Quantitative-structure based predictions of displacer and protein affinities in chromatographic systems

The field of biotechnology has grown drastically in the last two decades with chromatography being one of the main processes employed for the purification of bioproducts. Biotechnology as well as pharmaceutical companies require time reduction in process development. To achieve this goal, generating predictive quantitative structure based models for given chromatographic systems as well as determining the main interactions that occur are becoming increasingly important.; In this work a novel approach is described for the a-priori prediction of potential displacers, small molecules as well as proteins in ion exchange and hydrophobic interaction systems. Quantitative structure based models employing a genetic algorithm/partial least squares approach were developed using chromatographic data in concert with molecular descriptors based on molecular structure. In this thesis it was shown that with the appropriate data set size, the resulting quantitative structure based models were well correlated and the predictive power of these models was demonstrated with probes not included in the derivation of the models. Virtual libraries were constructed and potential leads were identified for their use in future phases of this new approach.