Unlocking the Potential of Clustering and Classification Approaches: Navigating Supervised and Unsupervised Chemical Similarity.

BACKGROUND: The field of toxicology has witnessed substantial advancements in recent years, particularly with the adoption of new approach methodologies (NAMs) to understand and predict chemical toxicity. Class-based methods such as clustering and classification are key to NAMs development and application, aiding the understanding of hazard and risk concerns associated with groups of chemicals without additional laboratory work. Advances in computational chemistry, data generation and availability, and machine learning algorithms represent important opportunities for continued improvement of these techniques to optimize their utility for specific regulatory and research purposes. However, due to their intricacy, deep understanding and careful selection are imperative to align the adequate methods with their intended applications. OBJECTIVES: This commentary aims to deepen the understanding of class-based approaches by elucidating the pivotal role of chemical similarity (structural and biological) in clustering and classification approaches (CCAs). It addresses the dichotomy between general end point-agnostic similarity, often entailing unsupervised analysis, and end point-specific similarity necessitating supervised learning. The goal is to highlight the nuances of these approaches, their applications, and common misuses. DISCUSSION: Understanding similarity is pivotal in toxicological research involving CCAs. The effectiveness of these approaches depends on the right definition and measure of similarity, which varies based on context and objectives of the study. This choice is influenced by how chemical structures are represented and the respective labels indicating biological activity, if applicable. The distinction between unsupervised clustering and supervised classification methods is vital, requiring the use of end point-agnostic vs. end point-specific similarity definition. Separate use or combination of these methods requires careful consideration to prevent bias and ensure relevance for the goal of the study. Unsupervised methods use end point-agnostic similarity measures to uncover general structural patterns and relationships, aiding hypothesis generation and facilitating exploration of datasets without the need for predefined labels or explicit guidance. Conversely, supervised techniques demand end point-specific similarity to group chemicals into predefined classes or to train classification models, allowing accurate predictions for new chemicals. Misuse can arise when unsupervised methods are applied to end point-specific contexts, like analog selection in read-across, leading to erroneous conclusions. This commentary provides insights into the significance of similarity and its role in supervised classification and unsupervised clustering approaches. https://doi.org/10.1289/EHP14001.

Identifying assessment criteria for in vitro studies: a method and item bank.

To support the development of appraisal tools for assessing the quality of in vitro studies, we developed a method for literature-based discovery of study assessment criteria, used the method to create an item bank of assessment criteria of potential relevance to in vitro studies, and analyzed the item bank to discern and critique current approaches for appraisal of in vitro studies. We searched four research indexes and included any document that identified itself as an appraisal tool for in vitro studies, was a systematic review that included a critical appraisal step, or was a reporting checklist for in vitro studies. We abstracted, normalized, and categorized all criteria applied by the included appraisal tools to create an "item bank" database of issues relevant to the assessment of in vitro studies. The resulting item bank consists of 676 unique appraisal concepts from 67 appraisal tools. We believe this item bank is the single most comprehensive resource of its type to date, should be of high utility for future tool development exercises, and provides a robust methodology for grounding tool development in the existing literature. While we set out to develop an item bank specifically targeting in vitro studies, we found that many of the assessment concepts we discovered are readily applicable to other study designs. Item banks can be of significant value as a resource; however, there are important challenges in developing, maintaining, and extending them of which researchers should be aware.

A tool to assess risk of bias in non-randomized follow-up studies of exposure effects (ROBINS-E).

