTox Substance IDs (DTXSID), chemical structures, and property data.
DataFishing Tool [8]	Python script/web form for automated data retrieval from multiple biological databases.	Efficiently obtains taxonomic, DNA sequence, and conservation status data.
ToxValDB [5]	Compiled resource of human health-relevant toxicity data.	Standardized format for experimental and derived toxicity values from multiple sources.
ADORE Dataset [7]	Benchmark dataset for machine learning in aquatic ecotoxicology.	Curated acute toxicity data with chemical, phylogenetic, and species-specific features.

Detailed Procedural Steps

Step 1: Data Acquisition and Harmonization

• Input Sources: Download core data files (e.g., species, tests, results) from authoritative sources like the ECOTOX downloadable ASCII files [7].
• Identifier Mapping: Retain and map critical chemical identifiers, including CAS numbers, DSSTox Substance IDs (DTXSID), and InChIKeys, to ensure traceability and integration with other chemical resources [7].

Step 2: Application of Inclusion/Exclusion Criteria

• Taxonomic Filtering: Retain data only for relevant taxonomic groups (e.g., Filter ecotox_group for "Fish", "Crusta", "Algae") [7].
• Endpoint Selection: Define and select relevant toxicity endpoints based on standardized test guidelines. For example:
  • Fish: Mortality (MOR), typically over 96 hours [7].
  • Crustaceans: Mortality (MOR) or Immobilization/Intoxication (ITX), typically over 48 hours [7].
  • Algae: Effects on growth (GRO), population (POP), or physiology (PHY), typically over 72 hours [7].
• Study Type Exclusion: Remove data from in vitro tests and assays based on early life stages (e.g., eggs, embryos) if the objective is to model adult organism toxicity [7].

Step 3: Vocabulary Control and Metadata Curation

• Taxonomy Curation: Ensure species entries have complete taxonomic data (Class, Order, Family, Genus, Species). Remove entries with missing critical classification data [7].
• Effect and Endpoint Mapping: Categorize all effects and endpoints using a controlled vocabulary. For example, map various reported effects to standardized categories like "MOR," "GRO," etc. [7].
• Experimental Condition Annotation: Extract and codify key experimental conditions such as exposure duration, test medium, and organism life stage using predefined terms [4].

Step 4: Data Standardization and Value Conversion

• Unit Standardization: Convert all concentration values to a standard unit (e.g., mg/L or mol/L). Molar concentrations are often preferred for QSAR and machine learning as they are more biologically informative [7].
• Value Deduplication: Implement a quality control (QC) workflow to identify and resolve duplicate records from multiple sources, a process used in ToxValDB development [5].

Step 5: Feature Expansion and Dataset Integration

• Chemical Descriptor Addition: Expand the dataset with chemical features such as canonical SMILES, molecular representations, and physicochemical properties from resources like PubChem and the CompTox Chemicals Dashboard [7].
• Species Feature Integration: Add phylogenetic and species-specific traits to enable analyses that consider evolutionary relationships [7].
• Data Structuring: Output the final curated data in a structured, machine-readable format (e.g., CSV, SQL database) for subsequent analysis and modeling.

Protocol: Implementing a Controlled Vocabulary for Data Curation

The implementation of a consistent controlled vocabulary is fundamental to overcoming the variety challenge in ecotoxicity data.

Logical Framework for Vocabulary Implementation

The relationship between core data entities and the controlled vocabularies that structure them is illustrated below.

Procedures for Vocabulary Management

• Vocabulary Development: Define and maintain lists of approved terms for critical data fields. This includes standardized terms for taxonomic classification, observed effects, measured endpoints, and experimental conditions [4] [7].
• Data Curation Pipeline: Incorporate vocabulary mapping as a distinct step in the data processing workflow, ensuring all incoming data is translated into the controlled terms before integration into the master database [5].
• Quality Control Checks: Implement automated checks to flag values that do not conform to the controlled vocabulary, allowing for curator review and corrective action, thereby maintaining data integrity [5].

Adherence to these detailed protocols enables the transformation of disparate, complex ecotoxicity data into a structured, standardized resource. This structured data is essential for advancing computational toxicology, developing predictive models, and supporting robust chemical safety assessments.

The ECOTOXicology Knowledgebase (ECOTOX) stands as the world's largest compilation of curated ecotoxicity data, housing over one million test results for more than 12,000 chemicals and 13,000 species from over 53,000 scientific references [9] [10]. This monumental achievement in data management is underpinned by a rigorous, systematic application of controlled vocabularies (CVs). This case study details how ECOTOX employs CVs to ensure data consistency, enhance interoperability, and support robust environmental research and chemical risk assessments, contributing to a broader framework for reliable ecotoxicity data research.

In the field of ecotoxicology, the diversity of terminology used across thousands of scientific studies presents a significant challenge for data integration and reuse. Controlled vocabularies are predefined, standardized sets of terms used to consistently tag and categorize data. Within ECOTOX, these CVs provide the necessary semantic structure to transform free-text information from disparate literature sources into a harmonized, query-ready knowledgebase [10]. This practice is fundamental to making the data Findable, Accessible, Interoperable, and Reusable (FAIR).

The ECOTOX Data Curation Pipeline: A Systematic Workflow

The process of incorporating data into ECOTOX is a meticulously designed pipeline that ensures only relevant, high-quality studies are added, with all information translated into a consistent language of controlled terms. The workflow, summarized in the diagram below, involves multiple stages of screening and extraction [10].

Protocol: Literature Review and Data Curation Pipeline

The ECOTOX team follows standardized protocols aligned with systematic review practices to identify and curate ecotoxicity data [10] [11].

• Step 1: Literature Search and Acquisition

  • Objective: Comprehensively identify potentially relevant ecotoxicity studies from the open scientific literature.
  • Methods: Develop and execute customized search strings for chemicals of interest across multiple bibliographic databases. Grey literature, including government technical reports, is also included [10].

• Step 2: Citation Screening for Applicability

  • Objective: Filter references to retain only those reporting on ecologically relevant species and single-chemical exposures.
  • Methods: Review titles and abstracts against predefined eligibility criteria (e.g., presence of an ecologically relevant species, a single chemical stressor, and a measurable effect) [10]. This step significantly reduces the volume of references for full-text review.

• Step 3: Full-Text Review for Acceptability

  • Objective: Assess the methodological quality and reporting adequacy of the study.
  • Methods: Review the full text of articles against criteria for scientific rigor. Studies must include, for example, documented control groups and reported concentration-response relationships to be accepted for data extraction [10].

• Step 4: Data Abstraction and Controlled Vocabulary Application

  • Objective: Systematically extract and standardize key information from accepted studies.
  • Methods: Trained curators extract pertinent details into structured data fields. This is the critical stage where controlled vocabularies are applied to describe:
    • Chemical: Using identifiers like CAS numbers and DSSTox Substance IDs (DTXSID) for interoperability with other databases like the CompTox Chemicals Dashboard [9] [7].
    • Species: Taxonomic information (species, genus, family) is verified and standardized using integrated taxonomic resources [10].
    • Effect and Endpoint: Observed biological effects (e.g., Mortality, Immobilization, Growth) and the quantified metrics (e.g., LC50, EC50) are mapped to standardized terms [7].
    • Test Conditions: Experimental media, duration, and other methodological parameters are also described using controlled terms [10].

• Step 5: Quality Assurance and Publication

  • Objective: Ensure data accuracy and consistency before public release.
  • Methods: Extracted data undergoes automated checks and manual quality control. Newly curated data is added to the public knowledgebase in quarterly updates [9] [10].

Quantitative Scope of the ECOTOX Knowledgebase

The systematic and sustained application of this curation pipeline has resulted in a knowledgebase of remarkable scale and diversity. The following table summarizes the core data content of ECOTOX.

Table 1: Quantitative Data Inventory of the ECOTOX Knowledgebase (as of 2025) [9]

Data Category	Count	Description
Scientific References	> 53,000	Peer-reviewed literature and grey literature sources.
Unique Chemicals	> 12,000	Single chemical stressors, with links to CompTox Dashboard.
Ecological Species	> 13,000	Aquatic and terrestrial plant and animal species.
Total Test Results	> 1,000,000	Individual curated data records on chemical effects.

The data covers a wide array of biological effects and endpoints, which are standardized using CVs. The table below illustrates common categories.

Table 2: Common Ecotoxicity Effects and Endpoints Standardized in ECOTOX [7]

Taxonomic Group	Standardized Effect (CV)	Standardized Endpoint (CV)	Typical Test Duration
Fish	Mortality (MOR)	LC50 (Lethal Concentration 50%)	96 hours
Crustaceans	Mortality (MOR), Intoxication (ITX)	LC50 / EC50 (Effective Concentration 50%)	48 hours
Algae	Growth (GRO), Population (POP)	EC50 (e.g., growth inhibition)	72-96 hours

The Scientist's Toolkit: Research Reagent Solutions

Researchers leveraging ECOTOX or building similar curated systems utilize a suite of key resources and tools. The following table details these essential components.

Table 3: Essential Research Reagents and Resources for Ecotoxicity Data Curation

Item Name	Function in Research / Curation	Relevance to ECOTOX
CompTox Chemicals Dashboard	A comprehensive chemistry database and web-based suite of tools.	Provides verified chemical identifiers (DTXSID) and properties, ensuring chemical data interoperability [9] [7].
Controlled Vocabularies (CVs)	Standardized lists of terms for effects, endpoints, species, etc.	The core system for normalizing data from thousands of disparate studies, enabling reliable search and analysis [10].
Systematic Review Protocols	A framework for identifying, evaluating, and synthesizing scientific evidence.	ECOTOX's curation pipeline is built on these principles, ensuring transparency, objectivity, and consistency [10] [11].
ECOTOX User Interface (Ver 5)	The public-facing website for querying the knowledgebase.	Allows users to Search, Explore, and Visualize curated data using the underlying CVs for precise filtering [9].
Azadirachtin B	Azadirachtin B, CAS:95507-03-2, MF:C33H42O14, MW:662.7 g/mol	Chemical Reagent
Elaiomycin	Elaiomycin, CAS:23315-05-1, MF:C13H26N2O3, MW:258.36 g/mol	Chemical Reagent

Functional Applications and Data Interoperability

The true value of a curated database is realized through its application. ECOTOX supports a wide range of ecological research and regulatory functions. The diagram below illustrates how the curated data flows to support key applications.

• Support for Regulatory Decisions: ECOTOX data is used by local, state, and tribal governments to develop site-specific water quality criteria and to interpret environmental monitoring data for chemicals without established regulatory benchmarks [9]. It also informs ecological risk assessments for chemical registration under statutes like TSCA and FIFRA [9] [10].

• Enabling Predictive Modeling: The high-quality, curated data in ECOTOX is essential for developing and validating Quantitative Structure-Activity Relationship (QSAR) models and other New Approach Methodologies (NAMs) [9] [12]. By providing reliable experimental data, it helps build machine learning models to predict toxicity, reducing reliance on animal testing [7] [12]. For instance, the ADORE dataset is a benchmark for machine learning derived from ECOTOX, specifically created to facilitate model comparison and advancement [7].

The ECOTOX Knowledgebase exemplifies the critical importance of controlled vocabularies in managing large-scale scientific data. Through a rigorous, systematic curation pipeline, ECOTOX transforms heterogeneous ecological toxicity information from the global literature into a structured, reliable, and interoperable resource. This foundational work not only supports immediate regulatory and research needs but also provides the essential empirical data required to develop the next generation of predictive toxicological models, thereby contributing to a more efficient and ethical future for chemical safety assessment.

The exponential growth of chemical substances in commerce necessitates robust frameworks for ecological risk assessment and research. Central to this challenge is the management of vast, heterogeneous ecotoxicity data. A controlled vocabulary serves as the foundational element, standardizing terminology for test methods, species, endpoints, and chemical properties to enable data integration and knowledge discovery [4]. This application note details how a well-defined controlled vocabulary system directly enables three core benefits within ecotoxicology: reliable data search, seamless data interoperability, and regulatory acceptance. Adherence to the protocols and utilizations of the resources described herein is critical for researchers, scientists, and drug development professionals engaged in chemical safety and ecological research.

The implementation of a controlled vocabulary is exemplified by the ECOTOXicology Knowledgebase (ECOTOX), the world's largest curated compilation of ecotoxicity data [4]. The scale and diversity of data managed within this system underscore the necessity of a standardized terminology framework. The table below summarizes the quantitative scope of data enabled by this approach.

Table 1: Quantified Data Scope of the ECOTOX Knowledgebase (as of 2022)

Data Category	Metric	Count / Volume
Chemical Coverage	Unique Chemicals	> 12,000 chemicals [4]
Biological Species	Aquatic & Terrestrial Species	> 12,000 ecological species [4]
Test Results	Individual Toxicity Results	> 1 million test results [4]
Scientific References	Source Publications	> 50,000 references [4]
Data Sources	Aggregated Public Sources	> 1,000 worldwide sources [6]

Core Benefit Analysis and Experimental Protocols

Enabling Reliable and Reproducible Search

A controlled vocabulary overcomes the challenge of inconsistent terminology in the scientific literature, which otherwise hampers data retrieval. By enforcing a unified set of terms for organisms, effects, and conditions, it ensures search queries are comprehensive and reproducible.

Experimental Protocol 1: Systematic Literature Search and Data Curation via ECOTOX

This protocol outlines the steps for identifying and curating ecotoxicity studies from the open literature, ensuring only relevant and acceptable data are incorporated into a knowledgebase [13] [4].

• Literature Sourcing: Conduct systematic searches of scientific databases (e.g., PubMed) using predefined search strategies tailored for ecologically relevant toxicity data for single chemicals [4].
• Initial Screening (Phase I): Apply strict acceptability criteria to identify relevant papers. A study must meet all the following minimum criteria to be accepted [13]:
  • The toxic effects are related to single chemical exposure.
  • The effects are on an aquatic or terrestrial plant or animal species.
  • There is a biological effect on live, whole organisms.
  • A concurrent environmental chemical concentration/dose or application rate is reported.
  • There is an explicit duration of exposure.
• Data Extraction and Curation: For accepted studies, extract pertinent methodological details and results following well-established controlled vocabularies. This includes [4]:
  • Chemical identification using standard identifiers (e.g., CAS RN, DTXSID).
  • Test organism species and life stage.
  • Detailed exposure conditions (duration, medium).
  • Measured toxicity endpoints (e.g., LC50, EC50, NOEC) and values.
• Quality Control and Entry: Enter curated data into the knowledgebase using controlled vocabulary terms. Data is subjected to quality control checks before being made publicly available [4].

Facilitating Data Interoperability and Reusability

Controlled vocabularies act as a universal translator, allowing disparate datasets and computational tools to communicate effectively. This interoperability is a cornerstone of modern, integrated approaches to toxicology [6] [4].

Experimental Protocol 2: Integrating Curated Data with High-Throughput Screening (HTS) and Computational Tools

This protocol describes how curated in vivo data, standardized through a controlled vocabulary, is used to support and validate new approach methodologies (NAMs) and computational models [4].

• Data Export and Standardization: From a source like the ECOTOX Knowledgebase, export curated in vivo toxicity data for a set of chemicals of interest. Data is structured using standardized formats and identifiers [4].
• HTS Data Acquisition: Access corresponding high-throughput screening (HTS) data from programs such as ToxCast. These assays provide rapid, in vitro toxicity signatures for thousands of chemicals [6].
• Computational Modeling: Use the paired in vivo and in vitro data to build and validate quantitative structure-activity relationship (QSAR) models or other in silico prediction tools. The curated in vivo data serves as the biological anchor for model training and evaluation [4].
• Tool Interoperability: Leverage the CompTox Chemicals Dashboard, which interlinks chemical structures, properties, toxicity data (e.g., ToxValDB), and exposure information. The shared use of a controlled vocabulary and standard chemical identifiers (DTXSID) across these EPA tools enables seamless navigation and data integration [6].

Table 2: Key U.S. EPA Tools for Integrated Chemical Safety Assessment

Tool / Database Name	Primary Function	Role in Interoperability
ECOTOX Knowledgebase	Curated in vivo ecotoxicity data repository [4]	Provides foundational ecological effects data for modeling and assessment.
CompTox Chemicals Dashboard	Centralized access to chemical property and toxicity data [6]	Integrates data from multiple sources (ECOTOX, ToxCast, ToxValDB) using standardized chemical identifiers.
ToxCast	High-throughput in vitro screening assays [6]	Generates mechanistic toxicity data for chemical prioritization and predictive model development.
ToxValDB	Database of in vivo toxicity values and derived guideline values [6]	Provides standardized summary toxicity data from over 40 sources for comparison and use in assessments.

Supporting Regulatory Acceptance and Standardization

Regulatory bodies require transparent, objective, and consistent data for risk assessment. A controlled vocabulary is integral to systematic review practices, providing the structure needed for study evaluation and use in regulatory decisions [13] [4].

Experimental Protocol 3: Evaluation of Open Literature Studies for Ecological Risk Assessment

This protocol, based on EPA Office of Pesticide Programs (OPP) guidelines, details the process for reviewing open literature studies for use in regulatory ecological risk assessments, particularly for Registration Review and endangered species evaluations [13].

