Practical Guide 5: How to report (Q)SARs - ECHA - Europa

A reference to a well-documented model or a (Q)SAR Model Reporting Form (QMRF),
attached to a dossier, will be beneficial for the dossier submitter and is highly
recommended. See the Guidance on information requirements and chemical safety
assessment, Chapter R.6: (Q)SARs and grouping of chemicals (R.6.1.9.1) for more
details on how to generate in the QMRF. Please note that the JRC (Q)SAR Model
Database (QMDB) is intended to provide information on (Q)SAR models submitted to
JRC for peer review (http://qsardb.jrc.it/qmrf/search_catalogs.jsp).
There is no formal adoption process foreseen for (Q)SAR models; the validity,
applicability and adequacy of (Q)SAR models for regulatory purposes will be assessed
on a case-by-case basis.
NOTE: A valid (Q)SAR model does not necessarily produce a valid prediction. It is
necessary to assess whether the substance falls within the applicability domain of the
(Q)SAR model, that the results are adequate for the purpose of classification and
labelling and/or risk assessment, and that adequate and reliable documentation of the
applied method is provided.
3.3. Does the substance fall within the applicability domain of
the (Q)SAR model?
In order to get a reliable prediction, it is important to verify that the chemical of interest
falls within the applicability domain of the model. The concept of applicability domain
was introduced to assess the probability of a chemical being covered by the (Q)SAR
training set and it is connected to the reliability of the prediction. The applicability of a
model could, for example, be based on the (sub)structures present in the molecules
and/or a range of molecular descriptor or prediction values. The applicability domain can
also be defined by chemicals (or their metabolites/degradation products) having the
same reaction mechanisms: e.g. types of reactivity or interaction with biomolecules.
Sometimes it is not the chemical itself which reacts with biomolecules but its
metabolites/degradation products. Therefore, it is also important to investigate
metabolic/degradation pathways and the reactivity of the metabolites/degradation
products. See the Guidance on information requirements and chemical safety
assessment, Chapter R.6: (Q)SARs and grouping of chemicals (R.6.1.5.3) for more
details.
3.4. Is the model prediction adequate for the purpose of
classification and labelling and/or risk assessment?
For a (Q)SAR prediction to be adequate, it should be not only reliable, but also relevant
for regulatory decision. The adequacy of the model prediction for the purpose of
5