Tab. 2:

Questions to assist in the identification of potential sources of uncertainty that may impair RAx prediction (Schultz et al., 2019)

1 The context of, and relevance to, the regulatory use of the read-across prediction as defined by appropriate problem formulation	– Is the regulatory purpose of the read-across prediction clearly defined? – Is the acceptable level or degree of uncertainty for the stated purpose defined? – Is the stated acceptable level or degree of uncertainty appropriate for the stated regulatory purpose?
2 Type of category/group including the definition of the applicability domain	– Is the read-across approach (e.g., analogue or category) clearly reported? – Are the target and source chemicals clearly identified? – Is the applicability domain of the analogue or category defined? – Do target and source chemicals fit within the defined applicability domain?
3 The premise or hypothesis of the read-across	– Is the hypothesis on which the read-across is based clearly stated and presented in sufficient detail to be assessed?
4 Mechanistic plausibility including completeness of the understanding of the MoA or AOP	– How clearly does the hypothesis state the chemical and biological mechanisms underpinning the toxic effect being read across? – Is there sufficient experimental information provided to support the proposed chemical and toxicological mechanisms? – How extensively does the experimental information provided support the mechanistic plausibility and/or the AOP or MoA on which the read-across is based?
5 Similarity in chemistry	– Are the chemical structures (i.e., 2D structure, isomers, SMILES and molecular formula) reported for the derivatives used in the read-across? – Are the dissimilarities in chemical structure reported, and are they toxicologically relevant? – Are the relevant molecular and physicochemical properties (e.g., molecular size, hydrophobicity, solubility, volatility, degradation, etc.) reported for the derivatives used in the read-across? – Are the dissimilarities in molecular and physicochemical properties reported, and are they toxicologically (or pharmacokinetically) relevant?
6 Toxicodynamic similarity	– Is there sufficient and consistent toxicodynamic information provided to establish similarity in the hazard of the derivatives used in the read-across?
7 Toxicokinetic similarity	– Is there sufficient ADME information provided to establish toxicokinetic similarity for the derivatives used in the read-across? – Are any dissimilarities in ADME properties (and, as appropriate, metabolism/degradation) toxicologically relevant?
8 The quality of the apical endpoint data used to fill the data gap	– Is the performance (e.g., reliability, accuracy, precision, repeatability and reproducibility) of the data read across reported clearly? – Has the quality of the data to be read across been assessed, and are they sufficient to meet the purpose of the exercise, i.e., complete and of sufficient quality?
9 The consistency in the effects and severity of the apical in vivo hazard and their concordance with regards to the intermediate and apical effects and potency data	– Is the qualitative expression of the data reported, and is it consistent among the source chemicals? – Is the potency of the hazard reported, and is it consistent among the source chemicals? – What are the temporal relationships between relevant endpoints? – What are the dose-response relationships between relevant endpoints?
10 Strength or robustness of the supporting data sets	– How extensively are the relevant or key events either empirically measured and/or modelled by appropriate in silico, in chemico and in vitro data? – Is the performance (e.g., reliability, accuracy, precision, repeatability and reproducibility) of the supporting methods adequately reported?
11 The weight-of-evidence (WoE) supporting the prediction	– Is there consistency in the supportive information (e.g., structural alerts) between analogues or within the category? – How many and how large are the dissimilarities in the supporting information (i.e., data gaps)?
12 Documentation and written evidence provided	– Is the read-across prediction adequately documented? – Does the evidence support the hypothesis that the uncertainty is acceptable for the stated purpose (as per Question 1)?