• Obtain Relevant Studies: Query the ECOTOX database to identify published open literature studies for the pesticide or chemical under review [13].
• Review Against Acceptance Criteria: Evaluate each study against a detailed set of OPP acceptance criteria, which expand upon the basic ECOTOX criteria. Key additional criteria include [13]:
  • The toxicology information is for a chemical of concern to OPP.
  • The article is a publicly available, full-text primary source in English.
  • A calculated endpoint (e.g., LC50) is reported.
  • Treatments are compared to an acceptable control.
  • The study location (lab/field) and test species are reported and verified.
• Study Classification and Documentation: Classify the study based on its quality and relevance. Complete an Open Literature Review Summary (OLRS) for tracking and transparency [13].
• Incorporate into Risk Assessment: Use the accepted quantitative data to derive toxicity values (points of departure) for risk characterization or use data qualitatively to inform mode of action or for use in a weight-of-evidence approach [13].

Table 3: Key Resources for Curated Ecotoxicity Data and Analysis

Resource Name	Type / Function	Brief Description
ECOTOX Knowledgebase	Curated Database	Authoritative source for single-chemical ecotoxicity data for aquatic and terrestrial species [4].
CompTox Chemicals Dashboard	Data Integration Tool	Web-based application providing access to chemical structures, properties, bioactivity, and toxicity data from multiple EPA databases [6].
ToxValDB	Toxicity Value Database	A large compilation of human health-relevant in vivo toxicology data and derived toxicity values from over 40 sources, designed for easy comparison [6].
Controlled Vocabulary	Data Standardization Framework	A standardized set of terms for test methods, species, and endpoints that enables reliable search and data interoperability [4].
OECD Document No. 54	Statistical Guidance	Provides assistance on the statistical analysis of ecotoxicity data to ensure scientifically robust and harmonized evaluations (currently under revision) [14].

Implementing Controlled Vocabularies: From Theory to Practice in Data Systems

Application Note: Integrating Controlled Vocabularies in Ecotoxicity Research

In ecotoxicity research, structured processes for literature search, review, and data curation are critical for ensuring data reliability, reproducibility, and reusability. The exponential growth of chemical substances and associated toxicity data necessitates robust methodologies that can efficiently handle vast information volumes. This application note examines established pipelines and protocols, emphasizing the central role of controlled vocabularies in standardizing ecotoxicity data across research workflows. By implementing systematic approaches, researchers can enhance data interoperability and support computational toxicology applications, including machine learning and new approach methodologies (NAMs) [15] [7] [4].

The ECOTOXicology Knowledgebase (ECOTOX) exemplifies the successful implementation of these principles, serving as the world's largest compilation of curated ecotoxicity data with over 12,000 chemicals and 1 million test results [4]. Similarly, the ADORE benchmark dataset demonstrates how structured curation practices facilitate machine learning applications in ecotoxicology [7]. These resources highlight how controlled vocabularies and standardized processes transform raw data into FAIR (Findable, Accessible, Interoperable, and Reusable) resources for the research community.

Table 1: Key Databases and Resources for Ecotoxicity Research

Resource Name	Primary Focus	Data Volume	Controlled Vocabulary System	Update Frequency
ECOTOX Knowledgebase	Ecological toxicity data	>12,000 chemicals, >1 million test results	EPA-specific taxonomy; standardized test parameters	Quarterly
ADORE Dataset	Acute aquatic toxicity ML benchmarking	3 taxonomic groups; chemical & species features	Taxonomic classification; chemical identifiers	Specific versions
MEDLINE/PubMed	Biomedical literature	>26 million citations	Medical Subject Headings (MeSH)	Continuous
CompTox Chemicals Dashboard	Chemical properties and toxicity	>350,000 chemicals	DSSTox Substance ID (DTXSID)	Regular updates

Table 2: Common Controlled Vocabulary Systems in Scientific Databases

Vocabulary System	Database Application	Scope and Coverage	Specialized Features
Medical Subject Headings (MeSH)	PubMed/MEDLINE	Hierarchical vocabulary for medical concepts	Automatic term mapping and explosion
Emtree	Embase	Biomedical and pharmacological terms	Drug and disease terminology
CINAHL Headings	CINAHL	Nursing and allied health	Intervention and assessment terms
EPA Taxonomy	ECOTOX Database	Ecotoxicology test parameters	Species, endpoints, experimental conditions

Protocols for Systematic Literature Review and Data Curation

Systematic Review Workflow for Ecotoxicity Literature

Systematic reviews in ecotoxicology employ transparent, objective methodologies to identify, evaluate, and synthesize evidence from multiple studies. The process involves five critical steps that ensure comprehensive coverage and minimize bias [16] [17].

Systematic Review Workflow Diagram

Step 1: Framing the Research Question

A well-structured research question is the foundation of any systematic review. For ecotoxicity studies, this typically follows the PICOT framework (Population, Intervention, Comparison, Outcome, Time) to define scope and key elements [16] [18]. The question should meet FINER criteria (Feasible, Interesting, Novel, Ethical, Relevant) to ensure practical and scientific value [18]. For example, in assessing chemical safety, a structured question would specify: the test species (population), chemical exposure (intervention), control groups (comparison), measured endpoints like LC50 (outcome), and exposure duration (time) [16] [7].

Step 2: Identifying Relevant Literature

Comprehensive literature search requires multiple strategies to capture all relevant studies. Best practices include:

• Search multiple databases including PubMed/MEDLINE, Embase, Cochrane Central, and specialized resources like ECOTOX [16] [18] [4]
• Combine controlled vocabulary and keywords to account for terminology variations [19]
• Implement citation tracking by examining references of relevant articles [18]
• Apply no language restrictions during initial search to minimize geographic bias [16]

For ecotoxicity research, specifically include specialized resources like the ECOTOX database, which employs systematic review procedures to curate toxicity data from published literature [4].

Step 3: Assessing Study Quality

Quality assessment evaluates potential biases and methodological robustness using established criteria [16]:

• Study design appropriateness for research question (e.g., randomized vs. observational)
• Exposure ascertainment accuracy and timing relative to intervention
• Outcome measurement validity, including blinding and follow-up duration
• Control of confounding factors through design or statistical adjustment

In ecotoxicology, the Klimisch score or similar systems categorize studies based on reliability, with high-quality studies providing definitive data for risk assessment [4].

Step 4: Summarizing the Evidence

Data synthesis involves extracting and combining results from included studies. Create standardized tables documenting:

• Study characteristics (author, year, design)
• Test organisms and chemical details
• Exposure conditions and durations
• Measured endpoints and effect values
• Statistical analyses and results

Synthesis can be narrative (descriptive summary) or quantitative (meta-analysis), depending on study homogeneity [16].

Step 5: Interpreting the Findings

Interpret results by considering quality assessments, potential biases, heterogeneity sources, and overall evidence strength. Evaluate publication bias and address implications for risk assessment and future research [16].

Data Curation Pipeline for Ecotoxicity Studies

Effective data curation ensures ecological toxicity data remain accessible and reusable for future applications. The CURATE(D) model provides a structured approach [20]:

Data Curation Pipeline Diagram

Check Files and Documentation

• Verify completeness of data transfer, especially for large datasets [21]
• Conduct file inventory to ensure all components are present
• Appraise and select appropriate files for curation and publication

Understand the Data

• Execute code/scripts to verify functionality and outputs [21]
• Perform quality control through calibration, validation, and normalization
• Review README files and documentation for completeness

Request Missing Information

• Identify data gaps requiring researcher clarification
• Track provenance of all changes and additions
• Establish communication with data creators for context clarification

Augment Metadata for Findability

• Assign persistent identifiers (DOIs) for dataset citation [21]
• Apply standardized metadata schemas appropriate for ecotoxicology
• Enhance discoverability through controlled vocabulary terms

Transform File Formats for Reuse

• Convert to open, non-proprietary formats (e.g., CSV instead of Excel) [21]
• Ensure long-term accessibility while preserving data integrity
• Consider publishing both raw and curated data when scientifically valuable

Evaluate for FAIRness

• Assess interoperability with related resources and tools
• Verify accessibility through appropriate licensing and access controls
• Ensure reusability through comprehensive documentation

Document All Curation Activities

• Maintain detailed records of all curation decisions and modifications
• Explain quality control methods applied to the data [21]
• Provide context for future users to understand data transformations

Implementation Example: The ECOTOX Database Pipeline

The ECOTOX database exemplifies a mature literature review and data curation pipeline for ecotoxicity data. Its systematic approach includes:

Literature Search and Acquisition

• Comprehensive source monitoring of over 1,200 scientific journals [4]
• Structured search strategies using controlled vocabulary and keywords
• Regular quarterly updates to incorporate new data

Data Extraction and Curation

• Systematic review procedures following documented guidelines [4]
• Structured data extraction using controlled vocabularies for:
  • Test organisms (species, taxonomy, life stage)
  • Chemical identifiers (CAS, DTXSID, InChIKey, SMILES)
  • Experimental conditions (duration, endpoints, media)
  • Results (effect values, statistical measures)
• Quality control through expert review and validation

Controlled Vocabulary Application

ECOTOX employs extensive controlled vocabularies to standardize:

• Taxonomic classification using standardized nomenclature
• Chemical identification with multiple identifier systems
• Endpoint categorization (e.g., LC50, EC50, NOEC)
• Effect types (mortality, growth, reproduction, etc.)
• Test media and conditions (freshwater, seawater, sediment)

This standardized approach enables interoperability with other resources like the CompTox Chemicals Dashboard and supports computational toxicology applications [4].

Table 3: Research Reagent Solutions for Ecotoxicity Studies

Resource Category	Specific Examples	Function and Application	Key Characteristics
Toxicity Databases	ECOTOX Knowledgebase, EnviroTox	Curated toxicity data for hazard assessment	Standardized test results; Quality-controlled data
Chemical Identification	CAS RN, DTXSID, InChIKey, SMILES	Unique chemical identifiers for tracking	Cross-database compatibility; Structural information
Benchmark Datasets	ADORE Dataset	Machine learning training and validation	Multiple taxonomic groups; Chemical and species features
Controlled Vocabularies	MeSH, Emtree, EPA Taxonomy	Standardized terminology for data retrieval	Hierarchical structure; Comprehensive coverage
Statistical Software	R, Python with pandas	Data analysis and modeling	Reproducible workflows; Extensive package ecosystems
Molecular Representations	SMILES, Molecular fingerprints	Chemical structure encoding for QSAR	Machine-readable formats; Structure-activity relationships

Experimental Protocol: Building a Curated Ecotoxicity Dataset

Protocol: Developing a Benchmark Dataset for Machine Learning

Experimental Aim

To create a comprehensive, curated dataset of acute aquatic toxicity values for machine learning applications, incorporating chemical, species, and experimental data with controlled vocabulary standards [7].

• Source Data: ECOTOX database (latest release)
• Taxonomic Coverage: Fish, crustaceans, algae
• Chemical Identifiers: CAS RN, DTXSID, InChIKey, SMILES
• Programming Tools: Python or R for data processing
• Metadata Standards: Domain-specific controlled vocabularies

Procedure

• Data Acquisition and Filtering

  • Download ECOTOX core tables (species, tests, results, media)
  • Filter entries for target taxonomic groups (fish, crustaceans, algae)
  • Select relevant endpoints (LC50, EC50) and exposure durations (48-96 hours)
  • Exclude in vitro tests and embryo-life stage tests [7]

• Data Harmonization

  • Standardize chemical identifiers using CAS RN, DTXSID, and InChIKey
  • Apply taxonomic classification using controlled vocabulary
  • Normalize effect values and units (convert to molar concentrations)
  • Categorize test media and conditions using standardized terms

• Feature Expansion

  • Add chemical descriptors (molecular weight, log P, functional groups)
  • Incorporate species traits (phylogenetic information, habitat preferences)
  • Include experimental conditions (temperature, pH, water hardness)
  • Generate molecular representations (SMILES, fingerprints)

• Quality Control and Validation

  • Implement outlier detection for extreme values
  • Verify chemical structure-identifier consistency
  • Cross-reference with other sources for data validation
  • Apply completeness assessment for critical fields

• Dataset Splitting and Documentation

  • Create predefined train-test splits based on chemical scaffolds
  • Develop comprehensive data dictionaries explaining all fields
  • Document all processing steps and decision rules
  • Publish in open, accessible formats with usage licenses

Expected Results

A standardized benchmark dataset (such as ADORE) containing:

• Core ecotoxicity measurements (LC50/EC50 values)
• Chemical characteristics and molecular representations
• Species information and phylogenetic context
• Experimental conditions and methodological details
• Predefined splits for model validation and comparison

This protocol supports the development of robust QSAR and machine learning models while ensuring FAIR data principles through comprehensive curation and controlled vocabulary application [7].

In ecotoxicology, the integration of data from diverse sources—including guideline studies and the open literature—is fundamental for robust ecological risk assessments (ERAs) [13]. However, the primary source language describing treatment-related endpoints is highly variable, creating significant barriers to data comparison, integration, and reuse [1]. A controlled vocabulary provides the solution: an authoritative set of standardized terms selected and defined to ensure consistent indexing and description of data [22]. Implementing such a vocabulary is essential for creating a findable, accessible, interoperable, and reusable (FAIR) dataset, which in turn is critical for regulatory decision-making, chemical prioritization, and the validation of predictive models [1] [9]. This document outlines the key components and protocols for building a controlled vocabulary for ecotoxicity data, providing a framework to enhance the consistency and transparency of ERA.

Core Components of an Ecotoxicity Controlled Vocabulary

A comprehensive controlled vocabulary for ecotoxicity data is built upon four foundational pillars. Standardizing these elements ensures that data from different studies can be systematically aggregated, queried, and interpreted.

Chemical Substance Identification

Unambiguous chemical identification is the cornerstone of any ecotoxicological database. Inconsistent naming (e.g., using trade names vs. systematic names) severely hampers data retrieval and integration.

• Structured Identifiers: Each chemical should be associated with unique identifiers from authoritative sources. The EPA's CompTox Chemicals Dashboard provides such identifiers, including DSSTox Substance Identifiers (DTXSID), which are crucial for linking chemical records across databases [6].
• Standardized Naming: The preferred chemical name should be consistent with international nomenclature standards (e.g., IUPAC). Synonyms and common names should be cataloged but linked to the primary identifier to ensure comprehensive searchability.

Test Species and Organism Profile

The test organism must be identified with sufficient taxonomic precision to allow for meaningful interspecies comparisons and extrapolations.

• Taxonomic Resolution: Species should be identified by their full binomial name (genus and species), and verified using authoritative taxonomic references [13] [23]. The National Center for Biotechnology Information (NCBI) Taxonomy Database provides unique taxonomy IDs that can be used for standardization [24].
• Life Stage and Source: The controlled vocabulary must include terms for life stage (e.g., neonate, juvenile, adult) and the source of the organisms (e.g., laboratory culture, field-collected), as these factors can significantly influence toxicity outcomes [23].

Ecotoxicological Endpoints

The biological effects measured in a study must be described using consistent terminology to enable cross-study analysis and meta-analysis.

• Endpoint Harmonization: Primary source language (e.g., "mortality," "death," "% dead") should be mapped to a single controlled term. Existing vocabularies such as the Unified Medical Language System (UMLS), OECD harmonized templates, and the BfR DevTox lexicon provide a strong foundation for describing prenatal developmental and other toxicological endpoints [1].
• Temporal and Statistical Descriptors: The vocabulary must include standardized terms for the effect measurement (e.g., EC50, LC50, NOEC, LOEC) and the exposure duration (e.g., 24-h, 48-h, 96-h, chronic) [13] [25]. This allows for clear differentiation between, for example, an acute 48-h LC50 and a chronic 28-day NOEC.

Test Conditions and Methodology

Detailed and standardized reporting of test conditions is necessary to evaluate the reliability and relevance of a study and to understand the context of the reported effects.

• Exposure System: Terms should describe the test location (laboratory vs. field), route of exposure (water, sediment, diet), and test system type (static, renewal, flow-through) [13] [26].
• Environmental Parameters: Key parameters such as water temperature, pH, hardness, and light regime must be included, as they can modulate chemical toxicity [23]. The vocabulary should standardize the units and measurement methods where possible.

Table 1: Core Components of an Ecotoxicity Controlled Vocabulary

Component	Description	Standardization Source Examples
Chemical Identity	Unique substance identification	CompTox Chemicals Dashboard (DTXSID), CAS RN [6]
Test Species	Taxonomic identity of organism	NCBI Taxonomy ID, Verified binomial name [24]
Ecotoxicological Endpoint	Measured biological effect	UMLS, OECD Templates, BfR DevTox Terms [1]
Test Conditions	Methodology & environment	CRED reporting criteria, EPA Evaluation Guidelines [13] [27]

Experimental Protocol: Implementing a Controlled Vocabulary for Data Integration

The following protocol describes a systematic approach for standardizing extracted ecotoxicity data using an augmented intelligence workflow, which combines automated mapping with expert manual review [1].

Protocol: Automated and Manual Vocabulary Mapping

Objective: To standardize raw endpoint descriptions from ecotoxicity studies into controlled terms, enabling the creation of a FAIR (Findable, Accessible, Interoperable, Reusable) dataset.

