Comparative study of (Q)SAR models for predicting the environmental fate of cosmetic ingredients: Persistence, bioaccumulation, and mobility

• Qualitative and quantitative analyses were performed, either within or without the applicability domain (AD). • As a general rule, qualitative predictions, as classified by the REACH and CLP regulatory criteria, are more reliable than quantitative predictions based on correlation and the Applicability Domain (AD) plays an important role in evaluating the reliability of a (Q)SAR models. • For predicting persistence of cosmetic ingredients, the Ready Biodegradability IRFMN model (VEGA), Leadscope model (Danish QSAR Model) and BIOWIN model (EPISUITE) showed the relevant results. • For predicting bioaccumulation of cosmetic ingredients, the ALogP (VEGA), the ADMETLab 3.0 and the KOWWIN (EPISUITE) models were found to be the most appropriate for Log Kow prediction while the Arnot-Gobas (VEGA) and KNN-Read Across (VEGA) models were best for BCF prediction. • For the mobility assessment of cosmetic ingredients, the OPERA v. 1.0.1 and the KOCWIN-Log Kow estimation models from VEGA were deemed relevant. With increasingly stringent regulatory requirements, the cosmetics industry is confronted with the growing expectation to assess the environmental fate of its ingredients (Persistence, Bioaccumulation, and Mobility). However, this assessment is often hampered by data gaps, a challenge exacerbated by the EU's ban on animal testing. In response to these limitations, in silico predictive tools, such as (Quantitative) Structure-Activity Relationship ((Q)SAR) models, are being promoted to provide essential data for environmental risk assessment. The objective of this study was to identify the most suitable freeware tools for predicting Persistence, Ready Biodegradation, Log Kow, BCF and Log Koc. Models integrated into VEGA, EPI Suite™, T.E.S.T., ADMETLab 3.0, and Danish QSAR Models were selected, and their performances were compared by using cosmetic ingredients dataset. Regarding the persistence property, Ready Biodegradability IRFMN (VEGA), Leadscope model (Danish QSAR Model) and BIOWIN (EPISUITE) models showed the highest performance. Moreover, higher performance for bioaccumulation assessment was observed in the ALogP (VEGA), ADMETLab 3.0 and KOWWIN (EPISUITE) models for the Log Kow parameter, as well as in Arnot-Gobas (VEGA) and KNN-Read Across (VEGA) models for BCF prediction. Finally, VEGA's OPERA and KOCWIN-Log Kow estimation models were identified as relevant models for predicting mobility. These results suggest that qualitative predictions are more reliable than quantitative ones when evaluated against REACH and CLP regulatory criteria. The study also highlighted the significant role of the Applicability Domain (AD) in assessing the reliability of (Q)SAR models. These (Q)SAR models showed promising results for assessing the environmental fate of cosmetic ingredients; however, further investigations are needed to confirm these findings.