Current experimental, statistical, and mechanistic approaches to optimizing biomolecule separations in aqueous two-phase systems

Abstract

Aqueous two-phase systems (ATPS) have been used to purify a range of biomolecules, including small molecules, monoclonal antibodies, viruses, and whole cells. They are known for selective separations, creating a stabilizing, low-shear environment, and high yields. Recently, as biomanufacturing attempts to adopt continuous processing, attention has shifted to ATPS for its ability to operate fully continuously while incurring lower costs than many chromatographic methods. But despite 60 years of exploration and development, the complex network of interlinked driving forces controlling these separations has prevented robust development of process understanding, and most ATPS separations are still optimized using slow and costly manual screening methods. As a result, industry has been unwilling to adopt ATPS. Fortunately, a growing body of literature is developing statistical and mechanistic models of ATPS to predict liquid-liquid equilibria and separations with reduced experimental burden. This review surveys the application of these models to ATPS, comparing their progress and potential to promote rapid development of bioseparations in the near and long term. The discussion evaluates the adaptability of statistical tools, like response surface methodology and artificial neural networks, and contrasts it with the process understanding generated through application of semi-empirical thermodynamic models. Strategies are explored to automate optimization of separations for new biomolecules using these models to create artificial data. By understanding the landscape of models applied to ATPS, this review will start a discussion about bringing this technology closer to commercialization and enabling continuous processing on a broader scale.