Materials and Reagents:

• Hardware: Standard computer workstation.
• Software: Python scripting environment (e.g., version 3.7 or higher) [1].
• Data Input: A dataset of ecotoxicity test results with endpoint descriptions recorded in the primary source language (e.g., from ECOTOX Knowledgebase, ECHA dossiers, or NTP reports) [1] [9].
• Key Resource - Controlled Vocabulary Crosswalk: A harmonized crosswalk file linking common endpoint descriptions to standardized terms from UMLS, OECD templates, and BfR DevTox [1].

Procedure:

• Data Extraction and Preparation: Assemble the legacy or newly extracted ecotoxicity data. Ensure the data on chemicals, species, endpoints, and test conditions are in a structured format (e.g., a spreadsheet or database table).
• Automated Mapping Execution: Run the pre-developed annotation code (e.g., Python script) designed to automatically match the primary source endpoint descriptions to the standardized terms in the controlled vocabulary crosswalk [1].
• Categorization of Mapped Data: Upon completion of the automated script, the data will be separated into two streams:
  • Stream A: Automatically Mapped. Endpoints for which the script found a direct and confident match in the crosswalk.
  • Stream B: Unmapped. Endpoints that were too general, ambiguous, or lacked a direct lexical match for automated mapping.
• Manual Review and Curation:
  • For Stream A, perform a quality control check on a subset of the automated mappings to identify potential extraneous matches or inaccuracies. It is estimated that about half of the automatically mapped terms may require this verification [1].
  • For Stream B, trained risk assessors or data curators must manually assign the appropriate controlled vocabulary terms using professional judgment and logic. This step is critical for complex or nuanced endpoint descriptions.
• Data Integration and Documentation: Merge the validated mapped data from Stream A and the manually curated data from Stream B into a final, standardized dataset. Document the entire process, including the version of the crosswalk used, the mapping script, and any manual decisions made, to ensure transparency and reproducibility.

The following workflow diagram illustrates this integrated process:

Successful implementation of a controlled vocabulary and the execution of high-quality ecotoxicity tests rely on specific, well-characterized materials and databases.

Table 2: Essential Research Reagents and Resources for Ecotoxicity Data Generation and Curation

Tool/Reagent	Function/Description	Application in Ecotoxicity
Reference Toxicant [23]	A standard chemical used to assess the sensitivity and performance consistency of a test organism batch.	Quality control; verifying organism health and test system reliability.
Certified Test Organisms [23]	Organisms of a known species, age, and life stage, sourced from reliable culture facilities.	Ensures test reproducibility and validity; required for guideline studies.
EPA ECOTOX Knowledgebase [9]	A comprehensive, publicly available database of single-chemical ecotoxicity effects.	Primary source for curated data; template for vocabulary structure.
Controlled Vocabulary Crosswalk [1]	A file mapping common terms to standardized vocabularies (UMLS, OECD, BfR).	Core resource for automating data standardization efforts.
CRED Evaluation Method [27]	A framework of criteria for evaluating the reliability and relevance of ecotoxicity studies.	Provides structured guidance for manual review and study inclusion.

The construction and implementation of a controlled vocabulary for the key components of ecotoxicity data are not merely an administrative exercise but a scientific necessity. By standardizing the language used to describe chemicals, species, endpoints, and test conditions, the ecotoxicology community can overcome significant barriers to data interoperability. The protocols and tools outlined herein provide a actionable path toward creating robust, FAIR datasets. This, in turn, enhances the reliability of ecological risk assessments, supports the development of predictive models, and ultimately informs better decision-making for the protection of environmental health.

In the domain of ecotoxicity data research, ensuring consistent terminology is paramount for data interoperability, systematic reviews, and computational toxicology. Controlled vocabularies (CVs) are organized arrangements of words and phrases used to index content and retrieve it through browsing or searching [28]. They provide a common understanding of terms, reduce ambiguity, and are essential for making data findable, accessible, interoperable, and reusable (FAIR) [1]. The Simple Knowledge Organization System (SKOS) is a World Wide Web Consortium (W3C) standard designed for representing such knowledge organization systems—including thesauri, classification schemes, and taxonomies—as machine-readable data using the Resource Description Framework (RDF) [29] [30] [31]. By encoding vocabularies in SKOS, concepts and their relationships become processable by computers, enabling decentralized metadata applications and facilitating the integration of data harvested from multiple, distributed sources [29] [32].

SKOS Core Components and Data Model

The SKOS data model is concept-centric, where the fundamental unit is an abstract idea or meaning, distinct from the terms used to label it [30] [33]. This model provides a standardized set of RDF properties and classes to describe these concepts and their interrelations.

Fundamental SKOS Constructs

• Concepts and Concept Schemes: A skos:Concept represents an idea or meaning within a knowledge organization system. Each concept is identified by a Uniform Resource Identifier (URI), making it a unique, web-accessible resource [33] [31]. Concepts are typically aggregated into a skos:ConceptScheme, which represents a complete controlled vocabulary, thesaurus, or taxonomy [30].
• Lexical Labels: Concepts are labeled using human-readable strings. SKOS defines three primary labeling properties:
  • skos:prefLabel (Preferred Label): The primary, authoritative name for a concept. A concept can have at most one prefLabel per language tag [30] [28].
  • skos:altLabel (Alternative Label): Synonyms, acronyms, or other variant terms for the concept. A concept can have multiple altLabels [30] [28].
  • skos:hiddenLabel: A variant string that is useful for text indexing and search but is not intended for display to end-users (e.g., common misspellings) [30].
• Documentation Properties: SKOS offers several properties to document concepts, all of which are sub-properties of skos:note. These include skos:definition for formal explanations, skos:scopeNote for information about the term's intended usage, and skos:example to illustrate application [30] [31].
• Semantic Relations: Concepts are interlinked through semantic relationships.
  • Hierarchical Relations: skos:broader and skos:narrower link a concept to others that are more general or specific, respectively. While not defined as transitive in the core model, SKOS also provides skos:broaderTransitive and skos:narrowerTransitive for inferring transitive closures [30] [28].
  • Associative Relations: skos:related links two concepts that are associatively related but not in a hierarchical fashion [30] [31].
• Mapping Properties: For interoperability between different concept schemes, SKOS provides mapping properties like skos:exactMatch, skos:closeMatch, skos:broadMatch, and skos:narrowMatch. These are used to declare mapping links between concepts in different vocabularies [30] [33].

Visualizing the SKOS Data Model

The following diagram illustrates the core structure and relationships within a SKOS-concept scheme, providing a visual representation of the components described above.

Implementation Protocols for Ecotoxicity Data

Implementing SKOS for standardizing ecotoxicity data involves a structured process from vocabulary selection to automated application. The following workflow outlines the key stages in this process.

Protocol 1: SKOS Vocabulary Development and Mapping Workflow

Detailed Methodological Steps:

• Extract Toxicological End Points: Begin by extracting treatment-related end points from primary study reports, legacy datasets, and database records. The original language used in the source documents should be recorded verbatim [1].
• Analyze Source Language Variability: Catalog the variation in terminology used to describe identical or similar end points. This analysis highlights the requirement for standardization to enable cross-study comparison and integration [1].
• Select Authority Vocabularies: Identify and select established, domain-specific controlled vocabularies to serve as the target for standardization. In ecotoxicity and developmental toxicology, relevant vocabularies often include:
  • Unified Medical Language System (UMLS): A compendium of many controlled vocabularies in the biomedical sciences, providing comprehensive coverage of clinical and biological terms [1].
  • BfR DevTox Database Lexicon: A harmonized set of terms specifically developed for application to developmental toxicity data [1].
  • OECD Harmonised Templates: Standardized terminology for reporting chemical test results [1].
• Build a Controlled Vocabulary Crosswalk: Create a crosswalk that maps terms from the selected authority vocabularies to each other and, where possible, to common phrases found in the extracted source data. A crosswalk is a structured document (e.g., a spreadsheet or RDF using skos:closeMatch/skos:exactMatch) that annotates the overlaps between different vocabularies [1]. This resource acts as a translation layer between source terms and standardized SKOS concepts.
• Automated SKOS Mapping: Develop and execute annotation code (e.g., in Python) to automatically process the extracted end points and map them to standardized SKOS concepts using the pre-defined crosswalk. This code typically employs string matching, lookup tables, and simple logic to assign the appropriate URI of a SKOS concept to each extracted end point [1].
• Manual Review and Quality Control: Manually review a significant portion of the automatically mapped end points to identify and correct inaccuracies or extraneous matches. Research indicates approximately 51% of automatically mapped terms may require such manual review. End points that are too general or require complex human logic to match often remain unmapped by automated processes and must be handled separately [1].
• Publish as FAIR Linked Data: The finalized, standardized dataset should be published using RDF serialization formats (e.g., RDF/XML, Turtle). SKOS concepts are identified with persistent URIs, enabling them to be linked and dereferenced on the web. This final step ensures the data adheres to FAIR principles [1] [31].

Quantitative Performance of Automated SKOS Mapping

The table below summarizes performance metrics from a real-world implementation of an automated SKOS mapping approach in toxicology, demonstrating its efficiency gains.

Table 1: Performance Metrics from an Automated Vocabulary Mapping Exercise in Toxicology [1]

Metric	NTP Extracted End Points	ECHA Extracted End Points
Total Extracted End Points	~34,000	~6,400
Automatically Standardized	75% (~25,500 end points)	57% (~3,650 end points)
Requiring Manual Review	~13,005 end points (51% of standardized)	~1,861 end points (51% of standardized)
Estimated Labor Savings	>350 hours	>350 hours

The Scientist's Toolkit: Essential SKOS Research Reagents

Implementing SKOS-based solutions requires a combination of conceptual resources, software tools, and technical standards. The following table details key components of the SKOS research toolkit.

Table 2: Key Research Reagents and Tools for SKOS Implementation

Item Name	Type	Function / Application
SKOS Core Vocabulary	Standard / Specification	The normative RDF vocabulary (classes & properties) for representing concept schemes, definitions, and semantic relations [32] [31].
Controlled Vocabulary Crosswalk	Data Resource	A mapping table that links terms from different source vocabularies (e.g., UMLS, DevTox, OECD) to enable automated translation and standardization of extracted data [1].
Annotation Code (e.g., Python Script)	Software Tool	Custom code that automates the application of the crosswalk to raw extracted data, matching source terms to standardized SKOS concept URIs [1].
RDF Triplestore	Database System	A database designed for the storage, query, and retrieval of RDF triples. Essential for managing and querying large SKOS vocabularies and linked data [33].
SPARQL Endpoint	Query Service	A protocol that allows querying RDF data using the SPARQL language. Enables complex queries over SKOS concepts and their relationships (e.g., finding all narrower terms) [31].
UMLS Metathesaurus	Authority Vocabulary	A large, multi-source vocabulary in the biomedical domain that can be leveraged as a target for standardizing ecotoxicity terms [1].
OECD Harmonised Templates	Authority Vocabulary	Standardized terminology for reporting chemical test results, providing authoritative terms for regulatory ecotoxicity data [1].
Pendulone	Pendulone, MF:C17H16O6, MW:316.30 g/mol	Chemical Reagent
Breviscapine	Breviscapine, MF:C21H18O12, MW:462.4 g/mol	Chemical Reagent

The implementation of SKOS provides a robust, standards-based framework for transforming disparate and variably labeled ecotoxicity data into a machine-readable, interoperable resource. By following the detailed protocols for vocabulary mapping, automation, and quality control, researchers can achieve significant efficiencies in data curation. The resulting FAIR datasets, structured as linked data, become a powerful foundation for advanced computational toxicology, predictive modeling, and integrative meta-analyses, ultimately accelerating research and informing regulatory decisions.

In ecotoxicity data research, Controlled Vocabularies (CVs) are standardized sets of terms and definitions that enable consistent annotation, retrieval, and integration of complex environmental health data. The practical workflow for querying and retrieving data using these vocabularies is foundational to computational toxicology and chemical risk assessment. This protocol details the application of CVs within key public data resources, outlining a standardized methodology for researchers, scientists, and drug development professionals to efficiently access high-quality, structured ecotoxicity data. The framework is built primarily on tools and databases provided by the U.S. Environmental Protection Agency's (EPA) CompTox initiative, which offers data freely for both commercial and non-commercial use [6].

The following tables summarize the core data resources that utilize controlled vocabularies for data query and retrieval. These resources provide the quantitative and qualitative data necessary for modern computational toxicology studies.

Table 1: Core Hazard and Exposure Data Resources

Resource Name	Data Type	Key Content & Coverage	Primary Use Case
ToxCast [6]	High-throughput screening	In vitro screening data for thousands of chemicals via automated assays.	Prioritization of chemicals for further testing; hazard identification.
ToxRefDB [6]	In vivo animal toxicity	Chronic, sub-chronic, developmental, and reproductive toxicity data from ~6,000 guideline studies on ~1,000 chemicals.	Anchoring high-throughput screening data to traditional toxicological outcomes.
ToxValDB [6]	Aggregated in vivo toxicology values	237,804 records covering 39,669 unique chemicals from over 40 sources, including toxicity values and experimental results.	Risk assessment; derivation of point-of-departure and safe exposure levels.
ECOTOX [6]	Ecotoxicology	Adverse effects of single chemical stressors to aquatic and terrestrial species.	Ecological risk assessment.

Table 2: Exposure, Chemistry, and Supporting Data Resources

Resource Name	Data Type	Key Content & Coverage	Primary Use Case
CPDat [6]	Consumer product & use	Mapping of chemicals to their usage or function in consumer products.	Chemical exposure assessment from product use.
SHEDS-HT & SEEM [6]	High-throughput exposure	Rapid exposure and dose estimates to predict potential human exposure for thousands of chemicals.	High-throughput exposure modeling for chemical prioritization.
DSSTox [6]	Chemistry	Standardized chemical structures, identifiers, and physicochemical properties.	Chemical identification and structure-based querying.
CompTox Chemicals Dashboard [6]	Aggregation & Curation	A centralized portal providing access to chemistry, toxicity, and exposure data for ~900,000 chemicals.	Primary interface for chemical lookup, data integration, and download.

Experimental Protocols for Data Query and Retrieval

This section provides detailed, step-by-step methodologies for executing key tasks within the researcher workflow, from chemical identification to advanced pathway analysis.

Protocol 1: Chemical Identification and List Assembly Using Controlled Vocabularies

Objective: To unambiguously identify a chemical of interest and its related substances (e.g., salts, hydrates) using standardized identifiers to assemble a target list for subsequent querying.

• Define Query Substance: Begin with a chemical name (e.g., "Bisphenol A"), CAS RN (e.g., "80-05-7"), or SMILES string.
• Access the CompTox Chemicals Dashboard: Navigate to the EPA's CompTox Chemicals Dashboard via its public URL.
• Perform Initial Search: Enter the query term into the main search bar. The Dashboard will resolve the query to a unique substance record using its internal controlled vocabulary (DSSTox Substance Identifier, or DTXSID).
• Identify Related Substances: On the resulting chemical summary page, locate the "Related Substances" list. This list, generated based on structural and registration rules, contains substances that are salts, hydrates, or other forms of the searched chemical. Note their DTXSIDs.
• Assemble Target List: Compile a list of all relevant DTXSIDs from the previous step. This list ensures that all relevant chemical forms are included in subsequent data queries, preventing data omission due to narrow identifier matching.

Protocol 2: Cross-Database Ecotoxicity Data Retrieval via Batch Query

Objective: To retrieve curated ecotoxicity data from the ECOTOX Knowledgebase for a pre-defined list of chemicals.

• Prepare Input File: Format your list of DTXSIDs (from Protocol 1) into a single-column text file.
• Access the ECOTOX Database: Navigate to the ECOTOXicology Knowledgebase (ECOTOX) interface. It is recommended to use a web browser compatible with the tool (e.g., if experiencing issues in Chrome, clear cache or try an alternative browser) [6].
• Initiate Advanced Search: Select the "Advanced Search" option.
• Upload Chemical List: In the "Chemical" section, choose the option to upload a list of identifiers. Select "DTXSID" as the identifier type and upload your prepared text file.
• Apply Taxonomic and Effect Filters: Use the controlled vocabulary provided in the interface to define your query scope:
  • In the "Species" section, select relevant taxonomic groups (e.g., "Osteichthyes" for fish, "Cladocera" for water fleas).
  • In the "Effect" section, select standardized endpoint terms (e.g., "Mortality", "Growth", "Reproduction").
• Execute Query and Review Results: Run the search. The system will return a tabular list of effects results. Review the data and utilize the export function to download the full dataset in a machine-readable format (e.g., CSV) for further analysis.

Protocol 3: In Vitro Bioactivity Profiling with ToxCast Data

Objective: To obtain and interpret high-throughput screening (HTS) bioactivity data for a target chemical list to inform potential modes of action.

