Quantifying Analogue Suitability for SAR-Based Read-Across Toxicological Assessment

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.

Estimation of acceptable daily intake values based on modeling and in vivo mutagenicity of NDSRIs of fluoxetine, duloxetine and atomoxetine

Nitrosamine drug substance related impurities or NDSRIs can be formed if an active pharmaceutical ingredient (API) has an intrinsic secondary amine that can undergo nitrosation. This is a concern as 1) nitrosamines are potentially highly potent carcinogens, 2) secondary amines in API are common, and 3) NDSRIs that might form from such secondary amines will be of unknown carcinogenic potency. Approaches for evaluating NDSRIs include read across, quantum mechanical modeling of reactivity, in vitro mutation data, and transgenic in vivo mutation data. These approaches were used here to assess NDSRIs that could potentially form from the drugs fluoxetine, duloxetine and atomoxetine. Based on a read across informed by modeling of physicochemical properties and mechanistic activation from quantum mechanical modeling, NDSRIs of fluoxetine, duloxetine, and atomoxetine were 10-100-fold less potent compared with highly potent nitrosamines such as NDMA or NDEA. While the NDSRIs were all confirmed to be mutagenic in vitro (Ames assay) and in vivo (TGR) studies, the latter data indicated that the potency of the mutation response was ≥4400 ng/day for all compounds-an order of magnitude higher than published regulatory limits for these NDSRIs. The approaches described herein can be used qualitatively to better categorize NDSRIs with respect to potency and inform whether they are in the ICH M7 (R2) designated Cohort of Concern.

Use of in vitro ADME methods to identify suitable analogs of homosalate and octisalate for use in a read-across safety assessment

Case studies are needed to demonstrate the use of human-relevant New Approach Methodologies in cosmetics ingredient safety assessments. For read-across assessments, it is crucial to compare the target chemical with the most appropriate analog; therefore, reliable analog selection should consider physicochemical properties, bioavailability, metabolism, as well as the bioactivity of potential analogs. To complement in vitro bioactivity assays, we evaluated the suitability of three potential analogs for the UV filters, homosalate and octisalate, according to their in vitro ADME properties. We describe how technical aspects of conducting assays for these highly lipophilic chemicals were addressed and interpreted. There were several properties that were common to all five chemicals: they all had similar stability in gastrointestinal fluids (in which no hydrolysis to salicylic occurred); were not substrates of the P-glycoprotein efflux transporter; were highly protein bound; and were hydrolyzed to salicylic acid (which was also a major metabolite). The main properties differentiating the chemicals were their permeability in Caco-2 cells, plasma stability, clearance in hepatic models, and the extent of hydrolysis to salicylic acid. Cyclohexyl salicylate, octisalate, and homosalate were identified suitable analogs for each other, whereas butyloctyl salicylate exhibited ADME properties that were markedly different, indicating it is unsuitable. Isoamyl salicylate can be a suitable analog with interpretation for octisalate. In conclusion, in vitro ADME properties of five chemicals were measured and used to pair target and potential analogs. This study demonstrates the importance of robust ADME data for the selection of analogs in a read-across safety assessment.

A systematic analysis of read-across within REACH registration dossiers

Read-across is a well-established data-gap filling technique used within analogue or category approaches. Acceptance remains an issue, mainly due to the difficulties of addressing residual uncertainties associated with a read-across prediction and because assessments are expert-driven. Frameworks to develop, assess and document read-across may help reduce variability in read-across results. Data-driven read-across approaches such as Generalised Read-Across (GenRA) include quantification of uncertainties and performance. GenRA also affords opportunities on how New Approach Method (NAM) data can be systematically incorporated to support the read-across hypothesis. Herein, a systematic investigation of differences in expert-driven read-across with data-driven approaches was pursued in terms of establishing scientific confidence in the use of read-across. A dataset of expert-driven read-across assessments that made use of registration data as disseminated in the public International Uniform Chemical Information Database (IUCLID) (version 6) of Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) Study Results were compiled. A dataset of ~5000 read-across cases pertaining to repeated dose and developmental toxicity was extracted and mapped to content within EPA's Distributed Structure Searchable Toxicity database (DSSTox) to retrieve chemical name and structural identification information. Content could be mapped to ~3600 cases which when filtered for unique cases with curated quantitative structure-activity relationship-ready SMILES resulted in 389 target-source analogue pairs. The similarity between target and the source analogues on the basis of different contexts - from structural similarity using chemical fingerprints to metabolic similarity using predicted metabolic information was evaluated. An attempt was also made to quantify the relative contribution each similarity context played relative to the target-source analogue pairs by deriving a model which predicted known analogue pairs. Finally, point of departure values (PODs) were predicted using the GenRA approach underpinned by data extracted from the EPA's Toxicity Values Database (ToxValDB). The GenRA predicted PODs were compared with those reported within the REACH dossiers themselves. This study offers generalisable insights on how read-across is already applied for regulatory submissions and expectations on the levels of similarity necessary to make decisions.

