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

2. HOW TO GET STARTED WITH (Q)SARS
The use of (Q)SAR models with software is rather straightforward. However, experience
and a thorough understanding of (Q)SARs will be needed when it comes to the
assessment of whether predictions are reliable and adequate for the purpose of
classification and labelling and/or risk assessment and can thus be used instead of
testing.
2.1. Substance characterisation
The chemical structure needs to be well defined, following the Guidance on identification
and naming of substances under REACH. All individual constituents of multi-constituent
substances should be addressed. The composition of the well defined substances has to
include also known impurities (and additives, if any). For UVCBs, expert judgement is
needed to decide whether structures can be identified which are representative for the
substance. In case of a stable transformation product, it should be identified. A suitable
structural representation for the chemical (SMILES, mol file etc.) is usually required.
Stereochemistry should be taken into account if necessary.
2.2. (Q)SAR models
SAR and QSAR models, collectively referred to as (Q)SARs, are theoretical models that
can be used to predict in a quantitative or qualitative manner the physico-chemical,
biological (e.g. a (eco)toxicological) and environmental fate properties of compounds
from the knowledge of their chemical structure. A SAR is a qualitative relationship that
relates a (sub)structure to the presence or absence of a property or activity of interest. A
QSAR is a mathematical model relating one or more quantitative parameters, which are
derived from the chemical structure, to a quantitative measure of a property or activity. In
order to predict a property of a chemical, the validity of the selected (Q)SAR model
should be assessed, and it should be verified that the chemical falls within the
applicability domain to give a reliable prediction.
2.3. Overall Assessment of adequacy and reporting
In order to generate a reliable and adequate prediction of the presence or absence of a
chemical’s property to fulfil the information requirements under REACH, (i) a (Q)SAR
model should be used whose scientific validity has been established and (ii) the
substance should fall within the applicability domain of the (Q)SAR model, (iii) the
2