Anion-exchange resins are widely used for removing anionic micropollutants from aqueous environments, yet predictive

understanding across structurally diverse pollutants remains limited. To address this gap, we present the first combined experimental–

computational QSAR framework for anion-exchange resins that explicitly incorporates a concentration-dependent

descriptor, namely activity degree of the ion (log α). Adsorption isotherms of 26 anionic compounds were systematically

measured on Amberjet™ 4200 at multiple initial concentrations, providing a robust experimental dataset. Two complementary

descriptor sets were employed for model development: (i) empirically derived linear free energy relationship (LFER)

parameters and (ii) in silico-calculated COSMOtherm descriptors. Incorporating log α into both frameworks substantially

improved accuracy, reflecting the critical role of ionic strength and activity effects in adsorption processes. The optimized

models achieved excellent predictive power, with training R2 values > 0.93 and external validation using test set yielding

R2 = 0.938 (SE = 0.193 log units) for the LFER-based model and R2 = 0.953 (SE = 0.150 log units) for the COSMOthermbased

model. Analysis of LFER descriptor contributions revealed that the excess molar refractivity term had a negative

coefficient—suggesting that electronic lone pairs, hydrogen-bond acidity, and volume-related lipophilic effects exert a

repulsive influence on adsorption—whereas polar interaction and hydrogen-bond basicity terms showed positive coefficients,

indicating that these interactions enhance adsorption affinity. Comparative analysis further indicated that COSMOtherm

descriptors more effectively captured electronic and solvation effects, whereas LFER descriptors provided clearer mechanistic

interpretability. This study establishes a versatile framework for predictive evaluation of anionic micropollutant adsorption,

providing mechanistic insights and supporting the preliminary assessment of adsorbent suitability for structurally novel or

data-scarce pollutants.