How can we address the challenges of global equivalence assessments for pesticide active ingredients? Unveiling a comprehensive strategy for (Q)SAR prediction and toxicological evaluation.

What is (Q)SAR?

Namely, the Quantitative Structure–Activity Relationship. In Chinese, it is referred to as “(Quantitative) Structure–Activity Relationship.” It is a method for investigating the relationship between the structure of a chemical substance (or structure-related properties) and its biological activity or physicochemical characteristics, and can be used to predict the toxicity of chemical substances.

How important are (Q)SARs in the equivalence assessment of active ingredients in pesticides?

Generic‑equivalence certification is a streamlined pathway for non‑patented pesticide active ingredients to enter markets in countries and regions worldwide.

Today, a two‑stage active‑ingredient equivalence assessment procedure is widely adopted worldwide to determine whether a new source (a newly registered active ingredient) is chemically and toxicologically equivalent to the reference source (an already registered active ingredient). In this two‑stage process, the first stage is a chemical evaluation, primarily involving the qualitative and quantitative characterization of the structures and levels of various impurities in the new source. If the first‑stage chemical assessment is not satisfactory—i.e., if the new source contains novel impurities or exhibits elevated levels of existing impurities compared with the reference source—the second stage, a toxicological evaluation, must be undertaken to ensure that the new source and the reference source elicit equivalent toxicological effects.

Countries and regions including the European Union, Australia, and Brazil have incorporated (Q)SAR predictions into the second‑stage toxicological assessment, with the primary objective of determining whether impurities from new sources pose potential health or ecological hazards. If all impurity (Q)SAR predictions are favorable, and when supplemented by requisite toxicological studies—such as in vitro genotoxicity assays and acute toxicity tests—or by hazard‑based classification and risk‑assessment approaches, it is possible to adequately demonstrate that the toxicity of the new source is equivalent to that of the reference‑source active ingredient. Consequently, integrating (Q)SAR predictions into equivalence assessments not only ensures a more comprehensive toxicological evaluation but also helps avoid costly toxicological testing and shortens the registration timeline.

(Q)SAR Principles and Methods

The fundamental assumption of (Q)SAR is that molecules with similar structures exhibit similar properties. By definition, (Q)SAR establishes a quantitative or qualitative relationship between molecular structure and biological activity; in other words, the goal is to use the developed model to predict the effects of an unknown compound directly from its molecular structure. This approach offers the advantage of obviating the need for traditional experimental assays, which are time-consuming and costly.

Generally speaking, the development of (Q)SAR models comprises the following five steps:

(Q)SAR Requirements in Global Pesticide Regulations

European Union

The requirements for (Q)SAR reporting vary across EU member states; however, the specific requirements for (Q)SAR models can be broadly categorized into rule‑based and statistic‑based approaches. In the assessment report, (Q)SAR predictions must be interpreted and justified from a toxicological perspective, and impurities should undergo appropriate hazard classification or risk assessment. Additionally, relevant reference data from active substance studies—such as the acute toxicity battery tests and the Ames test—may be used to provide supplementary evidence for determining whether an impurity is of concern. Particular attention should also be paid to genotoxicity and endocrine‑disrupting effects.

Brazil

In general, at least three distinct QSAR software tools (three expert systems) should be employed. The predictions generated by these (Q)SAR models must cover essential endpoints, such as teratogenic, carcinogenic, and mutagenic effects; where appropriate, additional endpoints—such as endocrine-disrupting effects—should also be included. Whenever feasible, it is preferable to predict all relevant endpoints across the (Q)SAR models. Furthermore, the (Q)SAR prediction results should be interpreted and substantiated from a toxicological perspective.

Australia

Australia requires the submission of both rule-based and statistic-based model predictions; however, unlike the European Union, Australia places particular emphasis on the assessment of genotoxicity and carcinogenicity and mandates that (Q)SAR prediction results be interpreted and justified from a toxicological perspective. When both types of predictions are deemed satisfactory, toxicological studies such as the Ames assay—used to characterize impurity toxicity—may be waived.

Russia

(Q)SAR models must be employed to predict the requisite toxicological endpoints, with particular emphasis on genotoxicity—especially data points that reflect mutagenic potential—though no specific requirements are imposed regarding the type of (Q)SAR model. If the (Q)SAR predictions are deemed unsatisfactory, the regulatory authorities may mandate the conduct of appropriate Ames assays directly on the impurity.

Mexico

(Q)SAR prediction nodes are required to provide only predictions of oral acute toxicity (LD50); other endpoints such as teratogenicity, carcinogenicity, and mutagenicity are neither mandated by Mexican authorities nor deemed necessary under appropriate conditions, though the possibility of their inclusion cannot be ruled out. The number of predictive models employed should be determined on the basis of the results: for instance, if a single model yields unsatisfactory predictions, one or more additional models should be submitted to characterize the substance until sufficient (Q)SAR evidence is available to address the issue. It is recommended that companies submit at least two model‑derived predictions; however, whether these meet the regulatory requirements must be assessed comprehensively in light of toxicological principles. Finally, the resulting LD50 prediction must be further evaluated using the FAO methodology to determine whether the toxicity is equivalent to that of the reference source.

An Analysis of the Challenges in (Q)SAR Applications

Differences between Rule-based Models and Statistic-based Models

Rule-based models establish relationships between structural features and effects based on expert knowledge, primarily through the identification of structural alerts. Statistic-based models, on the other hand, derive these relationships using statistical techniques; common approaches include linear regression, k-nearest neighbors (KNN), random forests, support vector machines (SVM), and neural networks.

