Predictive Quantitative Read-Across Structure-Property Relationship Modeling of the Retention Time (Log

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Retention time in HPLC correlates with lipophilicity, a key factor in ecotoxicity.
• Quantitative Structure-Property Relationship (QSPR) models have limitations in external predictivity.
• Novel modeling approaches are needed for accurate prediction of environmental chemical properties.

Purpose of the Study:

• To develop and validate a quantitative read-across structure-property relationship (q-RASPR) model for predicting pesticide retention times (log tR) using HPLC data.
• To assess the model's external predictivity and interpretability for environmental risk assessment.
• To explore the utility of q-RASPR as a cost-effective alternative to experimental methods for predicting ecotoxicity potential.

Main Methods:

• Utilized a dataset of 823 environmentally significant pesticide residues.
• Employed 0D-2D descriptors and read-across-derived similarity descriptors for model generation.
• Developed a Partial Least Squares (PLS) model and validated it using OECD-recommended internal and external metrics.

Main Results:

• The developed q-RASPR model demonstrated excellent performance: R² = 0.82, Q²LOO = 0.81, Q²F1 = 0.84.
• The model significantly outperformed previously reported QSPR models in external predictivity.
• Lipophilicity was identified as the primary descriptor influencing retention time, with other factors like multiple bonds and graph density also playing significant roles.

Conclusions:

• The q-RASPR model is a robust, externally predictive, and interpretable tool for retention time prediction.
• This methodology offers a cost-effective and efficient alternative to experimental approaches for predicting ecotoxicity potential.
• q-RASPR shows strong potential for transferability and application in environmental risk assessment and chemical safety evaluations.