The last resort requirement under REACH: From principle to practice

REACH (Registration, Evaluation, Authorisation and Restriction of Chemicals) is a European Union regulation that aims to protect human health and the environment from the risks posed by chemicals. Article 25 clearly states that: "[i]n order to avoid animal testing, testing on vertebrate animals for the purposes of this Regulation shall be undertaken only as a last resort." In practice, however, the standard information requirements under REACH are still primarily filled using animal studies. This paper presents examples illustrating that animal testing is not always undertaken only as a last resort. Six over-arching issues have been identified which contribute to this: (1) non-acceptance of existing animal or non-animal data, (2) non-acceptance of read-across, (3) inflexible administrative processes, (4) redundancy of testing, (5) testing despite animal welfare concerns and (6) testing for cosmetic-only ingredients. We, members of the Animal-Free Safety Assessment (AFSA) Collaboration, who work together to accelerate the global adoption of non-animal approaches for chemical safety assessment, herein propose several recommendations intended to aid the European Commission, the European Chemicals Agency and registrants to protect human health and the environment while avoiding unnecessary animal tests - truly upholding the last resort requirement in REACH.

Refine and Strengthen SAR-Based Read-Across by Considering Bioactivation and Modes of Action

Structure-activity relationship (SAR)-based read-across is an important and effective method to establish the safety of a data-poor target chemical (structure of interest (SOI)) using hazard data from structurally similar source chemicals (analogues). Many methods use quantitative similarity scores to evaluate the structural similarity for searching and selecting analogues as well as for evaluating analogue suitability. However, studies suggest that read-across based purely on structural similarity cannot accurately predict the toxicity of an SOI. As mechanistic data become available, we gain a greater understanding of the mode of action (MOA), the relationship between structures and metabolism/bioactivation pathways, and the existence of "activity cliffs" in chemical chain length, which can improve the analogue rating process. For this purpose, the current work identifies a series of classes of chemicals where a small change at a key position can result in a significant change in metabolism and bioactivation pathways and may eventually result in significant changes in chemical toxicity that have a big impact on the suitability of analogues for read-across. Additionally, a series of SAR-based read-across case studies are presented, which cover a variety of chemical classes that commonly link to different toxic endpoints. The case study results indicate that SAR-based read-across can be refined and strengthened by considering MOAs or proposed reactive metabolite formation pathways, which can improve the overall accuracy, consistency, transparency, and confidence in evaluating analogue suitability.

Maximizing use of existing carcinogenicity data to support acceptable intake levels for mutagenic impurities in pharmaceuticals: Learnings from N-nitrosamine case studies

The unexpected finding of N-nitrosamine (NA) impurities in many pharmaceutical products raised significant challenges for industry and regulators. In addition to well-studied small molecular weight NAs, many of which are potent rodent carcinogens, novel NAs associated with active pharmaceutical ingredients have been found, many of which have limited or no safety data. A tiered approach to establishing Acceptable Intake (AI) limits for NA impurities has been established using chemical-specific data, read-across, or a class-specific TTC limit. There are ∼140 NAs with some rodent carcinogenicity data, but much of it is older and does not meet current guidelines for what constitutes a 'robust' bioassay. Nevertheless, these data are an important source of information to ensure the best science is used for assessing NA impurities and assuring consumer safety while minimizing impact that can lead to drug shortages. We present several strategies to maximize the use of imperfect data including using a lower confidence limit on a rodent TD50, and leveraging data from multiple NAs. Information on the chemical structure known to impact potency can also support development of an AI or potentially conclude that a particular NA does not fall in the cohort of concern for potent carcinogenicity.

High Throughput Read-Across for Screening a Large Inventory of Related Structures by Balancing Artificial Intelligence/Machine Learning and Human Knowledge

Read-across is an in silico method applied in chemical risk assessment for data-poor chemicals. The read-across outcomes for repeated-dose toxicity end points include the no-observed-adverse-effect level (NOAEL) and estimated uncertainty for a particular category of effects. We have previously developed a new paradigm for estimating NOAELs based on chemoinformatics analysis and experimental study qualities from selected analogues, not relying on quantitative structure-activity relationships (QSARs) or rule-based SAR systems, which are not well-suited to end points for which the underpinning data are weakly grounded in specific chemical-biological interactions. The central hypothesis of this approach is that similar compounds have similar toxicity profiles and, hence, similar NOAEL values. Analogue quality (AQ) quantifies the suitability of an analogue candidate for reading across to the target by considering similarity from structure, physicochemical, ADME (absorption, distribution, metabolism, excretion), and biological perspectives. Biological similarity is based on experimental data; assay vectors derived from aggregations of ToxCast/Tox21 data are used to derive machine learning (ML) hybrid rules that serve as biological fingerprints to capture target-analogue similarity relevant to specific effects of interest, for example, hormone receptors (ER/AR/THR). Once one or more analogues have been qualified for read-across, a decision theory approach is used to estimate confidence bounds for the NOAEL of the target. The confidence interval is dramatically narrowed when analogues are constrained to biologically related profiles. Although this read-across process works well for a single target with several analogues, it can become unmanageable when, for example, screening multiple targets (e.g., virtual screening library) or handling a parent compound having numerous metabolites. To this end, we have established a digitalized framework to enable the assessment of a large number of substances, while still allowing for human decisions for filtering and prioritization. This workflow was developed and validated through a use case of a large set of bisphenols and their metabolites.