Table 1.

Common features of qAOP models in the scientific literature

Common feature	Description	Criteriaa
Problem formulation	• A qAOP should answer a well-defined question relevant to the AO of interest • The purpose of the model dictates how much mechanistic understanding is required, and the way a qAOP should be developed, validated and used	• Question addressed and/or purpose of modelling • AO studied
Mechanistic knowledge and associated data	• The OECD AOP-Wiki can support the development of a qAOP model to predict an endpoint of interest. Empirical data for model parametrisation, fitting and/or testing can be obtained from the description of KERs published in the AOP-Wiki • Whilst for quantification it is recommended to start with linear AOPs, it should not impede quantification of networks or highly connected KEs/KERs within an AOP network • A qAOP model relies heavily on data: not only bioactivity of a compound/mixtures but also, measurements of effects at relevant doses/concentrations and appropriate time scales including physicochemical properties and molecular descriptors. Data may come from a range of in vivo and in vitro studies specifically designed to test an AOP as a hypothesis and/or retrieved from a variety of sources to assist with this process • Both adjacent and non-adjacent KEs paired as upstream–downstream in a KER should be quantified even though each of them impacts differently on the modelling process, e.g. in the context of Bayesian network modelling. Adjacency refers to whether there are other KEs positioned in between of the linear construction of an AOP or not • Different biological level of organisations should be quantified if this is relevant to the AO of interest and available data allowed	• Presence of the AOP in the OECD AOP-Wiki • Type of AOP: linear or network • Type of chemical model applied to (single chemical(s)/mixtures) • Type of data: in vivo, in vitro, in silico and/or other variables • Dose/concentration–responses • (D/C–R) and time–responses (T–R) • Adjacency of KERs: adjacency and non-adjacency • Biological levels: cellular, tissue, organ, organism, population
Quantitative approach	• The modelling approaches can vary from being probabilistic to deterministic • The mathematical expression can take various forms including linear regressions and ordinary differential equations resulting in different graphical shapes, e.g. linear, sigmoidal, Gaussian-type plots	• Type of quantitative approach
Regulatory applicability	• A qAOP model should imply various applications to regulatory decision-making and acceptance	• Human health/ecological risk assessment
Additional considerations	• These considerations can influence the regulatory approval, reduce the uncertainties, and extend the applicability domain of the predictions of a qAOP model • It is not mandatory that the test methods used (models and measured endpoints) are adopted/validated following national/international guidelines. However, they should be performed in a quality-controlled environment where relevance of the model is proved based on scientific rationale and reproducibility of data • Even though none of the definitions identified referred to uncertainty and sensitivity analysis, this aspect should be considered as well for its value in validating the predictions of a qAOP model while giving confidence in its further applications	• Cross species extrapolation • Modulating factors • Positive/negative feedback loops • Compensatory mechanisms • Test method adopted/validated • Kinetics • Exposure assessment • Uncertainty evaluation • Sensitivity analysis • Availability: open access or not

^aThe criteria were used to characterise available qAOP models