Model development and validation for human serum proteins- membrane interactions: Membrane fouling applications in ultrafiltration of patients blood

Abstract

The DLVO theory, named after the founders Derjaguin, Landau, Verwey, Overbeek, and its derivatives have been successfully used to predict membrane fouling by macromolecules such as proteins. Despite the theory’s success, the calculations involved require extensive measurements. Furthermore, deviations are observed at certain experimental conditions such as at high ionic strengths. Our study aims to develop a linear regression model to address those limitations and reduce the number of required measurements required and associated multistep calculations related to DLVO. Our developed regression model correlates the pseudo-steady normalized flux to the charges, hydrophobicity, and size/ pore size of the foulant(s) (e.g., human serum protein) and the membrane. To verify the validity of the regression model, we extracted data from various peer-reviewed journal articles. The predicted flux using our developed model agreed with the experimental data within 90% confidence intervals. Furthermore, our developed regression model seems to be less sensitive to high ionic strength levels compared to the DLVO and xDLVO, resulting in better correlations. Overall, the results suggest that the fouling phenomena can be predicted using a simple regression equation with readily measurable parameters, such as human serum protein and membrane properties. To conclude, our model only requires particle and membrane’s zeta potentials, their size/pore size, and their water contact angles (compared to water contact angles, glycerol contact angles, and diiodomethane contact angles required for xDLVO) to predict fouling behavior.