Details

Autor(en) / Beteiligte

• Wassenaar, Pim N.H.
• Rorije, Emiel
• Janssen, Nicole M.H.
• Peijnenburg, Willie J.G.M.
• Vijver, Martina G.

Titel

Chemical similarity to identify potential Substances of Very High Concern – An effective screening method

Ist Teil von

• Computational toxicology, 2019-11, Vol.12, p.100110, Article 100110

Ort / Verlag

Elsevier B.V

Erscheinungsjahr

2019

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• [Display omitted] •Potential Substances of Very High Concern can be identified by chemical similarity.•High balanced accuracies (≥0.8) were obtained for all SVHC-subgroup models.•Improvement of the ED model by extending the database is considered necessary.•The best performing similarity models can be used for screening and prioritization. There is a strong demand for early stage identification of potential Substances of Very High Concern (SVHC). SVHCs are substances that are classified as carcinogenic, mutagenic or reprotoxic (CMR); persistent, bioaccumulative and toxic (PBT) or very persistent and very bioaccumulative (vPvB); or as substances with an equivalent level of concern, like endocrine disruption (ED). The endeavor to improve the identification of potential SVHCs is also acknowledged by the European Commission, in their long-term vision towards a non-toxic environment. However, it has been shown difficult to identify substances as potentially harmful. With this goal in mind, we have developed a methodology that predicts whether a substance is a potential SVHC based on chemical similarity to chemicals already identified as SVHC. The approach is based on the structural property principle, which states that structurally similar chemicals are likely to have similar properties. We systematically analyzed the predictive performance of 112 similarity measures (i.e. all different combinations of 16 binary fingerprints and 7 similarity coefficients) classifying the substances in the dataset as (potential) SVHC or non-SVHC. The outcomes were analyzed for 546 substances that we collected within the Dutch SVHC database – with identified CMR, PBT/vPvB and/or ED properties – and 411 substances that lack these hazardous properties. The best similarity measures showed a high predictive performance with a balanced accuracy of 85% correct identifications for the whole dataset of SVHC substances, and 80% for CMR, 95% for PBT/vPvB and 99% for ED subgroups. This effective screening methodology showed great potential for early stage identification of potential SVHCs. This model can be applied within regulatory frameworks and safe-by-design trajectories, and hence can contribute to the EU goal of achieving a non-toxic environment.