A technique is presented for the highthroughput screening of ion-exchange displacers. Potential displacers were employed to displace proteins in parallel batch ion-exchange experiments. The percentage of protein displaced from a particular stationary phase was then used as a parameter to rank the displacers. By employing this technique, a large number of molecules possessing a range of affinities and properties could be rapidly evaluated. This data was then used together with traditional and electron density-based transferable atom equivalent (TAE) molecular descriptors computed for the displacer molecules to produce quantitative structure-efficacy relationship (QSER) models using a genetic algorithm/partial least squares (GA/PLS) regression approach. The QSER models were generated using a portion of the protein-displacement data, with the remainder serving as a test set. Descriptor selection and model building was accomplished using a genetic algorithm/partial least squares approach. The resulting models were found to have high-correlation coefficients and could be used to accurately predict the behavior of molecules not included in the training set. In addition, the models were employed to examine a virtual library of displacers based on modifications of neomycin to provide further insight into displacer design. The results presented here indicate that it may be possible to design displacers that can dramatically improve the effective selectivity of ion-exchange chromatographic materials. © 2002 Wiley Periodicals, Inc.