BACKGROUND: Observational epidemiologic studies provide critical data for the evaluation of the potential effects of environmental, occupational and behavioural exposures on human health. Systematic reviews of these studies play a key role in informing policy and practice. Systematic reviews should incorporate assessments of the risk of bias in results of the included studies. OBJECTIVE: To develop a new tool, Risk Of Bias In Non-randomized Studies - of Exposures (ROBINS-E) to assess risk of bias in estimates from cohort studies of the causal effect of an exposure on an outcome. METHODS AND RESULTS: ROBINS-E was developed by a large group of researchers from diverse research and public health disciplines through a series of working groups, in-person meetings and pilot testing phases. The tool aims to assess the risk of bias in a specific result (exposure effect estimate) from an individual observational study that examines the effect of an exposure on an outcome. A series of preliminary considerations informs the core ROBINS-E assessment, including details of the result being assessed and the causal effect being estimated. The assessment addresses bias within seven domains, through a series of 'signalling questions'. Domain-level judgements about risk of bias are derived from the answers to these questions, then combined to produce an overall risk of bias judgement for the result, together with judgements about the direction of bias. CONCLUSION: ROBINS-E provides a standardized framework for examining potential biases in results from cohort studies. Future work will produce variants of the tool for other epidemiologic study designs (e.g. case-control studies). We believe that ROBINS-E represents an important development in the integration of exposure assessment, evidence synthesis and causal inference.

I-REFF diagrams: enhancing transparency in systematic review through interactive reference flow diagrams.

Systematic review methods are recognized for their rigor and transparency and are widely adapted to frameworks that cover review types such as systematic reviews, scoping reviews, and systematic evidence maps. Reporting guidelines help promote better systematic review practices and detailed documentation of the review process for different types of health research (e.g., PRISMA-Preferred Reporting Items for Systematic Reviews and Meta-Analyses; CONSORT-Consolidated Standards of Reporting Trials; and STROBE-Strengthening the Reporting of Observational Studies in Epidemiology). Transparency in the systematic review process and reporting of results is one of the key advantages of the methods and particularly important for hazard and risk assessments due to the high level of scrutiny these reviews face from scientific, political, and public communities. Data visualizations are important to clearly convey information from a review by helping readers perceive, understand, and assess the displayed information easily and quickly. The study flow diagram is a required element of a systematic review and maps out the number of included and excluded records identified, and the reasons for exclusion. Static literature flow diagrams help viewers readily understand the general review methodology and summarize the number of records included or excluded at each stage of the review. However, such diagrams can be time-consuming to develop and maintain during a systematic review or scoping review, and they provide limited summary-level information. We explored how the use of online systematic review tools such as DistillerSR coupled with visualization software such as Tableau can efficiently generate an Interactive REFerence Flow (I-REFF) diagram that is linked to the literature screening data, thus requiring minimal preparation, and resulting in a simplified process for updating the diagram. Furthermore, I-REFF diagrams enhance transparency and traceability by not only summarizing the records in the review but also allowing viewers to follow specific records throughout the review process. We present an example I-REFF diagram and discuss recommendations for key interactive elements to include in these diagrams and how this workflow can improve efficiency and result in an accessible and transparent interactive literature flow diagram without advanced programming.

Protocol for designing INVITES-IN, a tool for assessing the internal validity of in vitro studies.

This protocol describes the design and development of a tool for evaluation of the internal validity of in vitro studies, which is needed to include the data as evidence in systematic reviews and chemical risk assessments. The tool will be designed specifically to be applied to cell culture studies, including, but not restricted to, studies meeting the new approach methodology (NAM) definition. The tool is called INVITES-IN (IN VITro Experimental Studies INternal validity). In this protocol, three of the four studies that will be performed to create the release version of INVITES-IN are described. In the first study, evaluation of existing assessment tools will be combined with focus group discussions to identify how characteristics of the design or conduct of an in vitro study can affect its internal validity. Bias domains and items considered to be of relevance for in vitro studies will be identified. In the second study, group agreement on internal validity domains and items of importance for in vitro studies will be identified via a modified Delphi methodology. In the third study, the draft version of the tool will be created, based on the data on relevance and importance of bias domains and items collected in Studies 1 and 2. A separate protocol will be prepared for the fourth study, which includes the user testing and validation of the tool, and collection of users' experience.