A new design strategy with stochastic optimization on the preparation of magnetite cross-linked tyrosinase aggregates (MCLTA). (December 2020)

• Record Type: Journal Article
• Authors: Polatoğlu, İlker; Aydin, Levent
• Is Part Of: Process biochemistry. Volume 99(2020)
• Page Start: 131
• Page End: 138
• Publication Date: 2020-12
• Subjects: Cross-linked enzyme aggregates -- Magnetite -- Neuro regression modelling -- Optimization; Biochemical engineering -- Periodicals; Biotechnology -- Periodicals; Biochemistry -- periodicals; Biotechnology -- periodicals; Chemical Engineering -- periodicals; Génie biochimique -- Périodiques; Biotechnologie -- Périodiques; Biochemical engineering; Biotechnology; Periodicals; 660.63
• DOI: 10.1016/j.procbio.2020.08.019
• Languages: English
• ISSNs: 1359-5113

Abstract

Graphical abstract: Highlights: Optimal design of magnetite cross-linked tyrosinase aggregates was firstly presented. Multiple nonlinear neuro-regression analysis was performed. Realistic functional structures were obtained by examining the models. Deceptive results could be eliminated by this optimization approach. This novel approach is feasible for another modeling-design-optimization problems. Abstract: In this study, a new design strategy with a systematic optimization process is proposed for the preparation of magnetite cross-linked tyrosinase aggregates (MCLTA) by using the concentration of magnetite nanoparticle, glutaraldehyde and tyrosinase enzyme as design variables. A comprehensive study on multiple non-linear neuro-regression analysis has been performed as a compelling alternative to the insufficient approaches on modeling-design-optimization of MCLTA. For this aim, the experimental process has been modeled with 13 candidate functional structures by using a hybrid method to test the accuracy of their predictions. R 2 training, R 2 testing values, and boundedness of the functions have been checked to reveal the realistic ones. Then four different design approaches in terms of three distinct scenarios have been used to optimize the process. The results show that, all models define the process well, depending on R 2 training . However, only five and nine models are appropriate based on R 2 testing for the first use activity and residual activity, respectively. On the other hand, depending on to be a realistic value, model TON best describes the "first use activity, " while the best one is FONT for residual activity. It is also concluded that the scenario types and selection of constraints for design variables affect the optimization results.