• Navigate to ToxCast Data Download Page: Access the dedicated ToxCast data download page via the EPA CompTox website [6].
• Select Data Release: Choose the most recent version of the "InvitroDB" (e.g., InvitroDB v4.0).
• Download Data Files: Two key files are required:
  • Chemical Inventory: Download the file linking DTXSIDs to chemical names and structures.
  • Summary Hit-Call Data: Download the file containing "hit-calls" (binary activity outcomes) and potency values (e.g., AC50) across all assay endpoints.
• Filter and Merge Data: Using a computational environment (e.g., R, Python), filter the hit-call data for your list of DTXSIDs. Merge the activity data with the chemical inventory and assay annotation files.
• Map Activity to Pathways: Group the active assay endpoints using the provided controlled vocabulary for biological pathways (e.g., "Estrogen Receptor Agonism", "Mitochondrial Membrane Potential Disruption"). This creates a bioactivity profile for each chemical.

Protocol 4: Data Integration and Visualization for Chemical Prioritization

Objective: To integrate data from multiple streams (ecotoxicity, bioactivity, exposure) to support a holistic chemical assessment or prioritization decision.

• Compile Results: Gather the curated datasets from ECOTOX (Protocol 2) and ToxCast (Protocol 3). Incorporate additional data, such as high-throughput exposure predictions from the SEEM model or consumer product use information from CPDat, if available [6].
• Normalize and Scale Data: For visualization, normalize quantitative values (e.g., EC50 from ECOTOX, AC50 from ToxCast) using a logarithmic transformation. Scale the data to a consistent range (e.g., 0-1) for comparative heatmaps.
• Create an Integrated Visualization: Use a radar chart (also known as a spider web chart) to display multiple dimensions of the data for a single chemical, or a scatter plot to explore relationships between two key variables (e.g., ToxCast potency vs. ECOTOX potency) across a chemical set [34].
• Interpret Based on Weight-of-Evidence: Analyze the integrated visualization to identify consistent patterns across different data types. A chemical with high bioactivity in receptor assays, confirmed by in vivo ecotoxicity effects, and with widespread consumer exposure, would be flagged as a high priority for further investigation.

Workflow Visualization with Graphviz Diagrams

The following diagrams, generated using DOT language, illustrate the logical flow and key relationships within the data query and retrieval workflow. The color palette is restricted to the specified Google-derived colors for consistency and accessibility.

Chemical Data Query Workflow

CVs in Ecotox Data Model

The Scientist's Toolkit: Essential Research Reagent Solutions

This table details key materials, software, and data resources essential for executing the computational ecotoxicology workflows described in this protocol.

Table 3: Essential Reagents and Resources for Computational Ecotoxicology

Item Name	Type	Function & Application in Workflow
CompTox Chemicals Dashboard	Software / Data Portal	Primary interface for chemical identifier resolution, data aggregation, and batch downloading of chemistry, toxicity, and exposure data [6].
DSSTox Controlled Vocabularies	Data Standard	Standardized chemical identifiers (DTXSID) and nomenclature that enable precise linking of data across disparate sources [6].
ECOTOX Knowledgebase	Database	Curated source of single-chemical ecotoxicity test results for aquatic and terrestrial species, queryable using standardized taxonomic and effect terms [6].
ToxCast/Tox21 High-Throughput Screening Data	Database	In vitro bioactivity profiling data for hypothesizing molecular initiating events and potential modes of action for environmental chemicals [6].
R or Python Programming Environment	Software	Computational environment for data manipulation, statistical analysis, and custom visualization of integrated datasets obtained from the above resources.
Graphviz (DOT Language)	Software	Open-source tool for generating clear, reproducible diagrams of workflows and data relationships, as demonstrated in this protocol [35].
Olomoucine II	Olomoucine II, MF:C19H26N6O2, MW:370.4 g/mol	Chemical Reagent
Luteolin-4'-o-glucoside	Luteolin-4'-o-glucoside, MF:C21H20O11, MW:448.4 g/mol	Chemical Reagent

Overcoming Common Challenges and Optimizing Your Use of Ecotoxicity CVs

In ecotoxicity research, the integration of data from diverse sources—including standard guideline studies and non-standard (or "legacy") scientific investigations—presents a significant challenge due to inherent data heterogeneity. This heterogeneity arises from differences in experimental designs, measured endpoints, species, and reporting formats. Establishing a controlled vocabulary is a foundational step for normalizing this disparate information, making it findable, accessible, interoperable, and reusable (FAIR) [4]. This protocol details methods for curating and integrating ecotoxicity data using structured vocabularies and systematic processes, leveraging frameworks from established knowledgebases like the ECOTOXicology Knowledgebase (ECOTOX) and the Toxicity Values Database (ToxValDB) to support advanced research and risk assessment [4] [5].

Application Notes: Curating Data with a Controlled Vocabulary

The Role of a Controlled Vocabulary

A controlled vocabulary consists of predefined, standardized terms used to consistently tag and describe data. In ecotoxicity, this applies to key entities such as chemical identifiers, species names, measured endpoints, and experimental conditions.

• Purpose: It mitigates heterogeneity by ensuring that the same concept (e.g., "LC50," "mortality," "Oncorhynchus mykiss") is described identically across all datasets, regardless of the original terminology used in the source publication [4].
• Implementation: Major databases employ extensive controlled vocabularies. For example, the ECOTOX Knowledgebase uses predefined terms for over 12,000 chemicals and numerous ecological species, which allows for the systematic curation of over one million test results [4].

Key Components of an Ecotoxicity Vocabulary

The table below summarizes core components of a controlled vocabulary for ecotoxicity data integration.

Table 1: Core Components of a Controlled Vocabulary for Ecotoxicity Data

Vocabulary Component	Description	Example Terms
Chemical Identifiers	Standardized codes for unique chemical identification	DTXSID (DSSTox Substance ID), CAS RN, InChIKey, SMILES [5]
Species Taxonomy	Standardized organism names and taxonomic hierarchy	Scientific name (e.g., Daphnia magna), taxonomic family, common name [4]
Toxicity Endpoints	Standardized names for measured effects and outcomes	LC50 (Lethal Concentration 50), EC50 (Effect Concentration 50), NOEC (No Observed Effect Concentration), LOEC (Lowest Observed Effect Concentration) [4]
Experimental Conditions	Standardized descriptors of the test environment	"static", "flow-through", "renewal", "temperature", "pH", "light cycle" [4]
Effect Measurements	Standardized units and types of reported values	"mg/L", "µg/L", "% mortality", "inhibition of growth" [4]

Experimental Protocols

The following protocols outline the step-by-step process for integrating heterogeneous ecotoxicity data, from literature acquisition to finalized, accessible data records.

Protocol 1: Literature Review and Data Acquisition

This protocol describes the systematic process for identifying and acquiring relevant ecotoxicity studies.

• Objective: To comprehensively identify relevant scientific literature and extract raw data for subsequent curation.
• Materials: Access to scientific databases (e.g., PubMed, Scopus), reference management software, and data extraction forms.
• Procedure:
  • Search Strategy Development: Define search strings using key chemical names, species, and toxicity endpoints. Document the search strategy transparently [4].
  • Literature Retrieval: Execute searches across multiple bibliographic databases to ensure broad coverage.
  • Study Screening: Apply predefined inclusion/exclusion criteria (e.g., relevance of species, availability of dose-response data) to screen titles, abstracts, and full-text articles [4].
  • Data Extraction: For included studies, extract all pertinent methodological details and quantitative results into a structured staging database, preserving the original language from the source material [5].

Protocol 2: Data Curation and Standardization with Controlled Vocabulary

This critical protocol involves mapping the extracted raw data onto the standardized controlled vocabulary.

• Objective: To transform heterogeneous raw data into a consistent, structured format suitable for integration and analysis.
• Materials: Staging database with raw extracted data, controlled vocabulary lists, relational database management system (e.g., MySQL) [5].
• Procedure:
  • Vocabulary Mapping: Map free-text terms from the raw data to their corresponding standardized terms in the controlled vocabulary (e.g., map "fathead minnow" to Pimephales promelas).
  • Data Standardization: Execute scripts to convert all units to a standard system (e.g., all concentrations to µg/L), normalize chemical identifiers, and harmonize endpoint classifications [5].
  • Quality Control (QC): Implement a multi-tiered QC process. This includes automated checks for data type consistency and manual review by a second curator to verify mapping accuracy and identify any errors [5].
  • Database Integration: Load the standardized and QC-verified records into the master database (e.g., ToxValDB or ECOTOX), where they become accessible for querying and analysis [5].

Protocol 3: Data Integration and Analysis

This protocol covers the use of integrated, curated data to support research and assessment.

• Objective: To utilize the curated dataset for generating insights, such as developing species sensitivity distributions or supporting quantitative structure-activity relationship (QSAR) models.
• Materials: The finalized, curated ecotoxicity database.
• Procedure:
  • Query Construction: Use the standardized terms from the controlled vocabulary to construct precise database queries (e.g., "retrieve all LC50 values for DTXSID1020147 across all freshwater fish species").
  • Data Retrieval and Export: Execute the query and export the consistent results for analysis.
  • Meta-Analysis: Perform statistical analyses on the integrated data. For example, the curated in vivo data from ToxValDB can be used to benchmark and validate new approach methodologies (NAMs) like high-throughput in vitro assays [5].

Workflow Visualization

The following diagram illustrates the end-to-end data integration workflow, from initial literature search to final application in risk assessment and research.

The Scientist's Toolkit: Research Reagent Solutions

The table below lists essential resources and tools for conducting ecotoxicity data integration projects.

Table 2: Essential Resources for Ecotoxicity Data Integration

Tool / Resource	Function	Relevance to Data Integration
ECOTOX Knowledgebase	A curated database of single-chemical ecotoxicity data for aquatic and terrestrial species [4].	Provides a model for systematic review procedures and a vast source of already curated data for use in assessments.
ToxValDB	A compiled database of human health-relevant in vivo toxicology data and derived toxicity values [5].	Demonstrates the process of standardizing data from multiple sources into a singular resource for comparison and modeling.
CompTox Chemicals Dashboard	A portal providing access to chemical properties, hazard data, and links to toxicity databases [6].	A key tool for obtaining standardized chemical identifiers (DTXSIDs) and sourcing related hazard data.
Controlled Vocabularies	Predefined lists of standardized terms for chemicals, species, and endpoints.	The fundamental tool for ensuring consistency and interoperability across disparate datasets [4].
Relational Database (e.g., MySQL)	A structured system for storing and managing large, complex datasets.	Provides the technical infrastructure for housing the staged, raw, and finalized standardized data [5].
Fluostatin A	Fluostatin A|Dipeptidyl Peptidase III (DPP3) Inhibitor	Fluostatin A is a potent DPP3 inhibitor for research. This product is For Research Use Only and is not intended for diagnostic or personal use.

In ecotoxicology research, the precise and consistent description of chemicals, species, and toxicological effects is fundamental to data integrity, retrieval, and interoperability. A controlled vocabulary is a carefully selected list of predefined, authorized terms used to tag units of information so they may be more easily retrieved by a search [36]. These vocabularies solve critical problems of homographs (same spelling, different meanings), synonyms (different words for the same concept), and polysemes by establishing a one-to-one correspondence between concepts and preferred terms [37] [36].

The need for such control is particularly acute in ecotoxicology, where data from diverse sources—scientific literature, government reports, and laboratory studies—must be integrated and compared. The ECOTOXicology Knowledgebase (ECOTOX), a leading curated database, relies on systematic review and controlled vocabularies to provide reliable single-chemical toxicity data for over 12,000 chemicals and ecological species [10]. Without vocabulary control, searches may fail to retrieve relevant studies, and computational models may be built on inconsistent data, ultimately compromising chemical safety assessments and ecological risk characterizations.

Core Challenges in Ecotoxicity Terminology

Ecotoxicology data management faces several specific terminology challenges that controlled vocabularies are designed to overcome.

Synonymy and Variant Terminology

A single concept is often described using different terms across the scientific literature. For example, a sweetened carbonated beverage might be referred to as a "soda," "pop," or "soft drink" [37]. In ecotoxicology, this phenomenon extends to chemical names (e.g., "Dicamba" vs. its systematic IUPAC name), species nomenclature (common vs. scientific names), and effect descriptions. This inconsistency means that a search for one term may miss relevant data tagged with a synonym, adversely affecting the recall of information retrieval systems [36].

Homography and Ambiguity

The same term can have multiple meanings, leading to ambiguity and reduced precision in search results. The word "pool," for instance, could refer to a swimming pool or the game of pool, and must be qualified to ensure each heading refers to only one concept [36]. In a scientific context, "absorption" has a specific meaning in toxicology (uptake of a chemical into general circulation) that must be distinguished from its broader meanings [38].

Linguistic Evolution and Field-Specific Usage

Scientific language evolves, and controlled vocabularies must be updated to remain relevant, a process guided by the principles of user warrant (what terms users are likely to use), literary warrant (what terms are generally used in the literature), and structural warrant (considering the vocabulary's own structure) [36]. Furthermore, the level of specificity of terms must be carefully considered to balance detail with usability [36].

Table 1: Core Terminology Challenges in Ecotoxicology Data

Challenge Type	Description	Ecotoxicology Example	Impact on Data Retrieval
Synonymy	Multiple terms for the same concept.	"Immobilization" vs. "Intoxication" in crustacean tests [7].	Low recall: misses relevant data.
Homography	Same term for multiple concepts.	"LC50" in fish vs. algae tests (may represent different effect types) [7].	Low precision: retrieves irrelevant data.
Variant Spelling	Differences in spelling conventions.	American vs. British English (e.g., "behavior" vs. "behaviour").	Low recall and fragmented results.
Structural Variation	Different levels of term specificity.	"Fish" vs. "Rainbow trout" (Oncorhynchus mykiss).	Inconsistent hierarchical organization.

Foundational Strategies and Protocols

Establishing a robust controlled vocabulary requires a systematic approach to term selection, organization, and management.

Establishing the Preferred Term

The process begins by designating a single preferred term for each unique concept. This involves:

• Controlling Synonyms: All terms representing the same concept are brought together under one preferred term. For example, in the Library of Congress Subject Headings (LCSH), "Young adults" is the preferred term, with "Young people" and "Young persons" as non-preferred variants [37].
• Disambiguating Homographs: Homographs are clarified with qualifiers. For example, "Bridges (Dentistry)" is used for a partial denture, while "Bridges" alone refers to the structures crossing rivers [37].

Creating a Syndetic Structure

A controlled vocabulary is not a simple list; it is a network of relationships. This syndetic structure is created by identifying and linking related terms [37] [36]:

• Broader Terms (BT): Point to a concept with a wider meaning.
• Narrower Terms (NT): Point to a more specific concept.
• Related Terms (RT): Point to a conceptually associated term.

For instance, in an ecotoxicology thesaurus, "Acute toxicity" might have a narrower term "LC50," and a related term "Bioassay" [38].

Protocol for Vocabulary Maintenance

Controlled vocabularies are dynamic and require ongoing curation. The following protocol ensures their long-term utility:

• Regular Audits: Periodically review terms for relevance, usage frequency, and the emergence of new concepts.
• Update Cycle: Establish a quarterly or annual schedule for adding new terms and deprecating obsolete ones, as done with the USGS Thesaurus [39].
• Documentation: Maintain detailed Standard Operating Procedures (SOPs) for literature search, review, and data curation, as practiced by the ECOTOX team [10].
• Stakeholder Input: Incorporate feedback from users (scientists, indexers) to ensure the vocabulary meets real-world needs (user warrant).

Application in Ecotoxicology: The ECOTOX Model

The ECOTOX Knowledgebase exemplifies the application of controlled vocabulary in ecotoxicology research, following a meticulous pipeline for data curation.

Data Curation Workflow

The process of incorporating data into ECOTOX involves multiple stages of screening and extraction, ensuring only relevant and high-quality data is added [10].

Controlled Vocabulary in Action

Within this workflow, controlled vocabulary is applied during data extraction and curation to ensure consistency [10]:

• Chemical Identifiers: Chemicals are standardized using unique identifiers like CAS numbers, DSSTox Substance IDs (DTXSID), and InChIKeys to unambiguously link toxicity data to specific chemical structures [7] [10].
• Species Taxonomy: Species are verified and tagged with full taxonomic hierarchy (kingdom, phylum, class, order, family, genus, species) to enable grouping and searching by taxonomic level [7].
• Effects and Endpoints: Toxicological outcomes are categorized using controlled terms. For example, the effect "MOR" (Mortality) and the endpoint "LC50" (Lethal Concentration for 50% of the population) are used for fish, while for crustaceans, "ITX" (Intoxication/Immobilization) is an accepted effect comparable to mortality [7].

Table 2: Key Controlled Vocabularies and Standards for Ecotoxicity Research

Vocabulary Category	Purpose	Examples & Standards	Function in Research
Chemical Identifiers	Uniquely and unambiguously identify substances.	CAS Registry Number, DSSTox ID (DTXSID), InChIKey [7] [10].	Links toxicity data to specific molecular structures; enables data integration across databases.
Taxonomic Classification	Standardize species nomenclature and classification.	Integrated Taxonomic Information System (ITIS), species hierarchy (Kingdom->Species) [7].	Allows grouping of data by taxonomic group (e.g., all fish); supports cross-species comparisons.
Toxicological Endpoints	Define and standardize measured outcomes of tests.	LC50, EC50, NOEC, LOEC; Acute vs. Chronic [38] [7].	Ensures consistent interpretation and quantitative comparison of toxicity results across studies.
Experimental Parameters	Describe test conditions and methodologies.	Controlled terms for exposure duration, test medium, organism life stage [7] [10].	Provides necessary context for interpreting results and assessing study quality and relevance.

