A multi-scale modeling approach is presented for the a priori prediction of isotherm parameters and preparative chromatographic behavior in ion-exchange and hydrophobic interaction systems. Predictive QSPR models were generated for isotherm parameters using novel pH sensitive molecular descriptors and SVM learning algorithms. The important molecular property descriptors were examined to provide insight into selectivity in these systems. The predicted isotherm parameters were employed in column chromatography models to successfully predict column performance under a range of operating conditions (salt, pH, mode, resins). SVM classification models were also employed in concert with high throughput experimentation to develop highly selective displacers for the separation of difficult protein mixtures. These new classes of displacers enable the separation of similar proteins using low efficiency or even batch chromatographic systems. These developments represent the state-of-the-art in structure-property modeling as applied to chromatography and can provide insight into the nature of selectivity in various chromatographic systems.