Human immunoglobulin G adsorption in hydrophobic ligands: equilibrium data, isotherm modelling and prediction using artificial neural networks

Author(s)

Shinku, Caroline A. ; Martins, Tiago D. ; Bresolin, Igor T. L. ; Bresolin, Iara R. A. P.

• Document type: Journal article
• Source: CHEMICAL PAPERS; v. 77, n. 2, p. 17-pg., 2022-10-26.

Abstract

This work aimed to evaluate the adsorption of human Immunoglobulin G in the hydrophobic interaction adsorbents: Phenyl Sepharose 6 Fast Flow High Sub (High Sub) and Phenyl Sepharose 6 Fast Flow Low Sub (Low Sub). Kinetic tests were performed at 298 K. Equilibrium was evaluated at 277 K, 288 K, 298 K and 310 K. Data modelling was performed by using isotherm models and artificial neural networks (ANN). The highest adsorption capacity was 56.30 +/- 6.50 mg mL(-1) for the experiment performed using 1.0 mol L-1 ammonium sulphate buffer and High Sub adsorbent at 288 K. In this condition of highest ligand density and salt concentration were found the best results regarding the behaviour of the isotherms. When comparing the results obtained in different temperatures, there was practically no difference, considering the standard deviation, between the values of maximum capacity adsorption. Also, if either salt concentration or ligands density was increased, higher values of adsorption capacity were obtained. All the dissociation constants showed to be between intermediate and strong. The best modelling result was obtained by the ANNs, which accurately represented the experimental isotherms with mean percentage error equal to 0.68%, and R-2 above 0.999. It had 2 neurons in the hidden layer, and it was trained with the Levenberg-Marquardt algorithm. The ANN modelling was effective in predicting the adsorption isotherms, as it proved to be able to interpolate its behaviour in the studied temperature range. Furthermore, it outgrew possible errors arising from experimental procedures and/or from adsorption phenomenology. [GRAPHICS] . (AU)

• FAPESP's process: 20/06797-1 - Adsorption of human immunoglobulin G in hydrophobic interaction ligands: evaluation of equilibria data by artificial neural networks
• Grantee: Caroline Akina Shinku
• Support Opportunities: Scholarships in Brazil - Scientific Initiation