The Scientist's Toolkit: Research Reagent Solutions

Successful implementation of controlled vocabularies relies on both conceptual frameworks and practical tools.

Table 3: Essential Tools for Implementing Controlled Vocabularies

Tool / Resource	Category	Brief Description & Function
Library of Congress Subject Headings (LCSH)	Subject Heading List	A comprehensive, widely adopted subject heading system that provides a model for establishing preferred terms and syndetic structure [37] [36].
ECOTOX Knowledgebase	Domain-Specific Database	A curated database demonstrating the application of controlled vocabularies for chemicals, species, and endpoints in ecotoxicology; serves as a practical reference [7] [10].
USGS Thesaurus	Thesaurus	A structured, hierarchical controlled vocabulary for scientific concepts relevant to earth sciences, providing a template for building domain-specific term relationships [39].
Medical Subject Headings (MeSH)	Thesaurus	The U.S. National Library of Medicine's controlled vocabulary thesaurus used for indexing articles, illustrating deep indexing in a life science domain [36].
Chemical Abstracts Service (CAS) Registry	Chemical Database	The authoritative source for unique chemical identifiers (CAS Numbers), essential for normalizing chemical data [7].

Implementation Framework and Quality Control

Deploying a controlled vocabulary is a strategic process that requires careful planning and continuous quality assurance. The following diagram outlines the key stages in the lifecycle of a controlled vocabulary.

Implementation Protocol

• Scope and Governance: Define the boundaries of the vocabulary and assign a governance body responsible for its management and evolution.
• Pilot Testing: Apply the draft vocabulary to a subset of existing data. Measure indexing consistency by having multiple indexers tag the same documents and calculating the rate of agreement.
• Integration and Training: Integrate the vocabulary into data management systems, often as pull-down menus in cataloging interfaces [37]. Conduct training sessions for all users and indexers to ensure consistent application.
• Performance Monitoring: Regularly assess the vocabulary's effectiveness by tracking search success rates, user feedback, and the frequency of use for individual terms.

Quality Control Metrics

• Indexing Consistency: A key metric for quality control is the level of agreement between different indexers when applying the controlled vocabulary to the same document. High consistency indicates clear and unambiguous terms.
• User Warrant Tracking: Monitor the terms used by researchers in their searches and publications to ensure the controlled vocabulary remains aligned with the language of the community [36].
• Update and Stability Balance: Maintain a log of vocabulary changes. While updates are necessary to stay current, excessive changes can destabilize the system and confuse users. The USGS Thesaurus, for example, adds new terms annually to maintain this balance [39].

The expansion of open literature data presents both an opportunity and a challenge for ecotoxicity research. While data availability has increased dramatically, consistent application of reliability and relevance criteria remains limited, potentially compromising the validity of chemical hazard assessments and ecological risk evaluations. Controlled vocabulary serves as the foundational element that enables standardized data interpretation across different studies and platforms, ensuring that terminology describing toxicological effects, test organisms, exposure conditions, and experimental methodologies is consistently applied and computationally tractable. Without such standardization, meta-analyses and systematic reviews encounter significant interoperability challenges that can undermine evidence-based decision-making.

The ecotoxicological study reliability (EcoSR) framework has emerged as a comprehensive tool for assessing the inherent scientific quality of ecotoxicity studies, specifically designed for toxicity value development [40]. This framework addresses a critical gap in ecological risk assessment by providing a systematic approach for evaluating potential biases and methodological soundness—a process that has been more established in human health assessments than in ecotoxicology. By integrating this framework with controlled vocabulary protocols, researchers can achieve greater transparency, consistency, and reproducibility in their evaluations of open literature data.

Theoretical Foundation: The EcoSR Framework and Controlled Vocabulary

The EcoSR Framework Structure

The EcoSR framework employs a two-tiered approach to evaluate study reliability [40]. Tier 1 constitutes an optional preliminary screening that allows for rapid triage of studies based on predefined exclusion criteria, such as incomplete reporting or fundamental methodological flaws. Tier 2 involves a full reliability assessment that examines the internal validity of studies through evaluation of potential biases across multiple methodological domains. This structured approach enables researchers to consistently apply reliability criteria, thereby enhancing the objectivity of study evaluations.

The framework builds upon traditional risk of bias (RoB) assessment methods frequently applied in human health assessments but incorporates key criteria specific to ecotoxicity studies [40]. These domain-specific considerations include aspects unique to ecotoxicological testing, such as test organism husbandry, environmental relevance of exposure scenarios, and endpoint measurement techniques appropriate for various species and life stages. The flexibility of the EcoSR framework allows for customization based on specific assessment goals, chemical classes, and regulatory contexts.

Integration with Controlled Vocabulary

Controlled vocabulary establishes a standardized terminology system that enables precise communication of EcoSR application results and methodological details. The implementation of controlled vocabulary ensures that key concepts—including test organisms, life stages, exposure pathways, measured endpoints, and statistical analyses—are consistently described across studies and research groups. This semantic standardization is particularly crucial for computational approaches to data mining and evidence synthesis, as it enables automated extraction and categorization of experimental details from diverse literature sources.

The integration of controlled vocabulary with the EcoSR framework occurs at multiple levels:

• Standardized reliability ratings (e.g., "high reliability," "moderate reliability," "low reliability") with explicitly defined criteria for each category
• Consistent documentation of methodological elements subject to evaluation, including test substance characterization, experimental design, and statistical analysis
• Uniform reporting of relevance considerations pertaining to ecological realism and regulatory applicability
• Systematic organization of toxicity data for subsequent benchmarking and dose-response modeling

Table 1: Core Components of the EcoSR Framework Integrated with Controlled Vocabulary

Framework Component	Description	Controlled Vocabulary Application
Tier 1: Preliminary Screening	Rapid assessment using predefined exclusion criteria	Standardized exclusion reasons (e.g., "missing control group," "inadequate exposure verification")
Tier 2: Full Reliability Assessment	Comprehensive evaluation of internal validity	Uniform bias domains (e.g., "selection bias," "performance bias," "detection bias")
Risk of Bias Evaluation	Assessment of potential systematic errors in methodology	Standardized bias ratings (e.g., "low risk," "high risk," "unclear risk") with explicit criteria
Relevance Assessment	Evaluation of ecological and regulatory applicability	Consistent relevance categories (e.g., "species relevance," "endpoint relevance," "exposure relevance")
Reporting Standards	Documentation of assessment rationale and outcomes	Structured reporting templates for reliability and relevance determinations

Application Protocol: Implementing the EcoSR Framework

Data Identification and Preparation

The initial phase involves systematic literature retrieval using predefined search strategies aligned with the research question. Search syntax should incorporate controlled vocabulary terms specific to ecotoxicology, such as standardized chemical identifiers, taxonomic nomenclature, and endpoint terminology. Following identification, studies should be cataloged using a reference management system with consistent tagging based on preliminary characteristics (e.g., test species, chemical class, exposure duration).

Data extraction prerequisites include:

• Development of customized data extraction forms that reflect assessment-specific information needs
• Training of evaluators in both the EcoSR framework and applicable controlled vocabulary
• Pilot testing of the extraction and evaluation process on a subset of studies to refine protocols
• Establishment of conflict resolution procedures for addressing discrepant evaluations between assessors

Tier 1: Preliminary Screening Implementation

The preliminary screening involves sequential evaluation against exclusion criteria defined a priori based on assessment objectives [40]. The screening should be conducted by at least two independent evaluators, with disagreements resolved through consensus or third-party adjudication. Exclusion criteria typically include:

• Incomplete reporting of essential study elements (e.g., missing measures of variability, insufficient exposure characterization)
• Fundamental methodological flaws that invalidate results (e.g., inappropriate statistical methods, grossly contaminated controls)
• Irrelevance to assessment objectives (e.g., wrong taxonomic groups, unrelated endpoints)
• Duplicate publication or secondary reporting without original data

Studies proceeding beyond Tier 1 advance to full reliability assessment, while excluded studies should be documented with specific rationale for exclusion, using controlled vocabulary terms to ensure consistent recording.

Tier 2: Full Reliability Assessment Methodology

The full reliability assessment comprises multiple evaluation domains, each addressing specific potential biases [40]. For each domain, evaluators assign reliability ratings based on explicit criteria, with supporting documentation referencing specific aspects of the study methodology.

Table 2: EcoSR Evaluation Domains and Assessment Criteria

Evaluation Domain	Key Assessment Criteria	Reliability Indicators	Potential Bias Sources
Test Substance Characterization	Purity verification, stability testing, concentration verification	Analytical confirmation of test concentrations, documentation of vehicle compatibility	Contamination, degradation, inaccurate dosing
Test Organism Considerations	Species identification, life stage specification, health status, acclimation	Certified specimen sources, standardized culturing conditions, adequate acclimation period	Genetic heterogeneity, inappropriate life stage, poor organism health
Experimental Design	Randomization, blinding, control groups, replication	Random assignment to treatments, blinded endpoint assessment, appropriate control types	Selection bias, performance bias, confounding factors
Exposure Characterization	Duration, route, medium, loading, renewal frequency	Measured concentrations, stability maintenance, appropriate media renewal	Nominal instead of measured concentrations, unstable exposure conditions
Endpoint Measurement	Method validity, precision, timing, relevance	Standardized measurement protocols, appropriate timing relative to exposure, validated methods	Detection bias, measurement error, subjective scoring
Statistical Analysis	Appropriate methods, assumptions testing, reporting completeness	Assumption verification, adequate statistical power, complete results reporting	Selective reporting, inappropriate tests, insufficient power

Data Synthesis and Reliability Integration

Following individual study evaluations, reliability assessments should be incorporated into the overall data synthesis approach. Several methods are available for integrating reliability considerations:

• Weighting approaches that assign greater influence to higher-reliability studies in quantitative analyses
• Stratified analyses that present results separately for different reliability categories
• Sensitivity analyses that examine the robustness of conclusions to inclusion criteria based on reliability ratings

The integration of controlled vocabulary enables computational approaches to these syntheses by providing standardized descriptors for reliability ratings and methodological characteristics. Throughout this process, documentation should be maintained using structured templates that capture both the final reliability determinations and the rationale supporting these judgments.

Visualization Protocols for Evaluation Frameworks

EcoSR Application Workflow

The following diagram illustrates the sequential workflow for applying the EcoSR framework to open literature data, incorporating both reliability assessment and controlled vocabulary implementation:

Controlled Vocabulary Implementation Structure

The relationship between controlled vocabulary components and their application in reliability assessment is visualized below:

Research Reagent Solutions for Ecotoxicity Data Evaluation

The implementation of evaluation frameworks requires both conceptual methodologies and practical tools. The following table details key research solutions essential for applying reliability and relevance criteria to open literature data:

Table 3: Essential Research Reagent Solutions for Ecotoxicity Data Evaluation

Research Solution	Function in Evaluation Framework	Application Protocol
EcoSR Framework	Comprehensive tool for assessing inherent scientific quality of ecotoxicity studies	Apply two-tiered approach: preliminary screening (Tier 1) followed by full reliability assessment (Tier 2) with customization based on assessment goals [40]
Controlled Vocabulary Systems	Standardized terminology for consistent data description and computational interoperability	Implement structured terminologies for test organisms, methodologies, endpoints, and reliability ratings using domain-specific ontologies
Critical Appraisal Tools (CATs)	Structured instruments for evaluating methodological quality and potential biases	Adapt existing CATs to ecotoxicology context while addressing full range of biases relevant to internal validity [40]
Reference Management Software	Organization and tracking of literature sources throughout evaluation process	Utilize systems with customizable tagging fields aligned with controlled vocabulary and reliability assessment categories
Data Extraction Platforms	Systematic capture of study details and methodological characteristics	Employ structured electronic forms with predefined fields corresponding to EcoSR evaluation domains
Digital Color Contrast Checkers	Verification of accessibility standards in visualization components	Ensure minimum contrast ratio of 4.5:1 for normal text and 3:1 for large text in all research outputs [41]

The integration of the EcoSR framework with controlled vocabulary systems represents a significant advancement in the critical evaluation of open literature data for ecotoxicity research. This structured approach enhances the transparency, consistency, and reproducibility of reliability and relevance assessments, ultimately strengthening the scientific foundation for ecological risk assessment and regulatory decision-making. The standardized protocols and visualization strategies presented in this document provide researchers with practical methodologies for implementing these evaluation frameworks, while the specific reagent solutions offer tools for operationalizing these assessments in diverse research contexts. As ecotoxicology continues to evolve with increasing data availability and computational approaches, such standardized evaluation frameworks will be essential for ensuring that data quality keeps pace with data quantity.

The integration of New Approach Methodologies (NAMs) and complex emerging data types into ecotoxicity and drug development research necessitates a parallel evolution in how scientific careers are documented. A Curriculum Vitae (CV) must now function not only as a record of past experience but as a structured, computationally accessible dataset that demonstrates a researcher's proficiency with modern data standards. This protocol provides a detailed framework for creating CVs that are interoperable with the FAIR (Findable, Accessible, Interoperable, and Reusable) data principles, ensuring they effectively communicate expertise in NAMs and advanced data types to both automated screening systems and human reviewers within the context of controlled vocabulary for ecotoxicity data research [1].

Core Principles: Alignment with Controlled Vocabularies and FAIR Data

A future-proof CV should mirror the structured data annotation processes used in modern toxicology. The core principle involves treating each CV entry not as free-form text, but as a data point annotated with standardized terms from established controlled vocabularies and ontologies [1]. This approach ensures semantic clarity and enables computational parsing and comparison.

• Interoperability through Standardization: Just as the Unified Medical Language System (UMLS) and OECD harmonized templates provide a common language for toxicological endpoints, a CV should use consistent, industry-recognized terms for skills, techniques, and achievements [1]. This avoids ambiguity and ensures that both Applicant Tracking Systems (ATS) and specialist reviewers correctly interpret your expertise.
• Structured for Human and Machine Readability: The CV's format must balance ATS compatibility with visual clarity for human experts. A clean, single-column layout with standard section headings and devoid of complex tables or graphics is essential for ATS parsing [42]. This technical foundation can then be enhanced with subtle strategic visual elements to guide the human eye [43].

Application Notes: CV Design Protocol for NAM Researchers

Protocol 1: ATS-Optimized Document Structure

Objective: To create a CV that is correctly parsed and ranked by Applicant Tracking Systems, ensuring it reaches a human reviewer.

Methodology:

• File Format and Naming: Save the final CV as a PDF file unless another format is specified. Use a professional file name, e.g., YourName_CV_NAMs.pdf [42].
• Layout and Typography: Employ a single-column layout with standard, readable fonts such as Arial, Calibri, or Times New Roman (10–12 pt). Use bold for section headings and bullet points for listings. Ensure generous white space to improve readability [42] [44].
• Section Headings: Use conventional, machine-readable section headings (e.g., "Professional Summary," "Technical Skills," "Research Experience," "Publications," "Certifications") [43].
• Color and Contrast: If using color, ensure sufficient contrast between text and background. For accessibility and legibility, a contrast ratio of at least 4.5:1 is recommended for standard text [45]. The provided color palette (e.g., #4285F4, #34A853, #202124) should be applied with this rule in mind.

Troubleshooting:

• Problem: CV is rejected by an ATS portal.
• Solution: Verify the document does not contain tables, text boxes, headers, footers, or images. Use a simple, linear structure. Online ATS simulator tools can be used for pre-validation [43].

Protocol 2: Integration of Controlled Vocabulary for Skills and Experience

Objective: To annotate skills and research experiences using standardized terms, enhancing discoverability in keyword searches and demonstrating domain-specific knowledge.

Methodology:

• Skills Mapping:
  • Extract key terms from target job descriptions and relevant literature.
  • Map personal skills to standardized terms from ontologies like UMLS, BfR DevTox, or other relevant controlled vocabularies [1].
  • Create a dedicated "Technical Skills" section near the top of the CV, categorizing skills (e.g., "Bioinformatics," "In Vitro Toxicology," "Data Visualization") [44].
• Experience Annotation:
  • For each research position or project, describe accomplishments using the standardized vocabulary.
  • Employ the Challenge-Action-Result (CAR) framework to structure bullet points [44].
  • Challenge: Briefly state the research problem.
  • Action: Describe the specific NAMs or techniques used, employing the standardized terms.
  • Result: Quantify the outcome with measurable data and state the impact.

Experimental Results Summary: Table 1 demonstrates the application of this protocol, comparing traditional CV entries with those enhanced by a controlled vocabulary. This reflects the data standardization process used in automated toxicology data mapping, which successfully standardized 75% of extracted endpoints in a recent study [1].

