• • The PPR model achieved R2=0.923, Q2LOO=0.836, and Q2ext=0.873, demonstrating high predictive accuracy and robustness for estimating ozone reaction rate constants of aromatic compounds in water.
• • SHAP analysis identified that the presence of dssC (=C<) fragments and chlorine atoms significantly reduces ozone reactivity, guiding molecular design for enhanced degradability.
• • The model was validated using Williams plots to define its applicability domain, ensuring reliable predictions within the chemical space of the training set.
• • TMAP and ARKA analyses provided insights into structural diversity and activity cliffs, aiding in outlier detection and model refinement.