Format Requirements for (Q)SAR Prediction Reports

In the current global assessment of the equivalence of active ingredients in pesticides, (Q)SAR model reports are not required to submit QMRF- and QPRF‑format reports, as mandated under REACH. However, these specific formats—prepared in accordance with the OECD’s five‑criteria for (Q)SAR—are highly versatile and can meet the (Q)SAR reporting requirements of virtually all countries and regions. QMRF: (Q)SAR Model Reporting Format QPRF: (Q)SAR Prediction Reporting Format For prediction results generated by rule‑based models that rely on structural alerts, alternative reporting formats may also be used, provided that the model outcomes are thoroughly validated and supported by explicit toxicological justification.

Why do regulatory agencies worldwide place such a strong emphasis on genotoxicity?

In the equivalence assessment of active pharmaceutical ingredients, genotoxicity associated with impurities is almost invariably a key focus of regulatory evaluations worldwide. The primary reason is that genotoxic substances are considered to lack a threshold of toxicity; put more loosely, any exposure to a genotoxic compound may potentially cause direct DNA damage in humans, thereby substantially increasing the risk of cancer. Accordingly, even when impurity levels in APIs from different sources are extremely low, their genotoxic effects must be thoroughly assessed.

Further toxicological assessment

Toxicology Data Query

Using the Weight of Evidence approach, relevant toxicological data are retrieved through a database system, and the validity and reliability of all such data are assessed to determine the hazard values of impurities or target compounds.

Read-Across

Cross-referencing involves predicting the same node for another compound (or multiple compounds) with similar properties, based on the node information of one (or more) compound, thereby serving as a substitute for test data. For impurities or target compounds, cross-referencing is primarily conducted using expert knowledge, focusing on three aspects: structural similarity, metabolic similarity, and trend analysis.

Toxicological Threshold of Concern (TTC)

The TTC defines the acceptable intake level for an untested compound when the risk of carcinogenic or other toxic effects is negligible. The method used to determine the TTC is generally considered highly conservative: it involves a simple linear extrapolation of the dose associated with a 50% tumor incidence (TD50) to a one‑in‑a‑million incidence, based on TD50 data derived from the most sensitive species and from the tumor sites most responsive to induction.

Quantitative Toxicity Threshold Calculation

To further establish a reasonable and reliable limit for impurities in active pharmaceutical ingredients, it is necessary to determine the quantitative toxicity threshold of the impurities themselves. This can be achieved by consulting toxicological data, conducting cross-referencing, or applying the Threshold of Toxicological Concern (TTC) to quantitatively characterize the impurities’ toxicity. Subsequently, relevant uncertainties should be conservatively accounted for, enabling the calculation of a Permitted Daily Exposure (PDE/ADE) or an Acceptable Intake (AI) that serves as the basis for setting impurity limits.

Impurity Limit Control

Impurity limit control is based on the principle of risk assessment, comparing the hazard quotient of an impurity—derived from its quantitative toxicity threshold—with its potential exposure level—determined by the ADI/AOEL and accounting for the impurity’s concentration—to determine whether the impurity’s concentration falls within the established limits.

Assessment of Endocrine Disruption Effects

Based on the European Union’s guidelines for the assessment of endocrine disruption in pesticides and biocides, molecular docking and (Q)SAR predictions may be employed in Level 1 and Level 2 tests to evaluate whether a target compound exhibits potential endocrine-disrupting effects or to elucidate the underlying mechanisms of action.

Methods for the Determination of Skin Sensitization (OECD TG 497)

In June 2021, the OECD TG 497 was published, introducing for the first time a tiered approach to skin sensitization assessment. This tiered approach specifies three in vitro test methods and two (Q)SAR software tools that are relevant to key events (KEs) within the skin sensitization adverse outcome pathway (AOP), and outlines a standardized workflow for evaluating the skin sensitization potential of target compounds.

Computational Toxicology Assessment in New Pesticide Development

For newly discovered bioactive compounds in pesticide research and development, corresponding computational toxicology assessments are conducted to predict whether these compounds or their molecular fragments may exhibit potential toxicological effects, thereby optimizing molecular design and accelerating the development and market launch of new pesticides.

(Q)SAR prediction system

In the equivalence assessment of active ingredients in pesticides, REACH247 can provide clients with a range of up to a dozen (Q)SAR software‑based predictions and expert evaluations. Furthermore, to further enhance the validity and reliability of prediction outcomes, REACH247 has recently launched Derek Nexus and Sarah Nexus—world‑leading (Q)SAR prediction systems—enabling comprehensive, higher‑quality assessments of toxicological endpoints.

Derek Nexus Introduction – Rule-based Models

Derek Nexus is a (Q)SAR-based prediction system recommended by the European Union’s pesticide equivalence assessment and by pesticide regulatory authorities in other countries, such as Brazil, and is also widely adopted by pharmaceutical regulatory agencies including the FDA, EMA, and NMPA.

Features:

• Based on decades of accumulated expertise from toxicology experts

• It covers hundreds of structural alerts and is thoroughly optimized for hard-to-predict toxic structures such as aromatic amines.

• Extensive toxicological endpoints: genotoxicity, skin sensitization, carcinogenicity, reproductive/developmental toxicity, neurotoxicity, and others.

• Highly transparent; all toxicological assessment criteria and literature sources are publicly accessible.

Sarah Nexus Introduction – Statistic-Based Models

Sarah Nexus employs a novel hierarchical clustering model, the SOHN method, to learn from fragmented molecular structures in training datasets, thereby enabling sensitive identification of structure–activity relationships associated with genotoxicity in target molecules.

Features:

• Highly sensitive predictive methods, when used in conjunction with Derek Nexus, can significantly enhance the effectiveness and reliability of predictions.

• Based on reliable, large-scale Ames mutagenicity assay data

• The training set has undergone thorough reliability validation and can serve as a trustworthy data source.

• Highly transparent, with a complete reasoning process.

Source: Rui'ou Technology