Table 1: Comparison of Traditional vs. Standardized Vocabulary CV Entries

Research Aspect	Traditional CV Wording	Standardized Vocabulary Wording (Based on Controlled Terms)	Quantitative Impact
High-Throughput Screening	"Ran cell-based assays"	"Execated high-throughput screening (HTS) using 3D hepatocyte spheroids to assess hepatotoxicity. Challenge: Need for human-relevant liver model. Action: Applied high-content analysis (HCA). Result: Identified 3 lead compounds with reduced toxicity, accelerating candidate selection."	Accelerated candidate selection by 2 weeks.
Computational Toxicology	"Did computer modeling"	"Developed a quantitative structure-activity relationship (QSAR) model for developmental toxicity prediction. Challenge: High cost of in vivo testing. Action: Utilized OECD QSAR Toolbox and KNIME analytics platform. Result: Model achieved 85% concordance with in vivo data, reducing animal use by 50% for priority ranking."	Reduced animal use by 50%.
Data Curation & Integration	"Collected and organized data"	"Curated and annotated legacy in vivo developmental toxicity studies using a harmonized controlled vocabulary crosswalk (UMLS, OECD). Challenge: Non-FAIR data. Action: Applied automated annotation code (Python). Result: Standardized 75% of extracted endpoints, creating a computationally accessible dataset for predictive modeling [1]."	Automated standardization of 75% of endpoints.

Protocol 3: Visualizing Expertise and Workflows

Objective: To communicate complex technical workflows and logical relationships clearly and concisely, demonstrating a deep understanding of NAMs and data integration processes.

Methodology: The following diagrams, created using Graphviz with a specified color palette and contrast rules, illustrate key workflows a researcher might describe in their CV.

Diagram 1: NAMs Data Integration Workflow This diagram visualizes the pathway from experimental data generation to risk assessment, a core competency for scientists in this field.

Diagram 2: CV Data Parsing Logic This diagram outlines the logical process an ATS or reviewer uses to parse a well-structured CV, highlighting the importance of keyword and section optimization.

The Scientist's Toolkit: Essential Research Reagent Solutions

A proficient researcher's CV should reflect familiarity with key tools and platforms. The following table details essential "reagent solutions" for data generation, analysis, and standardization in the field of NAMs and ecotoxicology.

Table 2: Key Research Reagent Solutions for NAMs and Data Standardization

Item Name	Function/Brief Explanation	Application in Research
OECD QSAR Toolbox	Software designed to fill data gaps for chemical safety assessment without additional testing, using read-across and trend analysis.	Essential for computational toxicology; used to group chemicals, profile metabolites, and predict adverse effects [1].
UMLS (Unified Medical Language System)	A set of files and software that brings together many health and biomedical vocabularies and standards to enable interoperability.	Serves as a core controlled vocabulary for standardizing terms related to diseases, findings, and chemicals in ecotoxicity data annotation [1].
BfR DevTox Database	A lexicon providing harmonized terminology specifically for describing prenatal developmental toxicity findings.	Critical for ensuring consistent annotation of developmental endpoints across studies, facilitating data comparison and integration [1].
KNIME/Python/R Platforms	Open-source platforms for data analytics, integration, and the creation of predictive models.	Used to build and execute workflows for data cleaning, statistical analysis, QSAR modeling, and automated data annotation [1].
ECOTOX Database	A comprehensive database providing single-chemical ecological toxicity data for aquatic and terrestrial organisms.	A key resource for curating legacy ecotoxicity data and performing ecological risk assessments as part of a weight-of-evidence approach.

The transition from a static document to a dynamic, semantically structured representation of professional expertise is critical for researchers in the age of NAMs and big data. By adhering to the protocols outlined herein—optimizing for ATS, rigorously applying controlled vocabularies, and clearly visualizing expertise—scientists can create CVs that are not only future-proof but also actively demonstrate their proficiency in the very principles of data standardization and computational analysis that are defining the future of toxicology and drug development. This approach ensures their credentials are both discoverable and meaningful in an increasingly competitive and data-driven research landscape.

Assessing Data Quality and Comparative Frameworks in Ecotoxicology

Application Note: Systematic Review Workflow for Ecotoxicity Data

Ecotoxicity research requires rigorous systematic review methodologies to ensure comprehensive data collection and reliable risk assessments. Central to this process is the effective use of controlled vocabularies—organized sets of standardized phrases used to index database content for consistent information retrieval [46]. This application note synthesizes protocols from the U.S. Environmental Protection Agency (EPA) and international standards to establish a robust framework for identifying, evaluating, and incorporating ecotoxicity evidence.

Key Concepts: Controlled Vocabularies in Ecotoxicity Research

Controlled vocabularies provide critical infrastructure for systematic reviews by bringing uniformity to database indexing. Trained indexers read full-text publications and identify key concepts, which are then translated into standardized terms within the database's vocabulary system [47]. This process creates consistency and precision, enabling researchers to locate relevant studies regardless of the terminology authors used in their publications [46]. Major databases employ different controlled vocabulary systems:

• MEDLINE/PubMed: Medical Subject Headings (MeSH) [48]
• Embase: Emtree subject headings [47]
• CINAHL: CINAHL Subject Headings [46]

These systems help address terminology challenges where the same concept may be described differently across databases, such as "complementary therapies" in MeSH versus "alternative medicine" in Emtree [47].

Experimental Protocols

Protocol 1: Systematic Literature Search Strategy

Objective

To comprehensively identify relevant ecotoxicity studies while minimizing database-specific terminology bias.

Materials

• Access to multiple scientific databases (e.g., MEDLINE, Embase, ECOTOX)
• Reference management software
• Search strategy documentation tool

Procedure

• Concept Mapping: Identify core concepts and potential synonyms for your research question.
• Vocabulary Identification: For each database, identify relevant controlled vocabulary terms for each concept using database thesauri [46].
• Search Construction:
  • Combine controlled vocabulary terms using database-specific explosion features to include narrower terms [47].
  • Supplement with keyword searches using author terminology [46].
  • Utilize Boolean operators to combine concepts.
• Search Execution: Run searches across multiple databases and export results.
• Documentation: Record search strategies, dates, and result counts for each database.

Example Search Strategy

For identifying pesticide toxicity studies in MEDLINE via PubMed:

Protocol 2: EPA ECOTOX Data Screening and Evaluation

Objective

To screen and evaluate ecotoxicity studies from open literature using EPA validation criteria.

Materials

• Access to EPA ECOTOX Knowledgebase [49]
• Study evaluation checklist
• Data extraction forms

Procedure

Phase I: Initial Screening Apply EPA acceptance criteria to determine study relevance [13]:

• Confirm toxic effects relate to single chemical exposure.
• Verify effects on aquatic or terrestrial plants or animals.
• Ensure documented biological effect on live, whole organisms.
• Confirm reported concurrent environmental chemical concentration/dose or application rate.
• Verify explicit exposure duration.
• Assess whether toxicology information is for a chemical of concern.
• Confirm the article is published in English.
• Verify the study is presented as a full article (not abstract only).

Phase II: Quality Assessment Evaluate passing studies using additional EPA criteria [13]:

• Confirm the paper is publicly available.
• Verify the paper is the primary data source (not secondary analysis).
• Check for reported calculated endpoints (e.g., LC50, NOEC).
• Verify treatments compared to acceptable controls.
• Confirm study location (laboratory vs. field) is reported.
• Check that tested species is reported and verified.

Phase III: Data Extraction For accepted studies, extract:

• Test substance characteristics
• Test organism details (species, life stage)
• Exposure conditions (duration, route, medium)
• Endpoints measured
• Results (quantitative and statistical)
• Study limitations

Data Presentation

EPA Ecotoxicity Data Acceptance Criteria

Table 1: EPA Acceptance Criteria for Ecological Toxicity Data from Open Literature [13]

Criterion Category	Specific Requirement	Application Notes
Exposure Conditions	Single chemical exposure	Excludes complex mixtures unless the pesticide formulation itself is evaluated
	Concurrent concentration/dose reported	Must include measured exposure levels, not just application rates
	Explicit exposure duration	Clear temporal component for the exposure scenario
Test System	Aquatic or terrestrial species	Includes plants, animals, and microorganisms
	Biological effect on live, whole organisms	Excludes in vitro or suborganismal studies unless specified
	Tested species reported and verified	Taxonomic identification must be confirmable
Study Design	Comparison to acceptable control	Appropriate control group with identical conditions except for test substance
	Location reported (lab/field)	Critical for interpreting exposure conditions and environmental relevance
Publication Status	English language	English translation acceptable for non-English papers
	Full article publicly available	Conference abstracts, theses, and non-public reports excluded
	Primary data source	Excludes review articles and meta-analyses for data extraction

Standardized Ecotoxicity Test Guidelines

Table 2: Selected EPA Ecological Effects Test Guidelines [50]

Test Category	Guideline Number	Test Name	Key Measurements
Aquatic Fauna	850.1000	Aquatic Invertebrate Acute Toxicity Test	LC50, mortality
	850.1400	Fish Acute Toxicity Test	LC50, behavioral changes
Terrestrial Wildlife	850.2100	Avian Acute Oral Toxicity Test	LD50, mortality
	850.2300	Avian Reproduction Test	Reproduction success, egg viability
Beneficial Insects	850.3020	Honey Bee Acute Contact Toxicity Test	LD50, mortality
	850.3030	Honey Bee Toxicity of Residues on Foliage	Contact toxicity, residual effects
Plants	850.4100	Seedling Emergence and Seedling Growth	Emergence rate, growth parameters
	850.4400	Aquatic Plant Toxicity Test Using Lemna spp.	Growth inhibition, frond production

International Ecotoxicity Standards

Table 3: International Ecotoxicity Testing Standards and Their Applications

Standard Identifier	Title	Scope/Application
ISO 5430:2023 [51]	Plastics — Ecotoxicity testing scheme	Marine organisms across four trophic levels for plastic degradation products
ASTM E2361-13(2021) [52]	Standard Guide for Testing Leave-On Products Using In-Situ Methods	Antimicrobial efficacy testing
ASTM E2180-24 [52]	Standard Test Method for Determining the Activity of Incorporated Antimicrobial Agent(s)	Polymeric or hydrophobic materials with incorporated antimicrobials

Visualization: Systematic Review Workflow

Ecotoxicity Data Screening Workflow

EPA ECOTOX Database Evaluation Process

The Scientist's Toolkit: Research Reagent Solutions

Table 4: Essential Resources for Ecotoxicity Systematic Reviews

Tool/Resource	Type	Function	Access
ECOTOX Knowledgebase [49]	Database	Comprehensive database of single chemical toxicity to ecological species	https://cfpub.epa.gov/ecotox/
EPA Series 850 Guidelines [50]	Test Guidelines	Standardized ecological effects test protocols for regulatory submissions	EPA website
MeSH (Medical Subject Headings) [48]	Controlled Vocabulary	NLM's controlled vocabulary for indexing MEDLINE/PubMed articles	PubMed MeSH Database
Emtree [47]	Controlled Vocabulary	Elsevier's controlled vocabulary for Embase database	Embase platform
SeqAPASS [49]	Computational Tool	Predicts chemical susceptibility across species using protein sequence alignment	EPA website
Web-ICE [49]	Modeling Tool	Estimates acute toxicity to aquatic and terrestrial organisms for risk assessment	EPA website
SSD Toolbox [49]	Statistical Tool	Generates species sensitivity distributions for chemical risk characterization	EPA website
ASTM Environmental Toxicology Standards [52]	Standard Methods	Consensus standards for environmental toxicology testing procedures	ASTM standards store

Implementation Considerations

Integration of Controlled Vocabulary Searching

Effective systematic reviews in ecotoxicity must combine both controlled vocabulary and keyword searching approaches [46]. Since not all articles are immediately assigned controlled vocabulary terms, particularly newer publications, relying solely on subject headings risks missing relevant recent research. A comprehensive search strategy should include both the controlled vocabulary terms (e.g., "Dogs"[MeSH]) and keyword variants (e.g., dog*, canine) to ensure complete coverage of the literature [46].

EPA Data Evaluation Framework Implementation

The EPA's two-phase evaluation approach provides a robust framework for assessing study quality and relevance [13]. Risk assessors should apply best professional judgment when implementing these criteria, as the utility of open literature studies cannot be completely prescribed by guidance documents. Documentation of the evaluation process through Open Literature Review Summaries (OLRS) is essential for transparency and tracking on EPA's Storage Area Network [13].

Addressing Data Gaps with Computational Tools

For chemicals with limited toxicity data, EPA researchers develop ecological models to predict effects on endangered species and wildlife populations [49]. Tools such as SeqAPASS enable cross-species extrapolation of toxicity information, while Web-ICE and the Species Sensitivity Distribution Toolbox help characterize chemical risks based on available data [49]. These computational approaches are particularly valuable for assessing contaminants of immediate and emerging concern, such as PFAS chemicals, where traditional toxicity data may be limited.

Within ecotoxicology and regulatory science, the reliability of individual studies forms the cornerstone of robust hazard and risk assessments. The evaluation of ecotoxicity data ensures that regulatory decisions—from marketing authorizations for plant protection products to assessments under the REACH legislation—are based on sound, verifiable science [53]. For decades, the method established by Klimisch et al. in 1997 has been the predominant tool for this task, categorizing studies as "reliable without restrictions," "reliable with restrictions," "not reliable," or "not assignable" [53]. However, its reliance on expert judgement and limited criteria have raised concerns about consistency and transparency [53].

This landscape has spurred the development of alternative frameworks, including the Schneider method (Toxicological data reliability assessment Tool, or ToxRTool) and the more recent CRED (Criteria for Reporting and Evaluating Ecotoxicity Data) method [53] [54]. The ongoing evolution of these tools occurs within a critical broader context: the push for a standardized controlled vocabulary for ecotoxicity data research. Consistent terminology is not merely an academic exercise; it is essential for ensuring that data is Findable, Accessible, Interoperable, and Reusable (FAIR), thereby enabling computational toxicology, systematic reviews, and the validation of New Approach Methodologies (NAMs) [1] [4]. This article compares these key evaluation frameworks, detailing their application and highlighting their synergy with controlled vocabularies in modern toxicological research.

Detailed Description of Individual Methods

• Klimisch Method: Developed for evaluating both toxicological and ecotoxicological data, this method relies on 12 to 14 evaluation criteria and four categorical outcomes [53] [54]. Its primary strength was providing an initial step toward standardized reliability evaluation. However, it has been criticized for its lack of detailed guidance and for favoring Good Laboratory Practice (GLP) and standardized guideline studies, potentially leading to the automatic categorization of such studies as reliable even when specific flaws exist [53]. Its minimal guidance often results in evaluations that are heavily dependent on expert judgment, causing inconsistencies among assessors [53].

• Schneider Method (ToxRTool): This framework, known as the Toxicological data reliability assessment Tool, assesses toxicity data from in vivo and in vitro studies [54]. It employs 21 evaluation criteria, which include both recommended and mandatory questions, and requires scorings between 0 and 1 [54]. A key feature is the provision of additional guidance to the evaluator and a defined process for summarizing the evaluation, which is calculated automatically [54]. Compared to the Klimisch method, it matches the same number of OECD reporting criteria (14 out of 37) but offers a more structured and less subjective evaluation process [54].

• CRED Method: Developed specifically to address the shortcomings of the Klimisch method for aquatic ecotoxicity studies, the CRED method offers a significantly more detailed framework [53]. It evaluates 20 reliability criteria and, crucially, introduces 13 relevance criteria, ensuring a study's appropriateness for a specific hazard identification or risk characterization is assessed [53]. A ring test involving 75 risk assessors from 12 countries found that the CRED method was perceived as less dependent on expert judgement, more accurate and consistent, and practical in terms of time and criteria use compared to the Klimisch method [53]. It is considered a suitable replacement for the Klimisch method [53].

Comparative Analysis of Frameworks

Table 1: Comparative overview of reliability evaluation methods for toxicological and ecotoxicological data.

Characteristic	Klimisch et al.	Schneider et al. (ToxRTool)	CRED
Data Types	Toxicity (in vivo, in vitro) and ecotoxicity (acute, chronic) [54]	Toxicity data (in vivo, in vitro) [54]	Aquatic ecotoxicity [53]
Primary Coverage	Reliability [53]	Reliability and a few aspects of relevance [54]	Reliability and Relevance [53]
Evaluation Categories	Reliable without restrictions, reliable with restrictions, not reliable, not assignable [53] [54]	Reliable without restrictions, reliable with restrictions, not reliable, not assignable [54]	Qualitative evaluation of reliability and relevance [53]
Number of Criteria	12 (acute ecotoxicity), 14 (chronic ecotoxicity) [54]	21 [54]	20 reliability criteria, 13 relevance criteria [53]
Additional Guidance	No [54]	Yes [54]	Yes [53]
Alignment with OECD Criteria	14 out of 37 criteria [54]	14 out of 37 criteria [54]	37 out of 37 criteria [53]

The table above highlights the evolution from the broad but shallow Klimisch method toward more specialized and guided frameworks. The CRED method represents the most comprehensive option for aquatic ecotoxicity, fully incorporating OECD reporting standards and formally integrating relevance evaluation. The Schneider method offers a structured, scored approach for a broader range of toxicity data. The choice of method can directly impact the outcome of a hazard or risk assessment, influencing which studies are included in a dataset and potentially leading to unnecessary risk mitigation measures or underestimated environmental risks [53].

Application Notes: Implementing Evaluation Frameworks

Step-by-Step Experimental Protocols

Implementing a robust reliability evaluation requires a systematic, step-by-step approach. The following protocol is synthesized from best practices across the evaluated methods, with particular emphasis on the detailed procedures of the CRED method and the data curation pipeline of the ECOTOX knowledgebase [53] [4].

Protocol 1: Reliability and Relevance Evaluation of an Ecotoxicity Study

• Study Identification and Triage:

  • Identify the study to be evaluated, whether from the peer-reviewed literature or a regulatory dossier.
  • Determine the study's basic attributes: test substance, test organism, endpoints measured, and duration. This facilitates subsequent evaluation.

• Systematic Data Extraction:

  • Extract key methodological details using a predefined template. The ECOTOX knowledgebase exemplifies this with well-established controlled vocabularies to ensure consistency [4].
  • Key data points include: test substance characterization (e.g., purity, formulation), test organism (species, life stage, source), exposure conditions (duration, media, renewal), experimental design (controls, replicates, randomization), and measured endpoints with results.

• Reliability Assessment:

  • Use a selected evaluation method (e.g., CRED, ToxRTool) as a checklist. For each criterion (e.g., "test substance specifications are described," "control mortality is within acceptable limits"), determine if it is fulfilled.
  • The CRED method provides detailed guidance for assessing its 20 reliability criteria, reducing subjectivity [53].
  • Document the rationale for each decision, noting any deviations from guideline procedures or GLP.

• Relevance Assessment:

  • Evaluate the study's relevance for the specific assessment purpose using the 13 CRED criteria or similar [53].
  • Consider the appropriateness of the test organism, exposure pathway, measured endpoints (apical vs. sub-organismal), and the environmental relevance of the exposure concentrations.

• Final Categorization and Documentation:

  • Synthesize the findings from the reliability and relevance evaluations to assign an overall categorization (e.g., "reliable without restrictions," "reliable with restrictions").
  • Prepare a final report that transparently documents the evaluation process, including all supporting judgments and any limitations identified. This ensures the assessment is verifiable and reproducible.

Protocol 2: Data Curation and Vocabulary Standardization for Database Inclusion

This protocol, derived from recent work on standardizing developmental toxicology data, is essential for preparing evaluated data for computational use [1] [55].

• Primary Data Extraction:

  • Extract treatment-related endpoints and study parameters from the source document, recording them using the original authors' language.

• Application of Controlled Vocabulary:

  • Map the extracted free-text terms to standardized terms from established controlled vocabularies. Key resources include:
    • Unified Medical Language System (UMLS): A comprehensive set of biomedical terms [1].
    • OECD Harmonised Templates: Standardized terms for reporting chemical test data [1].
    • BfR DevTox Project: A harmonized lexicon for developmental toxicology data [1].
  • Use a pre-defined crosswalk (a mapping between vocabularies) to automate this process where possible. One study achieved automated standardization for 57-75% of extracted endpoints, saving over 350 hours of manual effort [1].

• Manual Review and Curation:

  • Manually review automated mappings for accuracy, particularly for complex or ambiguous terms.
  • Manually map terms that could not be standardized automatically, often because they are too general or require human logic.

• Data Integration and FAIRification:

  • Integrate the standardized data into a structured database (e.g., ToxValDB, ECOTOX) [4] [5].
  • Ensure the final dataset adheres to FAIR principles by providing rich metadata and persistent identifiers.

Table 2: Key databases, tools, and controlled vocabularies for ecotoxicity data evaluation and curation.

Item Name	Type	Function & Application
ECOTOX Knowledgebase [4]	Database	A comprehensive, curated database of single-chemical ecotoxicity data for aquatic and terrestrial species. Used to locate existing effects data and as a model for systematic curation.
ToxValDB [5]	Database	A compiled resource of experimental and derived human health-relevant toxicity data. Provides summary-level data in a standardized format for comparison and modeling.
CRED Evaluation Method [53]	Guideline	Provides detailed criteria and guidance for evaluating the reliability and relevance of aquatic ecotoxicity studies. Used to ensure consistency and transparency in regulatory assessments.
ToxRTool [54]	Tool	A standardized tool for evaluating the reliability of toxicological data (in vivo and in vitro). Uses a scored questionnaire to reduce subjectivity.
Controlled Vocabulary Crosswalk [1]	Tool	A mapping file (e.g., between UMLS, OECD, BfR DevTox terms) that enables the automated standardization of extracted endpoint data, enhancing interoperability.
OECD Harmonised Templates [1]	Vocabulary	Standardized terms and reporting formats for chemical test data. Used to ensure consistent data extraction and reporting across studies.

Integration with Controlled Vocabulary Systems

The quantitative and qualitative data generated through the evaluation frameworks above realize their full potential only when integrated with structured vocabulary systems. This integration is the linchpin for achieving interoperability and reusability in modern data-driven research.

The workflow from primary study to a FAIR (Findable, Accessible, Interoperable, and Reusable) dataset critically depends on this integration. Evaluated studies, whether categorized via Klimisch, CRED, or another method, have their key data extracted. This extracted data, often in free-text form, is then mapped to terms from controlled vocabularies like those from the OECD or the UMLS [1]. This process of standardization transforms subjective narrative descriptions into structured, computable data. For example, terms like "reduced pup weight," "lower fetal body weight," and "decreased offspring mass" can all be mapped to a single standardized term such as "fetal body weight decrease" [1]. This resolves ambiguity and allows for the aggregation and comparison of data across thousands of studies.

This practice is central to the operation of major toxicological databases. The ECOTOX knowledgebase curates data using controlled vocabularies, which supports its role in environmental research and risk assessment [4]. Similarly, ToxValDB employs a rigorous two-phase process where data is first curated in its original format and then standardized onto a common structure and vocabulary, enabling meta-analyses and serving as an index for public toxicology data [5]. The adoption of these vocabularies directly supports the development and validation of New Approach Methodologies (NAMs) by providing high-quality, structured reference datasets for benchmarking [4] [5].

The following diagram illustrates the logical workflow from primary study to a FAIR-compliant dataset, highlighting the roles of evaluation frameworks and vocabulary standardization.

The evolution from the Klimisch method to more sophisticated frameworks like CRED and ToxRTool marks a significant advancement in ecotoxicology and regulatory science. This transition is characterized by a move toward greater transparency, reduced subjectivity, and the formal incorporation of relevance alongside reliability. The comparative analysis and application protocols provided herein offer researchers a practical guide for implementing these critical evaluations.

Ultimately, the rigor of a single study evaluation is amplified when its data can be seamlessly integrated with other evidence. The synergistic relationship between robust evaluation frameworks and controlled vocabulary systems is what truly powers the future of toxicological research. By transforming evaluated studies into structured, standardized, and FAIR data, we enable more efficient and credible chemical assessments, inform the development of predictive models, and accelerate the adoption of New Approach Methodologies. This integrated approach is indispensable for meeting the demanding challenge of ensuring the safety of thousands of chemicals in commerce.

The Role of CVs in Supporting Chemical Alternatives Assessment and Green Chemistry

Controlled Vocabularies (CVs) serve as the foundational framework for standardizing ecotoxicity data, enabling interoperability, machine-readability, and advanced computational analysis in green chemistry and alternatives assessment. Within chemical research and regulation, inconsistent terminology for species, endpoints, and experimental conditions creates significant barriers to data integration, model development, and the reliable identification of safer chemical alternatives. CVs systematically address this challenge by providing standardized, structured terminology that tags data consistently across diverse sources [7]. This harmonization is critical for building robust datasets that power Quantitative Structure-Activity Relationship (QSAR) models, machine learning (ML) algorithms, and New Approach Methodologies (NAMs) aimed at reducing animal testing and guiding the design of benign chemicals [56] [57]. This document details practical protocols for implementing CVs and demonstrates their application through specific computational workflows for chemical alternatives assessment.

Key Concepts and Definitions

The Critical Role of Standardized Data in Ecotoxicology

The advancement of computational toxicology is heavily dependent on the quality and consistency of underlying data. Controlled Vocabularies (CVs) are curated, predefined lists of standard terms used to tag and categorize data, ensuring that all contributors describe the same concept, organism, or experimental condition using identical terminology. In ecotoxicology, this is paramount because models trained on heterogeneous data can produce unreliable predictions. For instance, the same lethal effect might be labeled as MOR, mortality, or lethality across different datasets, complicating data aggregation [7]. CVs remediate this by enforcing a single term, such as Effect: Mortality.

The synergy between CVs and computational approaches is a cornerstone of modern green chemistry. Computational toxicology employs in silico methods to predict the toxicity of chemicals, leveraging mathematical models and computer simulations [57]. These methods include:

• Quantitative Structure-Activity Relationship (QSAR): Models that establish a relationship between a chemical's molecular structure or properties (descriptors) and its biological activity [58] [57].
• Machine Learning (ML) and Deep Learning (DL): Advanced statistical techniques that learn from existing data to make predictions on new chemicals, often handling more complex, non-linear relationships than classical QSAR [57].
• New Approach Methodologies (NAMs): A broad suite of innovative tools, including in vitro assays and in silico models, intended to provide faster, more cost-effective, and human-relevant safety assessments while reducing animal testing [59] [56].

These methodologies are integral to the paradigm of "benign by design," a core principle of green chemistry where computational tools are used proactively to design chemicals and processes that are inherently low-hazard [60].

Core Controlled Vocabularies for Ecotoxicity Data

The following table summarizes essential CVs and identifiers required for structuring ecotoxicity data.

Table 1: Essential Controlled Vocabularies and Identifiers for Ecotoxicity Data

Vocabulary Category	Purpose	Standard Terms / Format	Example
Chemical Identifiers	Uniquely and unambiguously identify a chemical substance.	CAS RN, DTXSID, InChIKey, SMILES	InChIKey=BBQQJQOVCUSFPO-UHFFFAOYSA-N (for caffeine)
Taxonomic Classification	Standardize organism species using a hierarchical biological classification.	Kingdom; Phylum; Class; Order; Family; Genus; Species	Animalia; Chordata; Actinopterygii; Cyprinodontiformes; Poeciliidae; Poecilia; reticulata
Ecotox Group (CV)	Categorize test species into broad, ecologically relevant taxonomic groups.	Fish, Crustacean, Algae	Crustacean
Effect (CV)	Describe the observed biological response to chemical exposure.	Mortality (MOR), Immobilization (ITX), Growth (GRO), Population (POP)	Immobilization (ITX)
Endpoint (CV)	Define the measured quantitative value resulting from a test.	LC50, EC50, NOEC	EC50
Duration & Units	Standardize exposure time and concentration units.	h, d; mg/L, µg/L, mol/L	48 h, mg/L

Protocol: Implementing CVs for Curating a Model-Ready Ecotoxicity Dataset

This protocol provides a step-by-step methodology for curating a high-quality, computational-ready dataset from raw ecotoxicity sources (e.g., the US EPA ECOTOX database) [7]. The primary objective is to transform heterogeneous data into a structured, machine-readable format using CVs, enabling its direct use in QSAR and ML modeling for chemical alternatives assessment.

Materials and Reagent Solutions

Table 2: Essential Computational Tools for Data Curation and Modeling

Tool Name	Type	Primary Function in Protocol
KNIME [61] [57]	Data Analytics Platform	Visual workflow for data integration, curation, and transformation.
US EPA ECOTOX Database [7]	Data Repository	Source of raw ecotoxicity test results.
EPA CompTox Chemicals Dashboard [61] [7]	Chemistry Database	Source of curated chemical structures and identifiers (DTXSID, SMILES).
PubChem [61] [7]	Chemistry Database	Source of chemical structures and canonical SMILES.
RDKit [57]	Cheminformatics Library	Calculation of molecular descriptors and fingerprints within a programming environment.

Step-by-Step Procedure

• Data Acquisition and Initial Filtering:

  • Download the ECOTOX database in its pipe-delimited ASCII format.
  • Load the species, tests, and results tables into a data processing environment like KNIME or Python.
  • Filter the data to retain only entries for the three core taxonomic groups by setting the ecotox_group CV to Fish, Crustacean, or Algae [7].

• Chemical Identifier Curation and Standardization:

  • Extract all available chemical identifiers (CAS RN, DTXSID) from the source data.
  • Use a reliable resolver, such as the EPA CompTox Dashboard API or the Chemical Identifier Resolver (CIR) in KNIME, to obtain standard InChIKeys and canonical SMILES for each unique substance [61]. The InChIKey is preferred for deduplication as it is a unique, hash-based identifier.
  • Critical Step: Resolve any identifier conflicts and remove entries for which a valid structure cannot be obtained.

• Application of Effect and Endpoint CVs:

  • Map the raw effect descriptions from the source data to the standardized CV terms. For example:
    • Map "death", "lethality" to "MOR".
    • Map "immobilisation", "intoxication" to "ITX" [7].
  • Similarly, standardize endpoint names to LC50, EC50, etc.
  • Filter the dataset to include only the relevant effect-endpoint combinations (e.g., for acute toxicity, retain MOR-LC50 and ITX-EC50).

• Experimental Condition Standardization:

  • Standardize the duration field to a common unit (hours).
  • Convert all concentration values to a standardized unit, preferably molar (mol/L) for modeling, as it is more biologically informative [7].
  • Record the media of exposure (e.g., freshwater, marine) using a CV if available.

• Final Dataset Assembly and Validation:

  • Merge the curated chemical, species, effect, and endpoint data into a single, integrated table.
  • Perform final quality checks: remove duplicates, check for outliers in toxicity values, and ensure there are no missing critical values (SMILES, endpoint value, duration).
  • The final output is a structured dataset where each row represents a unique test result, tagged entirely with CVs, ready for feature generation and model training.

The following workflow diagram visualizes this multi-stage curation process.

Figure 1: CV-Driven Data Curation Workflow. This process transforms raw data into a structured, model-ready format.

Application Note: Utilizing CV-Standardized Data in an Alternatives Assessment Workflow

Background and Objective

This application note demonstrates how CV-standardized data is utilized in a computational workflow to assess the aquatic toxicity of a novel chemical (Chemical X) relative to a known hazardous compound, supporting a green chemistry alternatives assessment. The objective is to predict the acute toxicity of Chemical X for fish, crustaceans, and algae and compare it to the benchmark chemical.

Computational Workflow and Methods

• Data Retrieval and Model Training:

  • A CV-curated dataset, prepared per the protocol in Section 3, serves as the training data. Its standardization ensures reliable model development.
  • Molecular descriptors (e.g., topological, electronic) are calculated from the canonical SMILES strings using a tool like RDKit [57].
  • A machine learning model (e.g., Random Forest or a Deep Neural Network) is trained to predict the standardized endpoint LC50/EC50 based on the molecular descriptors [57].

• Toxicity Prediction and Mechanistic Insight:

  • The SMILES of Chemical X is input into the trained model, which outputs predicted LC50/EC50 values for the three taxonomic groups.
  • The Adverse Outcome Pathway (AOP) framework is used to contextualize predictions. An AOP is a structured representation of a sequence of events from a molecular initiating event (MIE) to an adverse outcome at the organism level [57]. CVs are critical for populating AOPs with consistent key event terminology.
  • In vitro to in vivo extrapolation (IVIVE) using physiologically based kinetic (PBK) models can be applied to relate bioactivity concentrations from assays to equivalent external exposure levels, strengthening the qualitative linkage between different data types [59].

The following diagram illustrates the integrated predictive workflow.

Figure 2: Predictive Toxicology Workflow for Alternatives Assessment. CV-curated data enables reliable ML model training for toxicity prediction.

Data Presentation and Interpretation

The results of the prediction and comparison are summarized in the table below.

Table 3: Predicted Acute Aquatic Toxicity for Alternative Assessment

Chemical	Taxonomic Group	Predicted LC50/EC50 (mg/L)	Confidence Score	GHS Category (Predicted)
Benchmark Chemical	Fish	0.5	High	Acute Toxicity 1
	Crustacean	0.8	High	Acute Toxicity 1
	Algae	1.2	Medium	Acute Toxicity 1
Chemical X	Fish	25.0	Medium	Acute Toxicity 2
	Crustacean	40.5	Medium	Acute Toxicity 3
	Algae	>100	Low	Not Classified

Interpretation: The data indicates that Chemical X is significantly less toxic than the Benchmark Chemical across all three taxonomic groups. Based on this computational assessment, Chemical X presents a potentially safer alternative. The confidence scores, which can be derived from the applicability domain of the QSAR/ML model, inform the user of the reliability of each prediction and highlight where further testing with NAMs might be prioritized [56].

Controlled Vocabularies are not merely an administrative data management tool; they are a critical enabler of modern, computational-driven green chemistry. By providing the necessary structure and consistency to ecotoxicity data, CVs unlock the potential of advanced in silico methodologies, from QSAR and machine learning to AOP development. The protocols and applications detailed herein provide researchers with a practical roadmap for leveraging CVs to design safer chemicals, conduct robust alternatives assessments, and ultimately support the transition toward a more sustainable and predictive toxicology paradigm.

Within ecotoxicity research, the challenge of integrating disparate data streams into a coherent risk assessment is a significant hurdle. The need for transparent, data-driven prioritization is paramount for directing resources toward the most pressing environmental hazards. Framing these assessments within a controlled vocabulary ensures consistency, reproducibility, and clear communication across interdisciplinary teams. The Toxicological Prioritization Index (ToxPi) framework emerges as a powerful solution, offering a standardized approach for integrating and visualizing diverse lines of evidence to support decision-making [62]. This protocol outlines the application of ToxPi, detailing its operation within a research context focused on comparative hazard assessment, and aligns its methodology with the principles of a structured data ontology.

The ToxPi framework transforms complex, multi-source data into an integrated visual and numerical ranking. It functions by combining diverse data sources—such as in vitro assay results, chemical properties, and exposure estimates—into a unified profile that facilitates the direct comparison of chemicals or other entities [62] [63]. Each data type is organized into a "slice" of the ToxPi pie, and the collective array of slices provides a transparent, weighted, and visual summary of the contributing factors to an overall hazard score [64].

The core output is a ToxPi profile, a variant of a polar diagram or radar chart, where the radial length of each slice represents its relative contribution to the overall score [64]. A fundamental principle is that a larger slice signifies a greater contribution to the measured effect (e.g., higher hazard or vulnerability), and a more filled-in area of the overall profile indicates a higher cumulative score [64]. This intuitive visual design allows for the rapid identification of key drivers of risk. The framework is supported by multiple software implementations, including a stand-alone Java Graphical User Interface (GUI), the toxpiR R package, and the ToxPi*GIS Toolkit for geospatial integration [62].

Table 1: ToxPi Software Distributions and Their Primary Uses

Software Distribution	Primary Function	Key Features	Ideal Use Case
ToxPi GUI [62]	Interactive visual analytics	User-friendly interface, dynamic exploration, bootstrap confidence intervals [63]	Desktop-based chemical prioritization and hypothesis generation
toxpiR R Package [62]	Programmatic analysis	Scriptable, integrates with R-based workflows, enables advanced statistical analysis	High-throughput or reproducible pipeline integration
ToxPi*GIS Toolkit [62] [64]	Geospatial visualization	Integrates ToxPi profiles with ArcGIS mapping, creates interactive web maps	Community-level vulnerability assessments and spatial risk mapping

Application Notes and Protocols

Protocol: Formulating a ToxPi Model for Ecotoxicity Hazard Ranking

This protocol describes the steps to create a ToxPi model for ranking the comparative ocular hazard of environmental contaminants using zebrafish data.

Research Reagent Solutions and Materials

Table 2: Essential Research Reagents and Materials for Zebrafish-Based Ocular Toxicity Testing

Item Name	Function/Description	Relevance to ToxPi Framework
Zebrafish (Danio rerio) [65]	A model organism with high genetic and anatomical similarity to humans, particularly in ocular structure.	Provides the in vivo data streams (behavioral, morphological) that feed into ToxPi slices.
Contrast-Optomotor Response (C-OMR) Assay [66]	A high-sensitivity behavioral test using graded contrast gray-white stripes to quantify visual function in zebrafish larvae.	Serves as a key functional endpoint; data from this assay populates a "Visual Function" slice in the ToxPi model.
Optical Coherence Tomography (OCT) [65]	A non-invasive interference technique for high-resolution retinal imaging.	Provides structural endpoint data for a "Retinal Morphology" slice in the ToxPi model.
Environmental Contaminants (e.g., EDCs, BFRs, heavy metals) [65]	Test articles used to induce ocular toxicity for model development.	The entities being ranked and compared by the ToxPi model.
ToxPi GUI Software [62]	The platform for integrating, modeling, and visualizing the data.	The analytical engine that transforms raw data into integrated hazard rankings and profiles.

Step-by-Step Methodology

• Data Acquisition and Curation: Collect data from relevant sources. For an ocular toxicity model, this would include:

  • Biochemical Data: Expression levels of key visual proteins (e.g., opsins) from molecular assays.
  • Histopathological Data: Retinal layer thickness measurements from OCT imaging [65].
  • Functional Behavioral Data: Response rates from the C-OMR assay, which provides a more sensitive measure of visual impairment than traditional OMR tests [66].
  • Exposure Data: Chemical properties and administered dose concentrations.
  • Organize all data into a tabular format, with rows representing each chemical or test condition and columns representing the individual data streams.

• Data Scaling and Normalization: Within the ToxPi GUI, transform all raw data values for each data stream to a consistent 0-1 scale, where 0 represents the minimum observed value and 1 represents the maximum [64]. For endpoints where a lower value indicates higher hazard (e.g., reduced response in C-OMR), instruct the software to invert the scale.

• Slice Formulation and Weighting: Group related data streams into conceptual slices. For our example:

  • Create a "Visual Function" slice containing the C-OMR data.
  • Create a "Retinal Integrity" slice containing OCT measurements and opsin expression data.
  • Assign a relative weight to each slice based on its deemed importance to the overall hazard assessment. Weights are flexible and can be adjusted through a semi-automated, guided optimization process [62].

• Model Execution and Visualization: Run the ToxPi model. The GUI will generate a sortable list of all tested chemicals alongside their circular ToxPi profiles [63]. The overall ToxPi score is a normalized composite of the weighted slice scores.

• Validation and Sensitivity Analysis: Utilize the built-in bootstrap resampling feature to calculate 95% confidence intervals for both the overall scores and the relative ranks. This assesses the stability and reliability of the prioritization [63].

Diagram 1: ToxPi model workflow for hazard ranking.

Protocol: Geospatial Hazard Mapping with ToxPi*GIS

This protocol extends the ToxPi framework for geographic visualization, ideal for identifying community-level environmental health vulnerabilities.

• Develop a Base ToxPi Model: First, create and finalize a ToxPi model using the GUI or toxpiR, ensuring each data record is linked to a specific geographic identifier (e.g., county FIPS code, census tract) [64].

• Data Preprocessing for GIS: Prepare the geographic boundary files (e.g., shapefiles) that correspond to the locations in your ToxPi model.

• Run the ToxPi*GIS Toolkit: Use the custom ArcGIS Toolbox (ToxPiToolbox.tbx) or the provided Python script (ToxPi_creation.py) within ArcGIS Pro. This tool consumes the ToxPi results and the spatial data to create a new feature layer [64].

• Generate the Interactive Map: The output is a map with ToxPi profiles drawn at their respective geographic locations. This layer can be styled and combined with other base maps or data layers in ArcGIS Pro [64].

• Share and Disseminate: Publish the map as a Web Map or Web Mapping Application to ArcGIS Online. This creates a public URL, allowing stakeholders without ArcGIS software to interact with the visualization, exploring the drivers of local hazard scores [64].

Diagram 2: Workflow for creating and sharing geospatial ToxPi maps.

Integration with a Controlled Vocabulary for Ecotoxicity

Integrating the ToxPi framework into a broader controlled vocabulary for ecotoxicity research standardizes the interpretation and communication of complex hazard data. The "Data Hazards" project provides a relevant model, offering an open-source vocabulary of ethical concerns—presented as hazard labels—to improve interdisciplinary communication about the potential for downstream harms from data-intensive technologies [67]. Aligning ToxPi outputs with such a vocabulary ensures that the factors driving a high hazard score (e.g., "High Environmental Persistence," "Evidence of Ocular Toxicity") are consistently named and understood across studies and institutions.

This integration creates a robust bridge between quantitative data integration and qualitative risk communication. For instance, a ToxPi slice integrating C-OMR and retinal histology data could be formally tagged with a controlled term like "Visual System Impairment". This allows the computational output of ToxPi to be seamlessly linked with broader safety assessment frameworks and regulatory guidelines, such as the ICH M7 guideline for pharmaceutical impurities, which relies on structured protocols for hazard assessment [68]. By mapping ToxPi components to a controlled ontology, researchers can more efficiently aggregate evidence, perform meta-analyses, and communicate findings with reduced ambiguity, thereby enhancing the reliability and translational impact of ecotoxicity research.

Conclusion

Controlled vocabularies are far more than a technical convenience; they are the fundamental infrastructure that enables the reliability, transparency, and reusability of ecotoxicity data in biomedical and environmental research. By providing a standardized language, CVs directly support critical tasks such as systematic review, ecological risk assessment, and the development of predictive models like QSARs. For drug development professionals, robust CVs ensure that environmental impact assessments are based on sound, comparable data, facilitating regulatory compliance and the design of safer chemicals. The future will see CVs evolve to integrate high-throughput in vitro data and support adverse outcome pathways, further bridging the gap between traditional ecotoxicology and modern computational toxicology. Embracing and contributing to these standardized systems is essential for advancing both scientific understanding and environmental protection.

References

• 1. Standardizing Extracted Data Using Automated Application of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10863721/]
• 2. SKOS Simple Knowledge Organization System - home page [https://www.w3.org/2004/02/skos/]
• 3. Term based comparison metrics for controlled and ... [https://informationr.net/ir/14-1/paper395.html]
• 4. A Curated Database of Ecologically Relevant Toxicity Tests to ... [https://academic.oup.com/etc/article/41/6/1520/7730623]
• 5. Development of the toxicity values database, ToxValDB [https://www.sciencedirect.com/science/article/abs/pii/S2468111325000258]
• 6. Downloadable Computational Toxicology Data [https://www.epa.gov/comptox-tools/downloadable-computational-toxicology-data]
• 7. A benchmark dataset for machine learning in ecotoxicology [https://www.nature.com/articles/s41597-023-02612-2]
• 8. dataFishing: An efficient Python tool and user-friendly web ... [https://www.sciencedirect.com/science/article/pii/S1574954124005120]
• 9. Ecotoxicology (ECOTOX) Knowledgebase Resource Hub [https://www.epa.gov/comptox-tools/ecotoxicology-ecotox-knowledgebase-resource-hub]
• 10. The ECOTOXicology Knowledgebase: A Curated Database of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9408435/]
• 11. A Curated Database of Ecologically Relevant Toxicity Tests to ... [https://academic.oup.com/etc/article-abstract/41/6/1520/7730623]
• 12. Computational toxicology in drug discovery: applications of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12499773/]
• 13. Evaluation Guidelines for Ecological Toxicity Data in the ... [https://www.epa.gov/pesticide-science-and-assessing-pesticide-risks/evaluation-guidelines-ecological-toxicity-data-open]
• 14. High time to Update Statistical Guidance in Ecotoxicology ... [https://pubmed.ncbi.nlm.nih.gov/40824079/]
• 15. An automated computational data pipeline to rapidly ... [https://pubmed.ncbi.nlm.nih.gov/38752675/]
• 16. Five steps to conducting a systematic review - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC539417/]
• 17. What are the Steps of a Systematic Review? [https://hslguides.osu.edu/systematic_reviews/steps]
• 18. Literature search for research planning and identification of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5037943/]
• 19. Choosing Search Terms - Expert Searching [https://browse.welch.jhmi.edu/searching/controlled-vs-keywords]
• 20. Data Curation - Research Data Management [https://researchguides.njit.edu/data-management/recom-practices/data-curation]
• 21. Best Practices for Data Curation and Publication [https://designsafe-ci.org/user-guide/curating/bestpractices/]
• 22. Controlled Vocabulary - an overview | ScienceDirect Topics [https://www.sciencedirect.com/topics/nursing-and-health-professions/controlled-vocabulary]
• 23. 252.403. Essential quality control requirements—toxicity ... [https://www.pacodeandbulletin.gov/Display/pacode?file=/secure/pacode/data/025/chapter252/s252.403.html&d=reduce]
• 24. Proposing Effective Ecotoxicity Test Species for Chemical ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10819827/]
• 25. Assessment of Ecotoxicity - A Framework to Guide ... - NCBI [https://www.ncbi.nlm.nih.gov/books/NBK253975/]
• 26. (Eco)toxicological tests for assessing impacts of chemical ... [https://www.sciencedirect.com/science/article/pii/S0048969721038481]
• 27. CRED: Criteria for reporting and evaluating ecotoxicity data [https://academic.oup.com/etc/article/35/5/1297/7742288]
• 28. Controlled Vocabularies and SKOS - DARIAH-Campus [https://campus.dariah.eu/resources/hosted/controlled-vocabularies-and-skos]
• 29. Introduction to SKOS [https://www.w3.org/2004/02/skos/intro]
• 30. Simple Knowledge Organization System [https://en.wikipedia.org/wiki/Simple_Knowledge_Organization_System]
• 31. SKOS Simple Knowledge Organization System Reference [https://www.w3.org/TR/skos-reference/]
• 32. SKOS Core Vocabulary Specification [https://www.w3.org/TR/swbp-skos-core-spec/]
• 33. Simple Knowledge Organization System (SKOS) (IEKO) [https://www.isko.org/cyclo/skos]
• 34. 24 Types of Charts And Graphs For Data Visualization [https://www.thoughtspot.com/data-trends/data-visualization/types-of-charts-graphs]
• 35. Node Shapes [https://graphviz.org/doc/info/shapes.html]
• 36. Controlled vocabulary [https://en.wikipedia.org/wiki/Controlled_vocabulary]
• 37. Controlled Vocabulary [https://www.librarianshipstudies.com/2020/03/controlled-vocabulary.html]
• 38. Glossary of key terms associated with environmental ... [https://www.ciimar.up.pt/Arcopol_2016/page_22.htm]
• 39. USGS Thesaurus [https://apps.usgs.gov/thesaurus/]
• 40. An integrated ecotoxicological study reliability framework ... [https://pubmed.ncbi.nlm.nih.gov/39873747/]
• 41. Overview of Color Contrast and Its Role in Web Accessibility [https://aeldata.com/role-of-color-contrast-in-web-accessibility/]
• 42. How to Write a CV That Gets You Noticed in 2025 [https://www.morganmckinley.com/article/how-create-cv-passes-ai-screening-and-impresses-recruiters-in-2025]
• 43. 5 Resume Formats That Will Dominate 2025 (And 3 ... [https://blog.theinterviewguys.com/5-resume-formats-that-will-dominate/]
• 44. How to Write a CV That Gets You Hired in 2025 [https://www.linkedin.com/pulse/how-write-cv-gets-you-hired-2025-esther-anagu-mba-tij3f]
• 45. Text has enhanced contrast | ACT Rule | WAI [https://www.w3.org/WAI/standards-guidelines/act/rules/09o5cg/]
• 46. Controlled vocabularies - Systematic Review [https://uow.libguides.com/systematic-review/controlledvocabularies]
• 47. Keyword vs. Subject Searching - Systematic Reviews [https://guides.library.ucdavis.edu/systematic-reviews/controlled-vocabularies]
• 48. LibGuides: Systematic Reviews: Controlled Vocabularies [https://researchguides.library.tufts.edu/c.php?g=249130&p=1659926]
• 49. Ecotoxicological Assessment & Modeling | US EPA [https://www.epa.gov/chemical-research/ecotoxicological-assessment-modeling]
• 50. Series 850 - Ecological Effects Test Guidelines [https://www.epa.gov/test-guidelines-pesticides-and-toxic-substances/series-850-ecological-effects-test-guidelines]
• 51. ISO 5430:2023 - Plastics — Ecotoxicity testing scheme for ... [https://www.iso.org/standard/81239.html]
• 52. Environmental Toxicology Standards - Standards Products [https://store.astm.org/products-services/standards-and-publications/standards/environmental-toxicology-standards.html]
• 53. Criteria for Reporting and Evaluating ecotoxicity Data ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5044958/]
• 54. Comparison of four different methods for reliability evaluation of ... [https://enveurope.springeropen.com/articles/10.1186/2190-4715-23-17/tables/12]
• 55. Standardizing Extracted Data Using Automated Application of ... [https://pubmed.ncbi.nlm.nih.gov/38349723/]
• 56. The rise of computational toxicology in chemical safety ... [https://www.sciencedirect.com/science/article/abs/pii/S2468111324000057]
• 57. Overview of Computational Toxicology Methods Applied in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11677645/]
• 58. Exploration of Computational Approaches to Predict the ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6468900/]
• 59. Computational Tools Applications [https://ntp.niehs.nih.gov/iccvamreport/2023/confidence/comp-tools-app]
• 60. Green chemistry and computational chemistry: A wealth of ... [https://www.sciencedirect.com/science/article/abs/pii/S2352554123001857]
• 61. Computational Tools to Facilitate Early Warning of New ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11511557/]
• 62. ToxPi - The Toxicological Prioritization Index | Official site for ... [https://toxpi.github.io/]
• 63. ToxPi GUI: an interactive visualization tool for transparent ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3988461/]
• 64. ToxPi*GIS Toolkit: creating, viewing, and sharing ... [https://www.nature.com/articles/s41370-022-00433-w]
• 65. Rapid Assessment of Ocular Toxicity from Environmental ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10459940/]
• 66. Development of an experimental model for ocular toxicity ... [https://www.sciencedirect.com/science/article/pii/S0006291X21007208]
• 67. Data Hazards: An open-source vocabulary of ethical ... [https://www.sciencedirect.com/science/article/pii/S266665962500006X]
• 68. Implementation of in silico toxicology protocols within a ... [https://www.sciencedirect.com/science/article/pii/S2468111321000475]